A multi-wheel pair multi-actuator integrated steering method for heavy-duty vehicles

By establishing a modular design and a multi-wheel-drive vehicle model, combined with an integrated steering control strategy, the problems of trajectory consistency and robustness in complex scenarios of steering systems for multi-axle heavy-duty vehicles were solved, achieving safe and reliable steering control and human-like autonomous driving decision-making.

CN123104773APending Publication Date: 2026-11-17JILIN UNIVERSITY
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202611595896.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-15
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies struggle to effectively address issues such as trajectory consistency, tire nonlinearity, roll instability risk, and robustness in steering systems of multi-axle heavy-duty vehicles with 10 or more axles under complex scenarios. In particular, there are significant differences in turning radii between wheel pairs during low-speed steering and high roll instability risk at high speeds. There is a lack of integrated steering methods to leverage the redundancy advantages of multiple wheel pairs and actuators.

Method used

This paper presents a multi-wheelset, multi-actuator integrated steering method for heavy-duty vehicles. By designing modular heavy-duty vehicle performance specifications, establishing a multi-wheelset vehicle model, and integrating steering control strategies, combined with a multi-wheelset integrated steering evaluation method, it enables safety decision-making with experience memory and human-like cognition.

Benefits of technology

It achieves safe and reliable steering control in complex scenarios, improves the robustness and safety of the steering system, is suitable for heavy-duty vehicles with more than 10 axles, and has human-like autonomous driving capabilities with experience memory and human-like cognition.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123104773A_ABST
    Figure CN123104773A_ABST
Patent Text Reader

Abstract

This invention discloses a multi-wheelset, multi-actuator integrated steering method for heavy-duty vehicles. The method includes the following steps: first, modular design of heavy-duty vehicle performance specifications; second, establishment of a multi-wheelset vehicle model; third, design of a multi-wheelset integrated steering control strategy; and fourth, construction of a multi-wheelset integrated steering evaluation method. Beneficial effects: It provides a quantitative design basis and performance boundary conditions for the steering system design of ultra-multi-axle heavy-duty vehicles with 10 or more axles; it provides a high-fidelity controlled object model for control strategy design; it achieves accurate, stable, and coordinated steering of ultra-multi-axle vehicles under all working conditions; it solves the problems of response lag and insufficient control accuracy of steering actuators in heavy-duty vehicles; it provides a theoretical basis for the optimization of wheelset configuration in multi-wheelset heavy-duty vehicles; and it provides a systematic evaluation tool for the design and verification of heavy-duty vehicle steering systems.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to a steering control technology for heavy-duty vehicles, and more particularly to a multi-wheel pair, multi-actuator integrated steering method for heavy-duty vehicles. Background Technology

[0002] Currently, heavy-duty vehicles are indispensable transportation equipment in fields such as national defense construction, energy development, and large-scale engineering construction, undertaking the transportation tasks of oversized, overweight, and overweight special cargo. With the continuous growth of transportation demand, the number of axles in heavy-duty vehicles continues to increase, and multi-axle heavy-duty vehicles with more than 10 axles have become the norm in the industry. These vehicles typically adopt a design scheme that combines a multi-axle hydraulic suspension system with a multi-mode steering mechanism. The overall length of the vehicle can reach more than 20 meters, and the fully loaded weight exceeds 100 tons. They are typical complex controlled objects with large inertia, slow response, and strong nonlinearity.

[0003] While multi-wheel-pair configurations offer significant advantages in load-bearing capacity and off-road capability, they also present unprecedented challenges to the steering system: the surge in the number of wheelsets leads to highly complex steering kinematic constraints, and the difficulty of kinematic coordination between wheelsets increases exponentially; heavy-duty vehicles have large masses and inertia, resulting in severe inter-axle load transfer during steering and significant tire nonlinearity, making it difficult for traditional open-loop control methods based on Ackermann steering geometry to guarantee trajectory consistency across multiple wheelsets; furthermore, multi-axle vehicles exhibit significant differences in turning radii among wheelsets at low speeds, leading to prominent tire wear issues, while the risk of roll instability increases dramatically during high-speed steering, placing extremely high demands on the robustness and safety of the steering system.

[0004] Currently, significant research achievements and patent technology reserves have been accumulated regarding steering control for multi-axle vehicles. In terms of patents, Chinese patent CN102514619A discloses an all-wheel steering control method for multi-axle vehicles. This method uses angle sensors to detect the front axle wheel angles and calculates the arithmetic mean of the wheel angles on both sides of the front axle, achieving coordinated control of the rear axle's electronically controlled steering. Chinese patent CN116001910A discloses a steering control method for towable heavy-duty vehicles. This method uses angle sensors installed on each wheel set to collect initial steering angle information, generates operating commands based on target sensing information and current vehicle speed, and decomposes them into sub-commands executable by each component. Chinese patent CN114560011A discloses a system and method for improving the safety of heavy-duty multi-axle vehicles when wheel steering is stuck. This method optimizes the distribution of tire forces to each normally steering wheel through secondary programming. While existing technologies have made some progress in steering control of multi-axle vehicles, there is still a lack of an integrated steering method that can be applied to multi-axle heavy-duty vehicles with more than 10 axles, fully utilize the redundancy advantages of multiple wheel pairs and multiple actuators, cover the entire chain of steering intention judgment, trajectory planning, trajectory following and motor control, and is equipped with a systematic evaluation method. Performance improvement and technological breakthroughs are urgently needed. Summary of the Invention

[0005] The main objective of this invention is to address the problems of existing autonomous driving decision-making systems, such as lack of memory capacity, low anthropomorphism, insufficient robustness of safety decisions in complex scenarios, and weak generalization ability.

[0006] Another objective of this invention is to realize a humanoid autonomous driving safety decision-making method based on neural memory that possesses experiential memory, human-like cognition, and verifiable safety features.

[0007] In order to achieve the above objectives and solve the above problems, the present invention provides a multi-wheelset multi-actuator integrated steering method for heavy-duty vehicles.

[0008] The multi-wheel-pair, multi-actuator integrated steering method for heavy-duty vehicles provided by this invention includes the following steps:

[0009] Step 1: Modular heavy-duty vehicle performance specification design;

[0010] The second step is to establish a multi-wheel vehicle model.

[0011] Step 3: Design of multi-wheel integrated steering control strategy;

[0012] Step 4: Construction of a multi-round integrated steering evaluation method.

[0013] The process of designing the performance specifications for modular heavy-duty vehicles in the first step is as follows:

[0014] Step 1: Assessment of the carrying capacity of modular heavy-duty vehicles. Step 1 consists of three parts, as follows:

[0015] Step 1: Calculation of the mapping relationship between the number of wheelsets and the carrying capacity. This step establishes a basic mapping relationship between the number of wheelsets and the carrying capacity of the entire vehicle based on the load-bearing characteristics of multi-wheelset configurations in heavy-duty vehicles. Let the rated carrying capacity of a single wheelset be m. rated With n wheelsets, considering the coupled effect of tire vertical stiffness and road surface unevenness, the actual load-bearing capacity of a single wheelset dynamically changes with vehicle speed V and road surface grade R. A dynamic load-bearing correction factor k is established. v (V, R), then the theoretical maximum carrying capacity M of the whole vehicle total As shown in formula (1):

[0016] (1);

[0017] In the formula, η load Let be the load uneven distribution coefficient, where k is the dynamic load correction factor. v The expression for calculating (V, R) is shown in formula (2):

[0018] (2);

[0019] In the formula, G q (r) represents the displacement power spectral density of road surface unevenness, α R This is the gain coefficient of the tire suspension coupled dynamic load;

[0020] The input signals for this stage are the number of wheelsets n and the rated load capacity m of a single wheelset. rated Vehicle speed V, road surface grade R; the output signal is the theoretical maximum carrying capacity M of the vehicle. total ;

[0021] Section Two: Multi-condition Carrying Capacity Calculation. This section calculates the carrying capacity under various typical operating conditions faced by heavy-duty vehicles in actual operation, including constant speed on a flat road, acceleration on a flat road, starting on an incline, crawling on an incline, and driving on curves. It establishes the maximum permissible carrying mass under constraints of traction force, resistance balance and adhesion, power, and stability. The carrying capacity M for constant speed driving on a flat road is calculated. flat, const As shown in formula (3):

[0022] (3);

[0023] In the formula, μ is the road surface adhesion coefficient, ψ is the static load distribution ratio of the drive axle, and ρ a C is the density of air. D Where A is the air resistance coefficient, f is the frontal area, g is the rolling resistance coefficient, and M is the acceleration due to gravity. The carrying capacity M is the acceleration on a straight road. flat, acc As shown in formula (4):

[0024] (4);

[0025] In the formula, ψ' is the dynamic load coefficient of the drive shaft during acceleration, δ is the rotational mass conversion coefficient, and a is the target acceleration. The carrying capacity M during ramp start-up is... ramp, start As shown in formula (5);

[0026] (5);

[0027] In the formula, F max The maximum driving force of the drive wheels is given by θ, where θ is the road gradient angle, and M represents the carrying capacity for crawling on the slope. ramp, climb As shown in formula (6):

[0028] (6);

[0029] In the formula, P max η is the maximum output power of the engine motor. t For the overall efficiency of the transmission system, the carrying capacity M when cornering curve As shown in formula (7):

[0030] (7);

[0031] In the formula, F tire Δlat represents the maximum allowable vertical load on a single wheel, and Δlat is the percentage increase in wheel load during curves.

[0032] The input signals for this stage are the road surface adhesion coefficient μ, the static load distribution ratio of the drive shaft ψ, and the air density ρ. a Air drag coefficient C D Frontal area A, rolling resistance coefficient f, dynamic load coefficient of drive shaft during acceleration ψ', rotational mass conversion coefficient δ, target acceleration a, maximum driving force of drive wheel F max Road slope angle θ, maximum output power of engine motor P max Overall efficiency η of the transmission system t Maximum allowable vertical load F per wheel tire The wheel load increment ratio Δlat in curves; the output signal is the carrying capacity M for the following driving conditions: constant speed on a straight road, acceleration on a straight road, hill start, hill crawl, and curve driving. flat, const M flat, acc M ramp, start M ramp, climb M curve ;

[0033] Step 3: Determining the carrying capacity boundary and safety margin. This step, based on the carrying capacity calculations for the five operating conditions in Step 2, takes the minimum allowable carrying mass for each operating condition as the lower limit boundary of the vehicle's carrying capacity, and takes the flat static benchmark value as the upper limit boundary, thus determining the safe carrying range of the entire vehicle [M]. min, safe M max, safe The carrying capacity safety margin SM is defined as shown in formula (8):

[0034] (8);

[0035] In the formula, M actual The actual carrying capacity is determined by the SM (Shipment Scale) threshold. When the SM reaches a certain threshold, an overload warning is triggered.

[0036] The input signal for this stage is the carrying capacity M under the following conditions: constant speed on a flat road, acceleration on a flat road, starting on a slope, crawling on a slope, and driving on a curve. flat, const M flat, acc M ramp, start M ramp, climb M curve The output signal represents the safe carrying range of the entire vehicle [M]. min, safe M max, safe Safety margin of carrying capacity (SM);

[0037] Step 2: Modular Heavy-Duty Vehicle Dynamics Assessment. Step 2 consists of three parts, as follows:

[0038] Step 1: Multi-wheel drive power matching calculation. This step calculates the total drive power required for three driving conditions—high-speed cruising, climbing, and acceleration—based on the vehicle's total mass, design maximum speed, and maximum climbing resistance. Then, it allocates the output power of each wheel pair according to the axle load ratio of each drive wheel axle. The total power demand P at the vehicle's steady-state maximum speed is calculated. total, max As shown in formula (9):

[0039] (9);

[0040] In the formula, M v V represents the unloaded mass. max At the maximum vehicle speed, the power distribution P of a single wheelset i As shown in formula (10):

[0041] (10);

[0042] In the formula, F i For vertical load on a single axle;

[0043] The input signal for this stage is the theoretical maximum carrying capacity M of the vehicle. total Unloaded mass M v Maximum speed V max air density ρ a Air drag coefficient C D Frontal area A, rolling resistance coefficient f, vertical load on a single axle F i The output signal is the total power demand P at the vehicle's steady-state maximum speed. total, max Power distribution P of a single wheelset i ;

[0044] Section Two: Acceleration and Climbing Ability Assessment. This section assesses the vehicle's acceleration and climbing ability based on its drive power and curb weight, with acceleration time t... acc The calculation is shown in formula (11):

[0045] (11);

[0046] Maximum stable gradient α max The calculation is shown in formula (12):

[0047] (12);

[0048] In the formula, V c To maintain a stable speed while climbing the hill;

[0049] The input signal for this stage is the theoretical maximum carrying capacity M of the vehicle. total Unloaded mass M0, vehicle speed V, air density ρ a Air drag coefficient C D Frontal area A, rolling resistance coefficient f, rotational mass conversion factor δ, maximum output power of engine / motor P max Overall efficiency η of the transmission system t Stable climbing speed V c The output signal is the acceleration time t. acc Maximum stable gradient α max ;

[0050] Step 3, Comprehensive Evaluation and Output of Dynamic Performance: This step integrates the calculation results from Step 1 and Step 2 to construct a performance evaluation system that includes the maximum vehicle speed V. max Maximum stable gradient α max Acceleration time t acc The dynamic performance evaluation system comprises three core indicators, and a weighted comprehensive score is applied based on five common operating conditions of heavy-duty vehicles. The analytic hierarchy process (AHP) is used, with the weight coefficient for each operating condition set as ω. i Finally, the integrated calculation yields the comprehensive power score S. power As shown in formula (13):

[0051] (13);

[0052] In the formula, λ1, λ2, and λ3 are weighting coefficients;

[0053] The input signal for this stage is the maximum vehicle speed V. max Maximum stable gradient α max Acceleration time t acc Weighting coefficient ω for operating conditions i The output signal is the comprehensive power score S. power ;

[0054] Step 3: Modular heavy-duty vehicle stability assessment. Step 3 consists of three parts, as follows:

[0055] Step 1: Handling Stability Assessment. This step focuses on the steering characteristics of multi-axle heavy-duty vehicles, establishing a handling stability evaluation index system, using understeer angle Us and yaw rate gain G as the main metrics. yr Lateral acceleration gain G la As a core evaluation indicator, for an n-axis vehicle, l is defined as i Let be the longitudinal distance of the i-th axis relative to the sprung center of mass of the entire vehicle, and let l be the distance between the axles in front of the center of mass. i <0, the axle behind the center of mass satisfies l i>0, the first axis is the equivalent front axle, serving as the main steering input axis of the vehicle, and the set of rear axles is defined as Ωr={i|l i For all load-bearing axes and active steering driven axes behind the center of mass, ≥0, 2≤i≤n}, a unique equivalent rear axle is constructed. The understeer angle Us is defined as the equivalent front axle tire slip angle α. f Equivalent rear axle tire slip angle α r The difference between the two and the lateral acceleration a y The ratio of the two values ​​is used to calculate the steady-state yaw rate gain G. yr As shown in formula (14):

[0056] (14);

[0057] In the formula, L eq For the equivalent wheelbase, K Us λ is the understeering coefficient. i Here are the steering weighting coefficients for each axis, and the lateral acceleration gain G. la As shown in formula (15):

[0058] (15);

[0059] The input signal for this stage is the front axle tire slip angle α. f Rear axle tire slip angle α r Lateral acceleration a y Equivalent wheelbase L eq Insufficient turning coefficient K Us Vehicle speed V; output signals are understeer angle Us and yaw rate gain G. yr Lateral acceleration gain G la ;

[0060] Step Two: Braking Longitudinal Stability Assessment. This step addresses the longitudinal stability issue of multi-axle heavy-duty vehicles during braking, establishing evaluation indicators for braking performance and braking directional stability, with the maximum braking deceleration 'a' as the criterion. brake, max Full-load braking distance S brake Axle load transfer braking lock-up boundary LTR brake The maximum braking deceleration a is used as an evaluation metric. brake, max As shown in formula (16):

[0061] (16);

[0062] In the formula, μ brake η is the peak adhesion coefficient of the braking surface. disk The multi-axis braking force distribution balance coefficient is constructed as shown in formula (17):

[0063] (17);

[0064] In the formula, F bi The braking force corresponding to the real-time braking pressure of each axle. The average braking force of all brake axles in the vehicle, and the fully loaded braking distance S. brake As shown in formula (18):

[0065] (18);

[0066] In the formula, t lag v0 is the initial braking velocity, and v0 is the braking system reaction time. Axle load transfer braking lock-up boundary LTR brake As shown in formula (19):

[0067] (19);

[0068] In the formula, ΔF z F represents the dynamic load change during braking, which is equivalent to an increase in load on the front axle and a decrease in load on the rear axle. z, static h represents the average vertical load on the front and rear axles under static full load. c The height of the vehicle's center of gravity when fully loaded;

[0069] The input signal for this stage is the peak adhesion coefficient μ of the braking surface. brake Multi-axis braking force distribution balance coefficient η disk Braking system reaction time t lag Initial braking velocity v0, dynamic load change ΔF during braking process (increased load on the front axle, decreased load on the rear axle) z The height of the vehicle's center of gravity when fully loaded (h) c Static full load, average vertical load F between front and rear axles z, static Equivalent wheelbase L eq The output signal is the maximum braking deceleration a. brake, max Full-load braking distance S brake Axle load transfer braking lock-up boundary LTR brake ;

[0070] Step 3, Comprehensive Evaluation of Braking Performance and Stability: This step integrates the evaluation results from Step 1 and Step 2 to construct a comprehensive vehicle stability score. stab As shown in formula (20):

[0071] (20);

[0072] In the formula, ω1, ω2, ω3, ω4, and ω5 are weighting coefficients, and std is the industry standard reference value;

[0073] The input signal for this stage is the yaw rate gain G. yr Lateral acceleration gain Gla Maximum braking deceleration a brake, max Full-load braking distance S brake Axle load transfer braking lock-up boundary LTR brake The output signal is the overall vehicle stability score. stab ;

[0074] Step 4: Modular heavy-duty vehicle passability assessment. Step 4 consists of three parts, as follows:

[0075] Section 1: Multi-Axle Geometric Passability Assessment for Heavy-Duty Vehicles. This section is based on the multi-span wheelbase structure of n-axle heavy-duty vehicles, and the unique failure modes of mid-section overhang, low-lying areas between multiple axles, and interference from multiple wheel sets. It establishes a multi-axle coupled geometric passability evaluation system adapted to ultra-long vehicle bodies and multi-wheelset arrangements, adding longitudinal passability characteristics within multi-axle sections. Core indicators include minimum ground clearance h. min Approach angle α approach Departure angle α depart Define the longitudinal pass angle α of multi-axis coupling. ramp Instead of using the traditional front and rear single axle constraint method, the weighted solution is based on the maximum span of adjacent axle groups and the minimum ground clearance of each segment, which characterizes the bottoming resistance of multi-span vehicle bodies when passing through uneven roads and continuous potholes, as shown in formula (21):

[0076] (twenty one);

[0077] In the formula, L wb, max This represents the maximum longitudinal distance between adjacent wheelsets.

[0078] The input signals for this stage are the ground clearance of each section of the heavy-duty multi-axle vehicle, the geometric parameters of the front end, the set parameters of the rear end, and the maximum longitudinal distance L between adjacent wheelsets. wb, max The output signal is the minimum ground clearance h. min Approach angle α approach Departure angle α depart Multi-axis coupling longitudinal through angle α ramp ;

[0079] Step Two: Assessment of the Passability of Flexible Supports with Multi-Axle Load Diversion. This step is based on the multi-wheel pair uniform load distribution, load diversion, and multi-point flexible grounding mechanism of ultra-multi-axle heavy-duty vehicles, as well as the issues of significantly reduced single-wheel grounding pressure, significantly reduced road subsidence, and traction adhesion characteristics that are completely different from ordinary vehicles. A passability evaluation model for soft road surface supports adapted to multi-axle heavy-duty vehicles is established, introducing a multi-axle load uniformity correction coefficient η. loadThis study introduces parameters such as the order of the number of axles and the dispersion of loads on each axle to reflect the coupled influence of multi-point pressure distribution and load uniformity of n-axle vehicles on road passability. It also constructs multi-axle coupled traction coefficient TC, multi-axle cooperative traction efficiency TE, and multi-axle heavy-duty fuel utilization index E. f Evaluation system, multiaxial load uniform distribution correction factor η load As shown in formula (22):

[0080] (twenty two);

[0081] In the formula, The average vertical load of all axles of the vehicle is given by the multi-axle coupling traction coefficient TC, as shown in formula (23):

[0082] (twenty three);

[0083] In the formula, F net The total vehicle hook traction force and the multi-axle cooperative traction efficiency TE are shown in formula (24):

[0084] (twenty four);

[0085] In the formula, P out For output power, P in For input power, the multi-axle heavy-duty fuel utilization index E f As shown in formula (25):

[0086] (25);

[0087] The input signal for this stage is the real-time vertical load F of each axle of the vehicle. zi Average vertical load on all axles of the vehicle The traction force F of the whole vehicle hook net Total vehicle weight M total g, Output power P out Input power P in The output signal is the multi-axis load distribution correction coefficient η. load Multi-axle coupling traction coefficient TC, multi-axle cooperative traction efficiency TE, multi-axle heavy-duty fuel utilization index E f ;

[0088] Step 3, Comprehensive Passability Rating: This step establishes a hierarchical comprehensive rating mechanism that combines hard thresholds for geometric constraints with soft evaluation of support performance. This aligns with the actual failure logic of heavy-duty vehicles. First, it determines whether multi-axle geometric bottoming interference exceeds limits, then it integrates the support and passage capacity of soft road surfaces to achieve a high-precision comprehensive passability assessment. A multi-axle geometric passability level R is constructed. geo As shown in formula (26):

[0089] (26);

[0090] In the formula, β1, β2, β3, and β4 are weighting coefficients, and ref is the reference value, which constructs the multi-axis support passability level R. support As shown in formula (27):

[0091] (27);

[0092] In the formula, ζ1, ζ2, and ζ3 are weighting coefficients, and ref is the reference benchmark value. The overall passability rating R is then calculated. mobility As shown in formula (28):

[0093] (28);

[0094] In the formula, λ1 and λ2 are weighting coefficients;

[0095] The input signal for this stage is the minimum ground clearance h. min Approach angle α approach Departure angle α depart Longitudinal passing angle α ramp Traction coefficient TC, traction efficiency TE, fuel utilization index E f The output signal is the geometric passability level R. geo Support passability rating R support Passability Comprehensive Rating mobility .

[0096] The process of building the whole vehicle model in the second step is as follows:

[0097] Step 1: Vehicle dynamics modeling. The model established in this step is applicable to the dynamic response of heavy-duty vehicles with 6 axles or more. A whole vehicle dynamics model with coupling in the longitudinal, lateral, and vertical directions is established. The number of wheelsets is n, and the whole vehicle's degrees of freedom is 6+2n, namely, longitudinal translational degrees of freedom, lateral translational degrees of freedom, vertical translational degrees of freedom, yaw rotational degrees of freedom, roll rotational degrees of freedom, pitch rotational degrees of freedom, and 2n degrees of freedom of the tires. The vehicle dynamics equation set is established as shown in formula (29):

[0098] (29);

[0099] In the formula, M v For the overall vehicle weight, a x a y These are the longitudinal and lateral accelerations, F, respectively. xi F yi F represents the sum of the longitudinal and lateral forces of each wheel on the i-th axis. areo For air resistance, F r R is the rolling resistance, r is the yaw rate, and I is the rolling resistance.z I x I y These are the yaw, roll, and pitch moments of inertia of the entire vehicle, respectively. i Let t be the longitudinal distance from the i-th axis to the centroid. i Let ϕ be the wheelbase of the i-th axis, φ be the roll angle, and c be the wheelbase of the yoke. ϕ k ϕ These are the roll damping coefficient and pitch stiffness coefficient, respectively, θ p c is the pitch angle. θp k θp These are the pitch damping coefficient and pitch stiffness coefficient, respectively. s F is the vertical displacement of the vehicle's center of gravity. zsi Let X be the vertical force exerted by the suspension on the vehicle body along the i-th axis. This is then rearranged to obtain the overall vehicle system state vector X. body and its state derivative ;

[0100] The input signal for this step is the vehicle structural parameter, the vehicle mass M. v The moment of inertia of the whole vehicle in terms of yaw, roll, and pitch I z I x I y Center of mass height h c Distance from each axis to the centroid l i Wheelbase t i The output signal is the vehicle system state vector X. body Vehicle system state derivative ;

[0101] Step 2, Wheel Dynamics Modeling. Step 2 consists of three parts, as follows:

[0102] Step 1: Modeling the Vertical Characteristics of a Single-Wheel Tire. This step establishes a model of the vertical force-deformation relationship between the tires of each wheelset based on the tire's vertical stiffness and damping characteristics. The vertical force F of a single-wheel tire is... zti As shown in formula (30):

[0103] (30);

[0104] In the formula, k ti For vertical stiffness, z ui For the vertical displacement of the wheel center, z ri For the vertical displacement of the road surface excitation, c ti This is the tire's vertical damping coefficient. The vertical velocity of the wheel center. The vertical excitation velocity of the road surface;

[0105] The input signal for this stage is the vertical stiffness k. ti Vertical displacement z of wheel center ui, road surface excitation vertical displacement z ri Tire vertical damping coefficient c ti Vertical velocity of wheel center Vertical excitation velocity of the road surface The output signal is the vertical force F of a single tire. zti ;

[0106] Section 2: Tire Dynamics Integration Model for Multi-Axle Heavy-Duty Vehicles. Based on the single-wheel model in Section 1, this section integrates the vertical, longitudinal, and lateral forces of the tires for all n wheelsets of the vehicle to establish a tire dynamics integration model for multi-axle heavy-duty vehicles. It defines the core coupling characteristic parameters of multi-axle heavy-duty vehicles and quantifies the unique mechanical interference effects of ultra-multi-axle vehicles, as shown in formula (31).

[0107] (31);

[0108] In the formula, η Fz, i To account for the unevenness of load distribution across shafts, ξ ij σ represents the mechanical interference coefficients along the i-th and j-th axes. The closer the axial distance, the stronger the coupling interference. slip The non-synchronous distortion coefficient of multi-axle slip is a unique phenomenon of heavy-load vehicles with inconsistent wheel slip states. The tire dynamics integrated equation of heavy-load vehicles is constructed, and the full-domain coupling superposition is achieved through the inter-axle interference characteristics to reflect the mechanical cluster distortion characteristics of multi-axle heavy-load tires, as shown in formula (32):

[0109] (32);

[0110] The input signal for this stage is the vertical force F of a single tire. zti Unevenness of load on each axis η Fz, i The mechanical interference coefficients ξ of the i-th and j-th axes ij Multi-axis slip asynchronous distortion coefficient σ slip Total longitudinal tire force F of each axle xi Total lateral tire force F of each axle yi The output signal is the total longitudinal tire force F of the multi-axis cluster coupling. x, sum Multi-axis cluster coupled total lateral tire force F y, sum ;

[0111] Step 3: Multi-axle Heavy-Duty Tire Temperature-Wear Coupling Modeling. This step addresses the unique characteristics of multi-axle heavy-duty vehicles, which operate under conditions of high load cluster pressure, continuous multi-wheel slippage, and prolonged continuous operation. These conditions result in significantly greater tire heat accumulation and wear degradation compared to ordinary vehicles. Furthermore, the coupled friction between the wheels of multiple axles exacerbates the differentiated evolution of temperature loss. A tire heat generation-dissipation balance equation adapted to ultra-multi-axle heavy-duty conditions is established, coupling tire temperature changes to tire peak force and stiffness parameters. Simultaneously, a wear rate calculation model is constructed to dynamically correct tire mechanical parameters according to temperature and wear degree. Tire heat generation power P heat As shown in formula (33):

[0112] (33);

[0113] In the formula, V slip, x V slip, y F represents the longitudinal and lateral slip velocities, respectively. xi For the total longitudinal tire force of each axle, F yi Based on the characteristics of accumulated heat generation and uneven heat dissipation under heavy multi-axle loads, a multi-axle coupled dynamic temperature rise balance equation is established for the total lateral tire force on each axle. Since the tires of ultra-multi-axle vehicles do not experience independent temperature rises, but rather clustered thermal radiation, close-range wheel-to-wheel thermal coupling, and continuous heat accumulation under heavy loads, a multi-axle heat accumulation coefficient ε is introduced. c The dynamic change of tire temperature is shown in formula (34):

[0114] (34);

[0115] In the formula, C th Here, T represents the tire's heat capacity, and h represents the tire's real-time temperature. conv T is the convective heat transfer coefficient. amb For ambient temperature, A s This refers to the effective heat dissipation area of ​​the tire. Tire wear rate W rate As shown in formula (35):

[0116] (35);

[0117] In the formula, K wear E is the wear coefficient. a The activation energy is R, and the gas constant is R.

[0118] The input signals for this stage are the longitudinal and lateral slip velocities V. slip, x V slip, y Total longitudinal tire force F of each axle xi Total lateral tire force F of each axle yi Tire temperature T; the output signal is the tire heat generation power P. heat Tire heat capacity C th Ambient temperature T amb Tire wear rate Wrate ;

[0119] Step 3: Multi-round kinematic modeling. Step 3 consists of three parts, as follows:

[0120] Step 1: Establishing Multi-Axis Coupled Kinematic Constraints. This step establishes the kinematic constraint relationships between each wheelset and between the wheelsets and the vehicle body, providing constraint boundary conditions for subsequent multi-body motion state solutions. Each wheelset is connected to the vehicle body through the suspension system, and the suspension kinematic constraints are embodied in the wheel alignment parameter kingpin inclination angle γ. kpi Kingpin inclination angle γ kpc Wheel camber angle γ camber γ front toe angle toe The dynamic relationship between suspension travel and suspension travel, let the vertical runout of the i-th axle be Δz. si The changes in the camber angle and toe angle of the wheel on this axle are shown in formula (36):

[0121] (36);

[0122] In the formula, γ camber, 0 γ toe, 0 To design the positioning angle, k camber k toe These are the suspension geometric kinematic parameters;

[0123] Adjacent wheelsets are coupled and constrained by the frame structure. For multi-axle rigid frame vehicles, the displacement of each axle relative to the frame is constrained by the frame deformation, generating an inter-axle displacement deviation ΔP. i, i-1 As shown in formula (37):

[0124] (37);

[0125] In the formula, L i, i-1 Let e ​​be the initial axis distance between the i-th axis and the (i-1)-th axis. x For the longitudinal unit vector of the vehicle, kinematic interference occurs when the inter-axle displacement deviation exceeds the allowable range, and constraint compensation is performed. Based on the attitude constraints of each wheelset and the coupling relationship between the inter-axle pose, the instantaneous equivalent steering angle δ of the entire vehicle with multiple wheelsets adapted to multi-axle vehicles is derived. ins As shown in formula (38):

[0126] (38);

[0127] The input signal for this stage is the kingpin inclination angle γ. kpi Kingpin inclination angle γ kpc Wheel camber angle γ camber γ front toe angle toe The vertical runout of the frame is Δz si The output signal is the inter-shaft displacement deviation ΔP.i, i-1 Multi-wheel instantaneous equivalent steering angle δ of the whole vehicle ins ;

[0128] Step Two: Kinematic Description of the Vehicle Multibody System. This step builds upon the suspension geometric constraints and multi-axle frame coupling constraints established in Step One. The entire vehicle is abstracted as a multibody system consisting of the vehicle body and multiple wheel pairs, achieving an accurate solution for the vehicle's multibody motion state. Considering the structural characteristics of ultra-multi-axle heavy-duty vehicles with 10 or more axles, the entire vehicle is viewed as a multibody system composed of the body and multiple wheel pairs. The vehicle's motion state is determined by the position of the center of mass P. G The yaw angle φ, pitch angle θp, roll angle ϕ, and the local motion state of each wheel pair are used to describe the vehicle's motion. A vehicle coordinate system Oxyz is defined, with the origin O located at the projection of the vehicle's center of mass onto the horizontal plane. The x-axis points along the vehicle's longitudinal direction in the forward direction, the y-axis points to the driver's left, and the z-axis is perpendicular to the ground and points upwards. For the i-th axis and j-th wheel, the contact point position vector P... ij As shown in formula (39):

[0129] (39);

[0130] In the formula, R G Let r be the vehicle body attitude rotation matrix. ij The effective rolling radius is given by R, where R is the vehicle body attitude rotation matrix. G The ZYX Euler angles are defined in the order shown in formula (40):

[0131] (40);

[0132] In the formula, R x R y R z Let x, y, and z be the fundamental rotation matrices about the x, y, and z axes, respectively. Differentiating these matrices with respect to the position vectors of each wheel's contact point yields the velocity vectors of each wheel's contact point. As shown in (41):

[0133] (41);

[0134] The input signal for this stage is the distance l from each axis to the centroid. i Wheelbase t i Location of the center of mass P G Yaw angle φ, pitch angle θp, roll angle ϕ, vehicle attitude rotation matrix R G Effective rolling radius r ij The output signal is the grounding point position vector P. ij Grounding point velocity vector ;

[0135] Step 3: Integration of the Augmented Kinematic Model of the Whole Vehicle. This step integrates the multi-axis coupled kinematic constraints of Step 1 with the multi-body system kinematic description of the whole vehicle in Step 2, and fuses the vehicle body dynamics state from Step 1 with the wheel dynamics state from Step 2 to construct a control-oriented multi-wheelset augmented kinematic model of the whole vehicle, selecting generalized coordinates q∈R. 3n+6 , representing all kinematic degrees of freedom of the entire vehicle, based on the kinematic description of link one and the constraint relationship of link two, the augmented kinematic mapping matrix S(q) is constructed as shown in formula (42):

[0136] (42);

[0137] In the formula, S body Map the dynamic parameters output in step one to the derivative of the vehicle body's generalized coordinates, S susp S represents the suspension kinematic constraints of the mapping. steer To map the turning angular velocity to the derivative of the generalized coordinates, a null space basis V2 is designed, and a reduced-order kinematic mapping matrix A is constructed. red (q)=V2·S(q), and the reduced-order augmented kinematic model is obtained as shown in formula (43):

[0138] (43);

[0139] To embed the dynamic equations of step one and the tire force balance equations of step two into the kinematic model, a coupling constraint vector Ψ is defined as shown in formula (44):

[0140] (44);

[0141] In the formula, q body Let be the vehicle body pose vector. For vehicle speed, Let C be the vehicle body acceleration, C be the system damping matrix, K be the system stiffness matrix, and F be the system acceleration. step1 F is a generalized force vector. ij For the actual tire force, the constraint vector is differentiated with respect to time and transformed into a linear constraint at the generalized acceleration level. Combined with the reduced-order kinematic model, the standard discrete-time augmented kinematic model for control is obtained as shown in formula (45):

[0142] (45);

[0143] In the formula, q(k+1) is the system generalized coordinate vector of the next discrete time. This model fully describes the kinematic coupling relationship of all wheelsets of a multi-axle heavy-duty vehicle with more than 10 axles. It realizes bidirectional deep coupling with the vehicle dynamics state in step one and the wheel dynamics state in step two. Its output state vector q(k) provides all the necessary kinematic feedback for the subsequent multi-wheelset system trajectory following controller.

[0144] The input signals for this stage are the generalized coordinate q and the generalized force vector F. stepl Vertical displacement of the vehicle body z s Mass matrix M v Damping matrix C, stiffness matrix K, tire force F ij Tire slip angle α ij Slip ratio λ ij The output signals are the augmented kinematic mapping matrix S(q) and the reduced kinematic mapping matrix A. red (q), coupling constraint vector Ψ, and the system generalized coordinate vector q(k+1) for the next discrete time step;

[0145] Step 4: Steering motor modeling. Step 4 consists of three parts, as follows:

[0146] Step 1: Establishment of the electromagnetic model of the steering motor. This step is based on the performance specifications of multi-axle heavy-duty vehicles for fast response, low hysteresis and large speed range of the steering system. The electromagnetic dynamic mathematical model of the steering motor is established, and the mathematical model in the dq coordinate system is adopted. The stator voltage equation is shown in formula (46):

[0147] (46);

[0148] In the formula, u d u q These are the stator voltages along the d-axis and q-axis, respectively, R s i is the equivalent resistance of the stator winding. d i q These are the stator currents along the d-axis and q-axis, respectively. d L q These are the equivalent inductances along the d-axis and q-axis, respectively, and ω e φ is the electric angular velocity of the motor. f The magnetic flux linkage is a permanent magnet. The electromagnetic torque T of the motor is... e As shown in formula (47):

[0149] (47);

[0150] In the formula, p n It is the extreme logarithm;

[0151] The input signals for this stage are the stator currents i along the d-axis and q-axis.d i q Stator winding equivalent resistance R s d-axis and q-axis equivalent inductance L d L q ω, the electric angular velocity of the motor e Permanent magnet flux linkage φ f The output signal is the d-axis and q-axis stator voltage u. d u q Motor electromagnetic torque T e ;

[0152] Step 2: Modeling the mechanical dynamics of the steering motor. This step establishes the mechanical dynamics equations of the steering motor, describing the dynamic relationship between the motor's output torque and the rotation angle and speed. The mechanical dynamics balance equation of the motor is shown in formula (48):

[0153] (48);

[0154] In the formula, J m Let ω be the moment of inertia of the motor rotor. m B is the mechanical angular velocity of the motor. m T is the mechanical damping coefficient. L This is the load torque;

[0155] The input signal for this stage is the rotor inertia J of the motor. m ω, mechanical angular velocity of the motor m Mechanical damping coefficient B m Load torque T L The output signal is the motor electromagnetic torque T. e ;

[0156] Step 3: Dynamic modeling of the steering system based on multiple motors and multiple degrees of freedom. This step couples the steering motor model with the steering actuator to establish a complete transmission model from motor torque to wheel steering angle. The steering wheel dynamic equation is shown in formula (49):

[0157] (49);

[0158] In the formula, J sw δ is the equivalent rotational inertia of the steering wheel and steering column. sw B is the steering wheel angle. sw τ is the equivalent damping coefficient of the steering column. f, sw T is the coulomb friction torque of the steering column. driver The input torque applied to the driver, T column The torque transmitted from the steering column to the steering gear. For the i-th axis, the steering angles δ of the left and right wheels. i, L δ i, RThe Ackermann steering geometry constraints are satisfied as shown in formula (50):

[0159] (50);

[0160] In the formula, t i L represents the wheelbase of each axle. i, axle Let be the distance from the instantaneous center of rotation of each axle to the axle. For multi-axle steering vehicles, there is a definite distribution relationship between the steering angles of each steering axle, and the kinematic constraints of adjacent axles are shown in formula (51):

[0161] (51);

[0162] In the formula, δ i, i-1 The steering actuator dynamics are determined by the motor driving torque, equivalent inertia, damping, friction, and tire return torque, with the steering angle distribution coefficient being the factor. By combining the motor mechanical dynamics equations and the steering actuator dynamics balance equations, a multi-motor, multi-degree-of-freedom steering system dynamic model is obtained, as shown in formula (52):

[0163] (52);

[0164] In the formula, G r J is the steering gear reduction ratio. eq For the equivalent rotational inertia of the steering actuator, B eq τ is the equivalent damping coefficient of the steering actuator. f Let τ be the Coulomb friction torque. align δ is the tire return torque. w This refers to the wheel steering angle;

[0165] The input signal for this stage is the rotor inertia J of the motor. m Steering gear reduction ratio G r Equivalent moment of inertia J of steering actuator eq Wheel steering angle δ w Coulomb friction torque τ f Tire return torque τ align Mechanical damping coefficient B m The output signal is the motor electromagnetic torque T. e .

[0166] The process of designing the multi-round integrated steering control strategy in the third step is as follows:

[0167] Step 1: Determining the vehicle's steering intention. Step 1 consists of three steps, as follows:

[0168] Step 1, Traffic Situation Awareness and Scene Prediction, utilizes multi-source fusion information from cameras, millimeter-wave radar, lidar, and high-precision maps to construct a traffic scene state space adapted for ultra-long, multi-axle, heavy-duty vehicles. It extracts high-dimensional scene feature vectors containing vehicle motion, road geometry, and traffic density, and introduces a sliding window temporal prediction algorithm to achieve accurate rolling predictions of future short-term driving scenarios, providing a priori environmental temporal sequence for subsequent risk identification. The multi-axle heavy-duty scene state vector S is constructed. scene (t)=[V ego V target , d target , ρ road , λ traffic ] T A fixed-length sliding time window is used to extract continuous time series samples, and the scene time series matrix within the window is constructed as shown in formula (53):

[0169] (53);

[0170] In the formula, T w Given a sliding window duration, a first-order temporal evolution prediction model is constructed based on the temporal state matrix to output a predicted sequence of future scenes in the time domain. As shown in formula (54):

[0171] (54);

[0172] In the formula, A1 and B1 are time-series prediction weight matrices, which are updated in real time through rolling iteration;

[0173] The input signal for this stage is the vehicle speed V. ego Target speed V target Relative distance d target Road curvature ρ load Traffic density λ traffic The output signal is the scene state vector S. scene (t), Road Environment Prediction Sequence ;

[0174] Step 2, Driving Risk State Estimation, uses the scenario prediction sequence output from Step 1 as environmental input. It integrates a recursive least squares system identification and a temporal fuzzy C-means AI clustering dual-core algorithm to identify time-varying lateral dynamic parameters and road adhesion parameters of multi-axle vehicles online. Combined with the unique risk characteristics of multi-axle heavy loads, it achieves adaptive driving risk quantification without fixed thresholds. The least squares algorithm is used to identify key dynamic parameters such as road adhesion coefficient and lateral disturbance stiffness online, constructing a high-dimensional clustering feature vector X that includes environmental risk, stability risk, and multi-axle cooperative risk. risk (t), as shown in formula (55):

[0175] (55);

[0176] In the formula, μ est To identify the road surface adhesion coefficient using the least squares algorithm, δ des, i -δ act, i To address the steering synchronization deviation of multi-axle wheelsets, a temporal fuzzy C-means artificial intelligence clustering objective function J is constructed. FCM Strengthen the weight of the current state and suppress the interference of overloaded state lag, as shown in formula (56):

[0177] (56);

[0178] In the formula, k is the cluster category number, α(t) is the temporal forgetting weight coefficient, and u kt Where m is the membership degree and v is the fuzzy coefficient. k Given the cluster center vector of the k-th class, the driving conditions are adaptively divided into three categories—safe, warning, and high-risk—by iteratively updating the membership degree and cluster center. The membership degree u of the high-risk condition is then output. risk By combining cluster membership degree and three types of risk indicators, a dedicated driving risk cost function R for multi-axle heavy-load vehicles is constructed. risk As shown in formula (57):

[0179] (57);

[0180] The input signal for this step is the road environment prediction sequence. Real-time lateral load transfer rate (LTR), target steering angle δ of each wheelset des Actual steering angle δ of each wheel pair act Real-time minimum distance d to obstacles target, min The output signal is a high-dimensional clustering feature vector X. risk (t), membership degree of high-risk working conditions u risk Driving risk cost function R risk ;

[0181] Step 3: Comprehensive Determination of Steering Intent. This step is based on the driving risk state estimation results from Step 2, combined with the driver's input commands and the planning commands δ from the autonomous driving decision module. cmd, auto The overall vehicle steering intention is determined by comprehensive analysis. The determination logic is as follows: when R... risk ≥R th, high Active steering intervention is triggered when an effective obstacle avoidance path exists; when |δ w |≥|δ th |and ≥ When δ is determined to be the driver actively steering; cmd, auto≠0 and when the autonomous driving mode is activated, the autonomous driving planning instruction is used, with the priority being safety emergency intervention, driver input, and autonomous driving instruction. The steering intention I is output based on the judgment logic and the actual input. steer ;

[0182] The input signal for this stage is the driving risk cost function R. risk Driver inputs operation commands and planning commands. cmd, auto The output signal is the steering intention I. steer ;

[0183] Step 2, Vehicle Trajectory Planning Strategy. Step 2 consists of three parts, as follows:

[0184] Step 1: Initial trajectory generation. This step is based on the turning intention I determined in Step 1. steer With scene state vector S scene (t) To address the unique characteristics of multi-axle heavy-duty vehicles, such as long vehicle body, large turning radius, coupling constraints between multiple wheel pairs' turning angles, wide overall vehicle path envelope, and inconsistent motion trajectories of each axle's contact points, this paper improves the traditional RRT random tree algorithm to form an enhanced RRT planning algorithm adapted for heavy-duty applications. This algorithm introduces multi-wheel contact point envelope constraints throughout the sampling, node expansion, and collision detection processes. Simultaneously, it verifies whether the contact points of all n axles' wheels intrude into obstacles and lane boundaries. The algorithm sequentially completes trajectory sampling, node expansion, cost evaluation, multi-wheel pair envelope collision detection, and path backtracking, generating multiple discrete initial candidate trajectory clusters that satisfy the kinematic constraints of the multi-axle vehicle. The enhanced RRT algorithm introduces an adaptive bias sampling strategy with a bias probability p. bias The dynamic adjustment with the number of iterations is shown in formula (58):

[0185] (58);

[0186] In the formula, k is the current iteration number, and τ bias As the bias rate growth time constant, during the expansion phase, five candidate nodes are generated in a single sampling, and the cost function C(x) is selected as shown in formula (59):

[0187] (59);

[0188] In the formula, g(x) is the actual path cost from the starting point to the current node, h(x) is the heuristically estimated cost, and o(x) is the obstacle risk cost;

[0189] Add hard kinematic constraints to the boundaries of the whole vehicle, prohibit sampling beyond the maximum steering angle of a single wheel and the left and right limit lateral coordinates of the lane, and add lateral safety constraints to the envelope of the multi-axle whole vehicle. The set of constraint inequalities is shown in formula (60):

[0190] (60);

[0191] In the formula, δ max For the single-wheel limit steering angle, y lane, min y lane, max W represents the left and right lateral boundaries of the lane. ve For the width of the entire vehicle body, y obs Let d be the lateral position of the obstacle. target, y To minimize the lateral safety distance between the trajectory point and obstacles, constraints are also imposed on the grounding point positions of all n axes to prevent scraping of the middle and rear wheels of the ultra-long multi-axle vehicle. For the discrete node sequence obtained by RRT sampling, a fifth-order polynomial is used as the basis function for the trajectory curve to construct a continuous time-domain analytical expression for a single candidate trajectory. The longitudinal curve x of the trajectory... traj Horizontal curve y traj As shown in formula (61):

[0192] (61);

[0193] In the formula, t∈[0, T] p [ ] represents the planning time, and a0~a5 and b0~b5 represent the polynomial coefficients of the longitudinal trajectory and the lateral trajectory, respectively. The coefficients are determined by the initial and final positions and the initial and final velocities of the boundary conditions output by RRT.

[0194] The input signal for this stage is steering intention I. steer Scene state vector S scene (t), cost function C(x); output signal is bias probability p bias Single-wheel limit steering angle δ max , the left and right lateral boundaries of the lane y lane, min y lane, max Trajectory longitudinal curve x traj Horizontal curve y traj ;

[0195] Step 2, Trajectory Post-processing and Smoothing Optimization: This step performs post-processing and smoothing optimization on the initial trajectory generated in Step 1, eliminating unnecessary nodes and sharp corners, and generating a smooth trajectory that meets the kinematic and comfort requirements of heavy-duty vehicles. Due to the large suspension stiffness and mechanical limitations of the steering actuator in multi-axle heavy-duty vehicles, the severe fluctuations in trajectory curvature will directly cause frequent changes in the steering angle of each wheel pair, resulting in frequent impacts on the steering motor and an increased risk of vehicle rollover. Using the polynomial trajectory output in Step 1 as the initial value, a multi-objective cost function that balances tracking accuracy and the smoothness of multi-wheel pair steering execution is constructed. The polynomial coefficients are iteratively corrected, and the smoothing optimization objective function is shown in Formula (62):

[0196] (62);

[0197] In the formula, ω a ω k ω δ These are the penalty weights for lateral acceleration, trajectory curvature change, and wheelset angle change, respectively. For the lateral acceleration of the trajectory, Let d be the time derivative of the trajectory curvature. δi, ref / dt represents the rate of change of the reference steering angle of the i-th wheel pair, and is a penalty term for multi-axle heavy-duty vehicles. The polynomial coefficients are updated through optimization, and the corrected continuous smooth trajectory curve T is output. smooth This ensures that position, velocity, and acceleration are continuously differentiable.

[0198] The input signal for this stage is the longitudinal curve x of the trajectory. traj Horizontal curve y traj The output signal is the corrected, continuous, smooth trajectory curve T. smooth ;

[0199] Step 3: Trajectory Feasibility Verification and Safety Assessment. This step involves verifying the smooth trajectory curve T generated in Step 2. smooth Conduct multi-dimensional feasibility verification and safety assessment, and smooth the trajectory curve T. smooth Substituting the multi-axle vehicle dynamic coupling model built in the second step, we perform dynamic feasibility verification, kinematic feasibility verification, and full-domain safety constraint verification. If the verification finds that the trajectory has an over-limit risk, we locate the over-limit time period and call the enhanced RRT algorithm in step one to locally regenerate the trajectory until the safety constraints are met in all time domains. Finally, we output the unique optimal reference trajectory T with the highest safety score. final ;

[0200] The input signal for this stage is the smooth trajectory curve T. smooth The output signal is the globally optimal vehicle reference trajectory T after verification and correction. final ;

[0201] Step 3: Multi-round system trajectory following control strategy. Step 3 is completed in three stages, as follows:

[0202] Step 1: Augmented State Space Modeling of the Multi-Wheelset System. This step builds upon the multi-wheelset vehicle model established in Step 2, considering the large inertia transient lag of multi-axle heavy-duty vehicles, the time-varying nature of heavy-load parameters, and the dynamic coupling disturbance characteristics of multi-wheelsets. An augmented state space model adapted for transient trajectory following control is constructed, selecting the state vector x=[Y, V, r, φ, ϕ, δ]. i , …,δ n ] T All are inherited from the output of the augmented kinematics model in the second step, with the control input u=[δ 1, cmd , …, δ n, cmd , ΔT1, …,ΔT n ]T This includes steering commands for each wheelset and compensation for transient torque under heavy load. The nonlinear state-space equation with transient disturbance correction is established as shown in equation (63):

[0203] (63);

[0204] In the formula, d tran (t) represents the transient coupling disturbance term under multi-axis heavy load, and the globally optimal vehicle reference trajectory T output in step two is used. final The trajectory parameters in the vector are used as reference inputs, and the augmented state vector x is used as the reference input. aug The integral term including the trajectory tracking error is used to eliminate the steady-state error, as shown in Equation (64):

[0205] (64);

[0206] The input signals for this stage are the output state variable x of the multi-wheel-pair augmented kinematic model of the whole vehicle, the control input u, and the globally optimal whole vehicle reference trajectory T. final The trajectory parameters are in the vector; the output signal is the augmented state vector x. aug ;

[0207] Section Two: Layered Sliding Mode Variable Structure Trajectory Following Controller. This section, based on the augmented state-space model of Section One, designs a layered sliding mode variable structure controller to achieve robust following of the reference trajectory for the multi-wheelset system. First, a dynamic coordination layer—stability boundary partitioning and priority weight generation—is designed. The stability boundary of the multi-axle heavy-duty vehicle is partitioned based on the phase plane method, and the phase plane state variable is defined as the centroid sideslip angle β. p angular velocity of the center of mass deflection The boundary equation of the phase plane stable domain is shown in equation (65):

[0208] (65);

[0209] In the formula, c1 and c2 are boundary coefficients related to vehicle speed and road surface adhesion coefficient. The zero-moment point roll instability judgment model position P is introduced. ZMP As shown in formula (66):

[0210] (66);

[0211] When P ZMP When the vehicle exceeds the boundary of the vehicle support polygon, it is determined to be a roll instability. Based on the phase plane stability domain division and ZMP roll risk assessment, a priority weighting factor ρ is dynamically generated. prior As shown in formula (67):

[0212] (67);

[0213] In the formula, sat is the saturation function, and ρ prior Constrained in the interval [0, 1], β crit , P ZMP, max Given the sideslip angle, sideslip angular velocity, and ZMP position critical threshold, design a variable structure control layer—a multi-constraint sliding mode control law—and define the tracking error sliding surface s. track As shown in formula (68):

[0214] (68);

[0215] In the formula, λ1~λ3 are strictly positive definite sliding surface parameters, and the stable sliding surface s is defined. stab As shown in formula (69):

[0216] (69);

[0217] In the formula, λ4~λ7 are strictly positive definite sliding surface parameters, which are determined by the pole placement method to ensure the rapid convergence of the sliding mode dynamics and good damping characteristics. The composite sliding surface S is constructed as shown in formula (70):

[0218] (70);

[0219] The superspiral approach law is used to weaken chattering while ensuring finite-time convergence, as shown in formula (71):

[0220] (71);

[0221] In the formula, k1 and k2 are strictly positive definite reaching rate parameters, based on the synovial reachability condition. =0, combining the first step vehicle dynamics equations and the second step tire model, the inverse solution yields the desired generalized torque ν. des The design incorporates a constraint processing layer—multi-constraint boundary handling and multi-wheel pair collaborative error correction. For multiple constraint conditions, a combination of boundary layer and saturation function is used to handle actuator saturation and state constraints, defining a boundary thickness T. b Adaptive adjustment, the saturation function is sat(s, T) b Define the comprehensive error E of multi-round transient cooperative tracking. syns As shown in formula (72):

[0222] (72);

[0223] The comprehensive error E syns As a feedback correction, it is added to the desired generalized torque ν des The closed-loop correction is formed as shown in formula (73):

[0224] (73);

[0225] In the formula, K sync For positive definite cooperative correction gain matrix, ν des, mod This is the desired generalized moment vector after collaborative correction;

[0226] The input signal for this stage is the augmented state vector x. aug Boundary coefficients c1 and c2 related to vehicle speed and road surface adhesion coefficient, and the position P of the roll instability judgment model. ZMP Tracking error sliding surface s track Stability of sliding surface s stab Comprehensive error E syn The output signal is the priority weighting factor ρ. prior Synthetic sliding surface S, desired generalized torque ν des The desired generalized moment vector ν after collaborative correction des, mod ;

[0227] Step 3: Multi-actuator control allocation and follow-up execution. This step will distribute the desired generalized torque ν output from Step 2. des The control is assigned to the steering actuators of each wheelset, and rolling time-domain updates and fault-tolerant reconfiguration are implemented. The control efficiency matrix M is defined. c Its elements are determined by the Jacobian matrix of the steering angle of each wheel pair and the tire force on the generalized force of the whole vehicle in the vehicle dynamics model of step one. The optimization problem J alloc As shown in formula (74):

[0228] (74);

[0229] In the formula, u act Let Q be the actual output vector of the actuator, R be the weight matrix for allocation accuracy, R be the weight matrix for load balancing, RΔ be the weight matrix for motion smoothing, and u be the weight matrix for motion smoothing. prev W is the actuator output vector at the previous time step, and W is the cumulative workload vector of each actuator. avg For multi-axle heavy-duty vehicles, the vertical load F of each axle during transient steering is the average load. zi Drastic dynamic changes directly affect the control efficiency of each actuator. Therefore, a transient load correction factor η is introduced into the control efficiency matrix. z, i (t), as shown in formula (75):

[0230] (75);

[0231] Corrected control efficiency matrix As shown in formula (76):

[0232] (76);

[0233] When a portion of the wheelset steering actuators fails, the failure detection and isolation module continuously monitors the residual error. i (k), when |ϵ i When (k)|≥ϵth and continues for more than 5 sampling periods, it is considered a failure, and the fault tolerance allocation problem J fault As shown in formula (77):

[0234] (77);

[0235] In the formula, M fail To reduce the order of the control efficiency matrix after removing the column corresponding to the failed actuator, u act, fail F represents the actual output vector of the remaining healthy actuators under failure conditions. act For the set of failed actuators, U fail For each sampling time T, there is a feasible set of constraints for the remaining healthy actuators under failure conditions. s Based on the latest state measurements, the sliding mode control law and control assignment problem are resolved to achieve rolling updates, and the desired steering angle command δ for each wheelset is re-output. i, cmd This forms a complete closed loop of "perception-decision-allocation-execution-feedback," enabling multi-round systems to achieve high-precision collaborative following of the reference trajectory.

[0236] The input signal for this stage is the desired generalized torque ν. des Control efficiency matrix M c The actual output vector u of the actuator act The executor output vector u at the previous time step prev The output signal is the actual output vector u of the remaining healthy actuator under failure conditions. act, fail Failed actuator set F act The feasible constraint set U for the remaining healthy actuators under effective operating conditions fail δ, the desired steering angle command for each wheelset i, cmd ;

[0237] Step 4: Multi-target following control strategy for motors. Step 4 consists of three parts, as follows:

[0238] Stage 1: Steering angle loop following control. This stage is the outermost closed loop of the motor execution system, accurately tracking the desired steering angle command δ assigned to each wheel pair by the upper-level multi-wheel pair system. i, cmd A closed-loop tracking controller for the upper-level rotation angle position is designed to achieve precise tracking of the motor rotor position to the target rotation angle θ. m, ref =δ i, cmd / Gr, the position loop adopts a proportional-integral-derivative control framework, and introduces a model-based feedforward compensation and a reduced-order extended state observer to improve the dynamic response speed and steady-state accuracy. The control law is shown in formula (78):

[0239] (78);

[0240] In the formula, ω m, ref e is the target angular velocity output by the position loop. θ K represents the position tracking error. p, pos K i, pos K d, pos These are the position loop proportional, integral, and differential gain coefficients, respectively. The rate of change of error, For feedforward velocity compensation term, The total perturbation compensation for the position loop is estimated by the reduced-order extended state observer, which is in second-order form, as shown in formula (79):

[0241] (79);

[0242] In the formula, This is a location estimate. For the total disturbance estimate, β z1 β z2 For observer gain, θ m This represents the actual electrical angle of the motor rotor.

[0243] The input signal for this stage is the desired steering angle command δ for each wheelset. i, cmd Steering gear reduction ratio Gr, position loop proportional, integral, and differential gain coefficient K p, pos K i, pos K d, pos Actual rotor electrical angle θ of the motor m The output signal is the target angular velocity ω output by the position loop. m, ref Position tracking error e θ Error change rate ;

[0244] Step 2: Smooth closed-loop control of motor speed loop. This step is based on the target angular velocity ω output by the position loop. m, ref A mid-level speed smoothing closed-loop controller is designed to achieve rapid and smooth tracking of the motor's actual speed to the target speed. The speed loop employs a proportional-integral control law and introduces a first-order low-pass filter ω. LPF / (s+ω LPF ), filtering out velocity measurement noise, filter cutoff angular frequency ω LPFThe selection should balance noise suppression effect and phase lag, and should be 5 to 10 times the speed loop bandwidth. The phase lag of the filter at the speed loop crossing frequency can be offset by acceleration feedforward compensation.

[0245] Introducing acceleration feedforward term As shown in formula (80):

[0246] (80);

[0247] In the formula, ω m, prev T represents the target angular velocity at the previous moment. vel For the speed loop control cycle, the feedforward term The current command directly reflects the changing trend of the target speed and is superimposed on the output of the PI controller in an open-loop manner, so that the current command obtains a corresponding increment at the beginning of the speed change, shortening the acceleration response time. The overall control law is shown in formula (81):

[0248] (81);

[0249] In the formula, K p, vel K i, vel For the speed loop proportional and integral gain coefficients, sat AW To resist integral saturation, the q-axis target current i output by the velocity loop is... q, ref With the underlying torque limit value T e, limit The combined effect, after online saturation limiting, is shown in formula (82):

[0250] (82);

[0251] In the formula, p n This represents the number of pole pairs of the motor.

[0252] The input signal for this stage is the target angular velocity ω output by the position loop. m, ref , Filter cutoff angular frequency ω LPF The target angular velocity ω at the previous moment m, prev The bottom torque limit value T e, limit Number of pole pairs p of the motor n The output signal is the trend of the target velocity. The target q-axis current i output by the speed loop q, ref ;

[0253] Section 3: Adaptive closed-loop control of the motor torque loop. This section is based on the q-axis target current i output by the speed loop. q, refA bottom-level torque adaptive closed-loop controller is designed to achieve accurate and rapid response of the motor's electromagnetic torque to the target torque. The current loop adopts a control architecture that includes feedforward decoupling and PI control, and superimposes current prediction compensation, adaptive field weakening control, torque-current adaptive calibration and active damping injection modules. The current loop control law is shown in formula (83):

[0254] (83);

[0255] In the formula, K p, cur K i, cur These are the current loop proportional gain and integral gain coefficient, u q, comp u d, comp This is the current prediction compensation term. When the motor terminal voltage approaches the DC bus voltage limit, adaptive field weakening control is activated, and the d-axis target current i output by the speed loop is... d, ref As shown in formula (84):

[0256] (84);

[0257] In the formula, φ f (T motor The permanent magnet flux linkage is temperature-corrected. The torque constant deviation is identified online using recursive least squares with a forgetting factor. The identification model is shown in formula (85).

[0258] (85);

[0259] In the formula, Let be the estimated torque constant to be identified at the k-th sampling time, and ε(k) be the identification residual at the k-th sampling time. The recursive process is shown in formula (86):

[0260] (86);

[0261] In the formula, P(k-1) is the covariance matrix at the (k-1)th step, and λ FF As the forgetting factor, the covariance matrix P(k) at the k-th step is constructed as shown in formula (87):

[0262] (87);

[0263] Corrected q-axis target current i q, ref, mod As shown in formula (88):

[0264] (88);

[0265] The transfer function of the active damped injection bandpass filter is shown in equation (89):

[0266] (89);

[0267] In the formula, G BPF (s) is the transfer function of the bandpass filter, s is the complex frequency domain variable, ζ is the damping ratio of the bandpass filter, and ω is the signal strength of the bandpass filter. n The center angular frequency of the bandpass filter is Δi, and the active damping injection current compensation is Δi. q, damp As shown in formula (90):

[0268] (90);

[0269] In the formula, K damp The active damping gain is dynamically adjusted according to the steering load as shown in formula (91):

[0270] (91);

[0271] In the formula, K damp, 0 K is the base-valued damping gain. damp, 1 T is the load-related damping gain coefficient. L For steering load torque. Corrected total q-axis current command i q, ref, total As shown in formula (92):

[0272] (92);

[0273] Add a torque overload graded protection and safety monitoring module, current RMS value I rms As shown in formula (93):

[0274] (93);

[0275] To suppress temperature rise and ensure safe operation of the motor by linearly reducing the torque limit when the motor temperature exceeds the rated value but does not reach the limit value, the first-level protection torque derating is activated, as shown in formula (94):

[0276] (94);

[0277] In the formula, T e, limit T is the torque limiting value after the first-level protection is activated. nom T is the rated operating temperature of the motor. max This refers to the maximum allowable temperature of the motor.

[0278] The input signal for this stage is the q-axis target current i output by the velocity loop. q, ref Current loop proportional gain and integral gain coefficient K p, cur K i, cur The estimated value of the torque constant to be identified at the k-th sampling time. The transfer function G of a bandpass filter BPF (s), Active damping injection current compensation Δi q, damp Rated operating temperature of the motor (T) nom Maximum allowable temperature T of the motor max The output signal is the d-axis target current i output by the speed loop. d, ref Corrected q-axis target current i q, ref, mod Corrected total q-axis current command i q, ref, total Torque limiting value T after first-level protection starts e, limit .

[0279] The process of constructing the multi-round integrated steering evaluation method in the fourth step is as follows:

[0280] Step 1: Minimum number of rounds verification calculation method. Step 1 consists of three steps, as follows:

[0281] Step 1: Single-wheelset capacity boundary calculation. This step takes the capacity requirements output from the first step of the vehicle performance specification design as input, and outputs the lower limit of the number of wheelsets configuration in the second step of multi-wheelset vehicle modeling, forming a two-way verification closed loop of "positive demand setting and negative lower limit setting". This is combined with the rated load capacity m of the single-wheelset in the first step. rated Wheelset dynamic load correction factor k v (V, R), dimensionless comprehensive slip variable σ of the second-step tire mechanics model, tire curvature factor E1, nonlinear correction coefficient σ tire (E1, σ), comprehensively considering four core influencing factors—road speed coupled steady-state load, tire curvature stiffness nonlinear degradation, tire slip mechanical attenuation, and heavy-load transient dynamic impact—a quantitative model of the single-wheelset ultimate load capacity adapted to complex dynamic conditions is constructed. This model integrates multi-dimensional degradation and transient response characteristics to establish a quantitative formula for the single-wheelset performance boundary. The dynamic ultimate load capacity m of the single-wheelset is... lim The calculation formula is shown in formula (95):

[0282] (95);

[0283] Single wheel pair ultimate lateral adhesion force F y, lim The calculation formula is shown in formula (96):

[0284] (96);

[0285] A tire lateral force wheel-pair rotation angle coupling equation is established to represent the dynamic relationship between rotation angle δ and the load-bearing and adhesion performance of a single wheel pair, as shown in formula (97):

[0286] (97);

[0287] In the formula, σ(δ) is the dimensionless composite slip variable that varies with the steering angle, and k δ δ0 is the steering slip correlation coefficient, where δ0 is the zero-slip reference steering angle of the wheelset;

[0288] The input signal for this stage is the rated load capacity (m) of a single wheelset. rated Wheelset dynamic load correction factor k v (V,R), dimensionless comprehensive slip variable σ of tire mechanics model, tire curvature factor E1, nonlinear correction coefficient σ tire (E1, σ), load unevenness distribution coefficient η load The output signal is the dynamic ultimate bearing capacity m of a single wheelset. lim Single wheel pair ultimate lateral adhesion force F y, lim , rotation angle δ;

[0289] Step Two: Calculation of Vehicle Capability Boundary Conditions under Specific Loads. This step, based on multi-condition carrying capacity, stability, and adhesion boundaries, calculates the basic vehicle capability boundary required to meet the load requirements under a specific load condition, including the total driving force F required by the vehicle. req From the highest vehicle speed V in the first step of the dynamics assessment max The maximum gradient α is determined as shown in formula (98):

[0290] (98);

[0291] In the formula, M cargo For cargo mass, the total braking force F required for the entire vehicle b, req The maximum braking deceleration 'a' given by the first step of the braking stability assessment x, max As shown in formula (99):

[0292] (99);

[0293] Total lateral force required for the vehicle F y, req The maximum safe lateral acceleration a given by the first step of the handling stability assessment y, max As shown in formula (100):

[0294] (100);

[0295] The input signal for this stage is the maximum vehicle speed V. max Maximum gradeability α, maximum braking deceleration a x, max Maximum safe lateral acceleration a y, max Vehicle weight M v Goods quality M cargo The output signal is the total driving force F required by the entire vehicle. req Total braking force required for the entire vehicle Fb, req Total lateral force F required for the entire vehicle y, req ;

[0296] Step 3: Reverse Derivation of Minimum Wheelset Count. This step, based on the microscopic performance boundaries of a single wheelset in Step 1 and the macroscopic boundaries of the entire vehicle in Step 2, employs a reverse derivation logic using performance parameters. Constrained by the compliance requirements of the first step (transportation safety) and the third step (steering stability), it independently derives the minimum wheelset count from four dimensions: load-bearing constraints, power constraints, braking constraints, and lateral stability constraints. This reverse derivation determines the minimum wheelset count n that satisfies safe driving under all operating conditions. load As shown in formula (101):

[0297] (101);

[0298] In the formula, F0 is the rated load of a single wheelset, and λ z The bearing capacity degradation coefficient is the road surface. The variance of the multi-axis load is given. The minimum number of wheelsets n based on driving force constraints is also given. drive As shown in formula (102);

[0299] (102);

[0300] In the formula, λ d To drive the dimensional coupling weight coefficients, ΔF dyn (n) represents the dynamic axle load transfer coupling term under accelerated conditions, indicating the degree of degradation of drive adhesion conditions caused by axle load redistribution during acceleration under different wheelset numbers. The minimum number of wheelsets n based on braking force constraints. brake As shown in formula (103):

[0301] (103);

[0302] In the formula, λ b Φ is the coupling weight coefficient for the braking dimension. brake (n) is the axle load mismatch function under braking conditions, characterizing the risk of wheel lock-up caused by axle load shifting during emergency braking, and is the minimum number of wheelsets n based on lateral stability constraints. lat As shown in formula (104):

[0303] (104);

[0304] In the formula, λ y , λ s For the lateral stability dimension coupling weight coefficients, Φ roll (n) is the multi-axis roll stiffness attenuation coupling term, characterizing the effect of wheelset number variation on the vehicle's equivalent roll stiffness and ultimate roll angle. P s(n) is the multi-axis steering geometry interference penalty function. The minimum number of wheelsets in the global envelope is n. min As shown in formula (105):

[0305] (105);

[0306] Take the maximum value of the required number of wheelsets for each boundary, ensuring that the number of individual wheelsets simultaneously meets all requirements for driving force, braking force, lateral force, and load-bearing capacity. Compare the calculated result with the actual number of wheelsets n configured in the second step. If n ≥ n min The current configuration meets the carrying requirements if n < n min If so, the "insufficient wheel pair configuration" signal needs to be returned to the second step again, and the wheel pair configuration needs to be increased again;

[0307] The input signal for this stage is the total driving force F required by the entire vehicle. req Total braking force required for the entire vehicle F b, req Total lateral force F required for the entire vehicle y, req Theoretical maximum carrying capacity of the whole vehicle M total The output signal is the minimum number of wheelsets n based on the bearing constraint. min ;

[0308] Step 2, Wheelset Failure Performance Assessment Method. Step 2 consists of three steps, as follows:

[0309] Step 1: Establishment of a wheelset failure mode discrimination mechanism. This step, based on the multi-wheelset steering system execution response, motor operating status, tire mechanical feedback, vehicle motion state parameters, and wheelset status monitoring data, identifies the failure modes of each wheelset. Model-based residual analysis is used for failure detection and isolation. The residual ϵ i (k)=u cmd, i (k)-u fb, i (k) represents the instruction value u cmd, i (k) and actual feedback value u fb, i The difference (k) when |ϵ i (k)|≥ϵ th When the failure persists for more than 5 sampling periods, it is determined that a failure has occurred. A quantitative classification criterion is established for the failure mode, as shown in formula (106):

[0310] (106);

[0311] In the formula, l i For the failure mode indication vector, ΔT lag To respond to the delay time, u max, sensor Δu is the upper limit of the sensor's measurement range. jump, max N represents the maximum allowable value for signal transitions. total, fail The cumulative number of judgment cycles for complete failure;

[0312] The input signal for this stage is the command value u. cmd, i (k) Actual feedback value u fb, i (k), Response delay time ΔT lag Sensor range upper limit u max, sensor Maximum allowable value of signal transition Δu jump, max The cumulative number of complete failure judgment cycles N total, fail The output signal is the failure mode indicator vector l. i ;

[0313] Step 2: Quantification of Performance Degradation of Failed Wheelsets. This step is based on the failure mode indicator vector l determined in Step 1. i The performance degradation degree of each failed wheelset is quantified, and the discrete failure type of each wheelset is mapped to a continuous performance degradation factor ρ. deg, i , representing the degree of performance loss of each failed wheelset relative to the healthy state, as shown in formula (107):

[0314] (107);

[0315] In the formula, T avail T represents the currently available torque. req For the target torque requirement, l i =0 indicates a normal state. After obtaining the degradation factors of each wheelset, the performance degradation level at the whole vehicle level is further evaluated. Different wheelsets bear different axle loads in the vehicle, and their impact on the overall vehicle performance has different weights. The comprehensive index of overall vehicle performance degradation ρ deg, total With the axle load F of each wheelset z, i The weighted average is calculated as shown in formula (108):

[0316] (108);

[0317] The input signal for this stage is the performance degradation factor ρ. deg, i Failure Mode Indication Vector l i Axle load F of each wheelset z, i The output signal is the overall vehicle performance degradation index ρ. deg, total ;

[0318] Step 3: Steering performance verification under failure conditions. This step uses the overall vehicle performance degradation index ρ output from Step 2 as the basis for verification. deg, total As input, it is embedded into the vehicle dynamics model established in the second step. By modifying the steering angle and tire force output boundary of the corresponding failed wheelset, the three key steering performance indicators under the failure condition are recalculated: minimum steering radius R min Maximum safe steering speed V safe Yaw rate tracking error er The performance degradation index is used to correct the steering angle and tire force output constraints of the failed wheelset, as shown in formula (109):

[0319] (109);

[0320] In the formula, δ i, max, fail F y, i, max, fail The maximum steering angle and maximum lateral tire force after correction for the failed wheelset;

[0321] When a steering actuator fails, the dynamic response of a multi-axle heavy-duty vehicle exhibits strong nonlinearity, multi-axis coupling, and multi-timescale characteristics. A physical information neural network (PIN) is introduced to perform a high-precision evaluation of steering performance under failure conditions using a data model fusion-driven approach. The PIN model consists of four parts: an input layer, a feature extraction layer, a physical constraint layer, and an output layer. The input layer receives the current vehicle state vector X. fail As shown in formula (110):

[0322] (110);

[0323] In the formula, failID is the identifier encoding of the failed wheel pair. The feature extraction layer uses a bidirectional long short-term memory network to extract temporal features and embeds a self-attention mechanism to automatically identify the key features that have the greatest impact on the performance evaluation of the failure condition. The high-dimensional temporal input is mapped to the latent feature space, and the enhanced feature vector H is obtained after self-attention weighting. att As shown in formula (111):

[0324] (111);

[0325] In the formula, Att is the self-attention operator, BIL is the dynamic feature extracted by the bidirectional long short-term memory network, H(t-1) is the historical hidden state vector of the bidirectional long short-term memory network at time t-1, the physical constraint layer uses the prior constraints of the yaw and roll dynamics of the multi-axle vehicle to correct the network output, and the physical constraint loss term Loss phys As shown in formula (112):

[0326] (112);

[0327] In the formula, f dyn The multi-wheelset vehicle dynamics mapping function is obtained by inputting the corrected failed wheelset constraints and real-time vehicle state into a trained physical information neural network. The output layer directly maps the predicted values ​​of three key steering performance parameters under the failure condition. , , As shown in formula (113):

[0328] (113);

[0329] The input signal for this stage is the overall vehicle performance degradation index ρ. deg, total Maximum steering angle and maximum lateral tire force δ after correction of the failed wheelset i, max, fail F y, i, max, fail The current vehicle state vector X fail Physical constraint loss term Loss phys The output signal is the enhanced feature vector H after self-attention weighting. att Physical constraint loss term Loss phys Predicted minimum turning radius under failure conditions Predicted maximum safe steering speed under failure conditions yaw rate tracking error prediction value ;

[0330] Step 3: Low-speed steering wheel pair coordination evaluation method. Step 3 consists of three parts, as follows:

[0331] Step 1: Low-Speed ​​Steering Wheelset Status Acquisition. This step collects real-time status parameters of each wheelset under low-speed steering conditions. At low speeds, the tire slip angle is small, and the wheelset kinematics are closer to the Ackerman steering geometry. Therefore, the data collected under this condition more accurately reflects the steering coordination between multiple wheelsets. The theoretical Ackerman angle δ of each wheel... i, theory As shown in formula (114):

[0332] (114);

[0333] In the formula, R ack δ1 is the turning radius of the first axis, and δ1 is the actual turning angle of the first axis;

[0334] Based on this, real-time status parameters of each wheelset are collected and calculated, including the actual steering angle δ of each wheel. i Rate of change of rotation angle of each wheel Tire slip angle α of each wheel i Tire wear rate W for each wheel rate, i After aligning the four types of data by timestamp, they are organized into a low-speed steering wheelset state matrix S, with wheelsets as the unit. low As shown in formula (115):

[0335] (115);

[0336] In the formula, each row in the matrix corresponds to a wheelset, the row number is the axle number, the first axle to the nth axle, and the four columns are the actual steering angle, the rate of change of steering angle, the tire slip angle and the tire wear rate of each wheel, respectively.

[0337] The input signal for this stage is the first axle steering radius R. ack 1. Actual steering angle δ1 of the first axis; 2. Longitudinal distance L from the first axis to the i-th axis. i Actual steering angle δ of each wheel i Rate of change of rotation angle of each wheel Tire slip angle α of each wheel i Tire wear rate W for each wheel rate, i The output signal is the theoretical Ackermann angle δ for each wheel. i, theory Low-speed steering wheel pair state matrix S low ;

[0338] Step 2: Calculation of wheel set wear degree. This step is based on the low-speed steering wheel set state matrix S collected in Step 1. low The wear rate is used to assess the unevenness of wear among wheelsets. The wear rate of each wheelset reflects its wear rate per unit time. The wear rates collected in step one are compared to determine if the wear is reasonable. The average wear rate W of all wheelsets is then calculated. rate, avg As shown in formula (116):

[0339] (116);

[0340] The wear index W of the i-th pair uneven, i Defined as the relative degree to which the wear rate of the wheelset deviates from the average wear rate, as shown in formula (117):

[0341] (117);

[0342] In the formula, ε W For custom small quantities, to prevent the denominator from being zero, W uneven, i A larger value indicates a more severe deviation of the wheelset's wear from the average level, and a higher risk of uneven wear. The wheelet number i with the most severe uneven wear is also identified. max =argmax i W uneven, i and its wear index W uneven, max The overall vehicle wear uniformity index U wear Defined as the average compensation value of the wear index of all wheelsets, as shown in formula (118):

[0343] (118);

[0344] U wear The closer to 1, the more uniform the wear of each wheel pair; the closer to 0, the more severe the uneven wear. When the result indicates severe uneven wear, the steering control strategy needs to be optimized and adjusted, and the weight of the steering angle distribution of each axle needs to be re-optimized to reduce the load on the high-wear wheelsets.

[0345] The input signal for this stage is the low-speed steering wheel pair state matrix S. low The output signal is the average wear rate W of the wheelset. rate, avg The wear index W of each wheelset uneven, i The wheelset with the most severe uneven wear is numbered i max Overall vehicle wear uniformity index U wear ;

[0346] Step 3: Comprehensive Calculation of Wheelset Coordination. This step comprehensively calculates the low-speed steering wheelset coordination based on three dimensions: the deviation of each wheelset's steering angle from the theoretical Ackermann steering angle, the consistency of each wheelset's sideslip angle, and the uniformity of uneven wear. The coordination degree C of each wheelset is calculated. sync, i As shown in formula (119):

[0347] (119);

[0348] In the formula, ω δ ω α ω W α is the weighting coefficient. max The critical sideslip angle is the low-speed steering wheel pair coordination degree C of the entire vehicle. sync, total The arithmetic mean of the coordination of all wheelsets is given by formula (120):

[0349] (120);

[0350] The input signal for this stage is the actual steering angle δ. i Theoretical Ackermann rotation δ i, theory Actual sideslip angle α i Mean sideslip angle α avg Critical sideslip angle α max The wear index W of each wheelset uneven, i The output signal is the coordination degree C of each wheelset. sync, i Low-speed steering wheel pair coordination degree C of the whole vehicle sync, total ;

[0351] Step 4: High-speed steering stability assessment method. Step 4 consists of three parts, as follows:

[0352] Step 1: Extraction of Dynamic Response Characteristics for High-Speed ​​Steering. This step focuses on the high-speed steering conditions of heavy-duty vehicles, constructing a standardized test case set S that includes constant speed lane change, high-speed step steering, and high-speed sinusoidal steering. cThe constant speed lane change test is used to evaluate the vehicle's lateral stability and yaw response quality in highway scenarios; the high-speed step steering test is used to obtain the transient response characteristics of the vehicle's yaw rate, including response time and overshoot; the high-speed sinusoidal steering test is used for frequency domain identification, extracting amplitude attenuation and phase lag characteristics at different excitation frequencies. The extracted time-domain response signal includes the steering wheel angle δ. i Yaw angular velocity r(t), roll angle ϕ(t), lateral acceleration a y (t), Vertical force F of each wheel zi (t) , The collected data is filtered and denoised to remove outliers, forming a standardized test dataset S. d ;

[0353] The input signal for this stage is the steering wheel angle δ. i Yaw angular velocity r(t), roll angle ϕ(t), lateral acceleration a y (t), Vertical force F of each wheel zi (t); the output signal is the standardized test condition set S. c Standardized test dataset S d ;

[0354] Step 2: Calculation of high-speed steering stability index. This step is based on the standardized test dataset S extracted in Step 1. d The core stability evaluation indices are calculated, and the yaw oscillation attenuation coefficient, high-speed trajectory tracking accuracy, high-speed lateral stability margin, and yaw oscillation attenuation rate ζ are solved respectively. yaw The calculation is shown in formula (121):

[0355] (121);

[0356] In the formula, n is the total number of oscillation peaks counted in the yaw rate decay oscillation waveform, and A r1 A represents the amplitude of the first peak of the yaw oscillation. rn This represents the amplitude of the nth peak of the yaw oscillation. High-speed trajectory tracking accuracy error E y, high The calculation is shown in formula (122):

[0357] (122);

[0358] In the formula, N is the total number of sampling points, Y ref Y represents the desired lateral trajectory coordinates. act These are the actual lateral trajectory coordinates. High-speed lateral stability margin K. stab, high The calculation is shown in formula (123):

[0359] (123);

[0360] The input signal for this step is the standardized test dataset S. d The output signal is the yaw oscillation decay rate ζ. yaw High-speed trajectory tracking accuracy error E y, high High-speed lateral stability margin K stab, high ;

[0361] Section 3: Comprehensive Rating of High-Speed ​​Steering Stability. This section will assess the stability of the three core indicators calculated in Section 2: yaw rate attenuation ζ. yaw High-speed trajectory tracking accuracy error E y, high High-speed lateral stability margin K stab, high A weighted fusion method is used to establish a comprehensive rating for high-speed steering stability; three indicators characterize high-speed steering stability from three dimensions: transient response quality, trajectory holding capability, and roll safety margin. yaw E reflects the vehicle's damping characteristics in response to steering input; a larger value indicates faster yaw rate convergence. y, high This reflects the vehicle's ability to track the desired path; a smaller value indicates higher tracking accuracy. K stab, high Reflecting a vehicle's ability to resist rollover, a higher value indicates a greater lateral stability margin. The overall rating for high-speed cornering stability is R. high As shown in formula (124):

[0362] (124);

[0363] In the formula, λ h1 , λ h2 , λ h3 ζ is the weighting coefficient. yaw, ref For reference to the yaw oscillation decay rate, E y, allow To allow for the maximum trajectory tracking error, K stab, ref For reference to lateral stability margin, the rating criteria are as follows: R high A value ≥0.85 is considered excellent, indicating high safety during high-speed steering, superior handling quality, rapid yaw convergence, accurate trajectory tracking, and ample roll margin; 0.70≤R high A value <0.85 is considered good, meeting the basic safety requirements for high-speed steering; 0.55≤R high A value <0.70 is acceptable, but this indicates a risk of roll or trajectory deviation, requiring limitation of the maximum steering speed or optimization of tire lateral stiffness matching; R high A value <0.55 is considered unacceptable, indicating a safety hazard during high-speed steering. The suspension roll stiffness, wheelset steering angle distribution strategy, or tire selection need to be reassessed.

[0364] The input signal for this stage is the yaw rate decay ζ. yaw High-speed trajectory tracking accuracy error E y, high High-speed lateral stability margin K stab, high Reference yaw oscillation decay rate ζyaw, ref Maximum allowable trajectory tracking error E y, allow Reference lateral stability margin K stab, ref The output signal is the comprehensive rating of high-speed steering stability, R. high Rating results.

[0365] The beneficial effects of this invention are:

[0366] 1) The multi-wheel pair multi-actuator integrated steering method for heavy-duty vehicles described in this invention establishes a multi-dimensional performance index system covering carrying capacity, power, stability and passability through modular heavy-duty vehicle performance specification design, providing quantitative design basis and performance boundary conditions for the steering system design of heavy-duty vehicles with more than 10 axles;

[0367] 2) The multi-wheel pair multi-actuator integrated steering method for heavy-duty vehicles described in this invention establishes a multi-wheel pair whole vehicle model and constructs an integrated model of vehicle body dynamics, wheel dynamics, multi-wheel pair kinematics and steering motor for ultra-multi-axle vehicles with more than 10 axles. It solves the technical problem of increased degrees of freedom and complex constraints in modeling ultra-multi-axle vehicles and provides a high-fidelity controlled object model for control strategy design.

[0368] 3) The multi-wheel pair multi-actuator integrated steering method for heavy-duty vehicles described in this invention constructs a four-layer hierarchical control architecture from steering intention judgment, trajectory planning, trajectory following to multi-level motor control through the design of multi-wheel pair integrated steering control strategy. It gives full play to the system redundancy advantage of multi-wheel pair multi-actuator and realizes accurate, stable and coordinated steering of multi-axle vehicles under all working conditions.

[0369] 4) The multi-actuator integrated steering method for heavy-duty vehicles described in this invention constructs a three-loop cascade control architecture of position loop, speed loop and current loop through a multi-level motor following control strategy, realizing high-precision and fast-response torque output and angle following of permanent magnet synchronous steering motor under extreme working conditions, and solving the problems of response lag and insufficient control accuracy of steering actuators in heavy-duty vehicles.

[0370] 5) The multi-wheelset multi-actuator integrated steering method for heavy-duty vehicles described in this invention establishes a reverse deduction mechanism for the minimum number of wheelsets under specific load and vehicle capacity boundary constraints through a minimum wheelset number verification calculation method, providing a theoretical basis for the optimization of wheelset configuration for multi-wheelset heavy-duty vehicles;

[0371] 6) The multi-wheelset multi-actuator integrated steering method for heavy-duty vehicles described in this invention constructs a full-scenario steering performance evaluation system covering failure conditions, low-speed conditions, and high-speed conditions through wheelset failure performance evaluation method, low-speed steering wheelset coordination degree evaluation method, and high-speed steering stability evaluation method, providing a systematic evaluation tool for the design and verification of heavy-duty vehicle steering systems. Attached Figure Description

[0372] Figure 1 This is a schematic diagram illustrating the overall steps of the multi-wheelset multi-actuator integrated steering method for heavy-duty vehicles described in this invention.

[0373] Figure 2 This is a schematic diagram of the overall architecture of the multi-wheel pair multi-actuator integrated steering method for heavy-duty vehicles described in this invention.

[0374] Figure 3 This is a schematic diagram of the overall architecture of the first step described in this invention.

[0375] Figure 4 This is a schematic diagram of the overall architecture of the second step described in this invention.

[0376] Figure 5 This is a schematic diagram of the overall architecture of the third step described in this invention.

[0377] Figure 6 This is a schematic diagram of the overall architecture of the fourth step described in this invention.

[0378] Figure 7 This is a flowchart of the algorithm for step three in the second step of the present invention.

[0379] Figure 8 This is a flowchart of the algorithm for step three in the third step of the present invention.

[0380] Figure 9 This is a flowchart of the algorithm for step four in the third step of the present invention. Detailed Implementation

[0381] Please see Figures 1 to 9 As shown:

[0382] The multi-wheel-pair, multi-actuator integrated steering method for heavy-duty vehicles provided by this invention is as follows:

[0383] Step 1: Modular heavy-duty vehicle performance specification design;

[0384] The second step is to establish a multi-wheel vehicle model.

[0385] Step 3: Design of multi-wheel integrated steering control strategy;

[0386] Step 4: Construction of a multi-round integrated steering evaluation method.

[0387] The process of designing the performance specifications for modular heavy-duty vehicles in the first step is as follows:

[0388] Step 1: Assessment of the carrying capacity of modular heavy-duty vehicles. Step 1 consists of three steps.

[0389] Step 1: Calculation of the mapping relationship between the number of wheelsets and the carrying capacity. This step establishes a basic mapping relationship between the number of wheelsets and the carrying capacity of the vehicle based on the load-bearing characteristics of multi-wheelset configurations in heavy-duty vehicles. Let the rated carrying capacity of a single wheelset be m. rated With n wheelsets, considering the coupled effect of tire vertical stiffness and road surface unevenness, the actual load-bearing capacity of a single wheelset dynamically changes with vehicle speed V and road surface grade R. A dynamic load-bearing correction factor k is established. v (V, R), then the theoretical maximum carrying capacity M of the whole vehicle total As shown in formula (1):

[0390] (1)

[0391] In the formula, η load is the load unevenness distribution coefficient. Wherein, the dynamic load correction factor k v The expression for calculating (V, R) is shown in formula (2):

[0392] (2)

[0393] In the formula, G q (R) represents the displacement power spectral density of road surface unevenness, α R This is the gain coefficient of the dynamic load coupled to the tire suspension.

[0394] The input signals for this stage are the number of wheelsets n and the rated load capacity m of a single wheelset. rated Vehicle speed V, road surface grade R; the output signal is the theoretical maximum carrying capacity M of the vehicle. total .

[0395] Section Two: Multi-condition Carrying Capacity Calculation. This section calculates the carrying capacity of heavy-duty vehicles under various typical operating conditions, including constant speed on a straight road, acceleration on a straight road, starting on an incline, crawling on an incline, and driving on curves. It establishes the maximum permissible load capacity under constraints of traction force, resistance balance and adhesion, power, and stability. The carrying capacity M for constant speed driving on a straight road is also calculated. flat, const As shown in formula (3):

[0396] (3)

[0397] In the formula, μ is the road surface adhesion coefficient, ψ is the static load distribution ratio of the drive axle, and ρ a C is the density of air. D Let M be the air resistance coefficient, A be the frontal area, f be the rolling resistance coefficient, and g be the acceleration due to gravity. The carrying capacity M during acceleration on a straight road is... flat, acc As shown in formula (4):

[0398] (4)

[0399] In the formula, ψ' is the dynamic load coefficient of the drive shaft during acceleration, δ is the rotational mass conversion coefficient, and a is the target acceleration. The carrying capacity M during ramp start-up is... ramp, start As shown in formula (5);

[0400] (5)

[0401] In the formula, F max The maximum driving force is θ, where θ is the road gradient angle. The carrying capacity M for crawling on a slope is... ramp, climb As shown in formula (6):

[0402] (6)

[0403] In the formula, P max η is the maximum output power of the engine motor. t For the overall efficiency of the transmission system. Carrying capacity (M) when cornering. curve As shown in formula (7):

[0404] (7)

[0405] In the formula, F tire Δlat represents the maximum allowable vertical load on a single wheel, and Δlat is the percentage increase in wheel load during cornering.

[0406] The input signals for this stage are the road surface adhesion coefficient μ, the static load distribution ratio of the drive shaft ψ, and the air density ρ. a Air drag coefficient C D Frontal area A, rolling resistance coefficient f, dynamic load coefficient of drive shaft during acceleration ψ', rotational mass conversion coefficient δ, target acceleration a, maximum driving force of drive wheel F max Road slope angle θ, maximum output power of engine motor P max Overall efficiency η of the transmission system t Maximum allowable vertical load F per wheel tire The wheel load increment ratio Δlat in curves; the output signal is the carrying capacity M for the following driving conditions: constant speed on a straight road, acceleration on a straight road, hill start, hill crawl, and curve driving. flat, const M flat, acc M ramp, start M ramp, climb M curve .

[0407] Step 3: Determining the Carrying Capacity Boundary and Safety Margin. This step, based on the carrying capacity calculations for the five operating conditions in Step 2, takes the minimum allowable carrying mass for each operating condition as the lower limit boundary of the vehicle's carrying capacity, and takes the flat static benchmark value as the upper limit boundary, thus determining the safe carrying range of the entire vehicle [M]. min, safeM max, safe The carrying capacity safety margin SM is defined as shown in formula (8):

[0408] (8)

[0409] In the formula, M actual The actual carrying capacity is determined by the SM threshold. When the SM reaches a certain threshold, a carrying capacity over-limit warning is triggered.

[0410] The input signal for this stage is the carrying capacity M under the following conditions: constant speed on a flat road, acceleration on a flat road, starting on a slope, crawling on a slope, and driving on a curve. flat, const M flat, acc M ramp, start M ramp, climb M curve The output signal represents the safe carrying range of the entire vehicle [M]. min, safe M max, safe ] , Safety margin of carrying capacity SM.

[0411] Step 2: Modular Heavy-Duty Vehicle Dynamics Assessment. Step 2 consists of three parts.

[0412] Step 1: Multi-wheelset drive power matching calculation. This step calculates the total drive power required for three driving conditions—high-speed cruising, climbing, and acceleration—based on the vehicle's total mass, design maximum speed, and maximum climbing resistance. Then, the output power of each wheelset is allocated according to the axle load ratio of each drive wheel. The total power demand P at the vehicle's steady-state maximum speed is calculated. total, max As shown in formula (9):

[0413] (9)

[0414] In the formula, M v V represents the unloaded mass. max This represents the maximum vehicle speed. Power distribution P per wheel pair i As shown in formula (10):

[0415] (10)

[0416] In the formula, F i This refers to the vertical load on a single axle.

[0417] The input signal for this stage is the theoretical maximum carrying capacity M of the vehicle. total Unloaded mass M v Maximum speed V max air density ρ a Air drag coefficient C D Frontal area A, rolling resistance coefficient f, vertical load on a single axle F i The output signal is the total power demand P at the vehicle's steady-state maximum speed. total, maxPower distribution P of a single wheelset i .

[0418] Section Two: Acceleration and Climbing Ability Assessment. This section assesses the vehicle's acceleration and climbing ability based on its drive power and curb weight. Acceleration time t acc The calculation is shown in formula (11):

[0419] (11)

[0420] Maximum stable gradient α max The calculation is shown in formula (12):

[0421] (12)

[0422] In the formula, V c To maintain a stable speed while climbing the hill.

[0423] The input signal for this stage is the theoretical maximum carrying capacity M of the vehicle. total Unloaded mass M0, vehicle speed V, air density ρ a Air drag coefficient C D Frontal area A, rolling resistance coefficient f, rotational mass conversion factor δ, maximum output power of engine / motor P max Overall efficiency η of the transmission system t Stable climbing speed V c The output signal is the acceleration time t. acc Maximum stable gradient α max .

[0424] Step 3: Comprehensive Evaluation and Output of Dynamic Performance Indicators. This step integrates the calculation results from Step 1 and Step 2 to construct an evaluation system that includes the maximum vehicle speed V. max Maximum stable gradient α max Acceleration time t acc The dynamic performance evaluation system comprises three core indicators, and a weighted comprehensive score is applied based on five common operating conditions of heavy-duty vehicles. The analytic hierarchy process (AHP) is used, with the weight coefficient for each operating condition set as ω. i Finally, the integrated calculation yields the comprehensive power score S. power As shown in formula (13):

[0425] (13)

[0426] In the formula, λ1, λ2, and λ3 are weighting coefficients.

[0427] The input signal for this stage is the maximum vehicle speed V. max Maximum stable gradient α max Acceleration time t acc Weighting coefficient ω for operating conditionsi The output signal is the comprehensive power score S. power .

[0428] Step 3: Overall stability assessment of modular heavy-duty vehicles. Step 3 consists of three parts.

[0429] Step 1: Handling Stability Assessment. This step establishes a handling stability evaluation index system based on the steering characteristics of multi-axle heavy-duty vehicles. The understeer angle Us and yaw rate gain G are used as the evaluation indices. yr Lateral acceleration gain G la As a core evaluation indicator. For an n-axis vehicle, define l i Let be the longitudinal distance of the i-th axis relative to the sprung center of mass of the entire vehicle, and let l be the distance between the axles in front of the center of mass. i <0, the axle behind the center of mass satisfies l i >0, the first axis is the equivalent front axle, serving as the main steering input axis of the vehicle, and the set of rear axles is defined as Ωr={i|l i For all load-bearing axes and active steering driven axes behind the center of mass, {>0, 2≤i≤n}, a unique equivalent rear axle is constructed. The understeer angle Us is defined as the equivalent front axle tire slip angle α. f Equivalent rear axle tire slip angle α r The difference between the two and the lateral acceleration a y The ratio of the two values ​​is used to calculate the steady-state yaw rate gain G. yr As shown in formula (14):

[0430] (14)

[0431] In the formula, L eq For the equivalent wheelbase, K Us λ is the understeering coefficient. i These are the steering weighting coefficients for each axis. Lateral acceleration gain G. la As shown in formula (15):

[0432] (15)

[0433] The input signal for this stage is the front axle tire slip angle α. f Rear axle tire slip angle α r Lateral acceleration a y Equivalent wheelbase L eq Insufficient turning coefficient K Us Vehicle speed V; output signals are understeer angle Us and yaw rate gain G. yr Lateral acceleration gain G la .

[0434] Step Two: Braking Longitudinal Stability Assessment. This step addresses the longitudinal stability issue of multi-axle heavy-duty vehicles during braking, establishing evaluation indicators for braking performance and braking directional stability. The maximum braking deceleration *a* is used as the criterion. brake, max Full-load braking distance S brake Axle load transfer braking lock-up boundary LTR brake The maximum braking deceleration a is used as an evaluation metric. brake, max As shown in formula (16):

[0435] (16)

[0436] In the formula, μ brake η is the peak adhesion coefficient of the braking surface. disk The multi-axis braking force distribution balance coefficient is defined as follows: The multi-axis braking force distribution balance coefficient is constructed as shown in formula (17):

[0437] (17)

[0438] In the formula, F bi The braking force corresponding to the real-time braking pressure of each axle. This represents the average braking force across all brake axles of the vehicle. Fully loaded braking distance S brake As shown in formula (18):

[0439] (18)

[0440] In the formula, t lag v0 is the initial braking velocity, and v0 is the braking system reaction time. Axle load transfer braking lock-up boundary LTR brake As shown in formula (19):

[0441] (19)

[0442] In the formula, ΔF z F represents the dynamic load change during braking, which is equivalent to an increase in load on the front axle and a decrease in load on the rear axle. z, static h represents the average vertical load on the front and rear axles under static full load. c The height of the vehicle's center of gravity when fully loaded.

[0443] The input signal for this stage is the peak adhesion coefficient μ of the braking surface. brake Multi-axis braking force distribution balance coefficient η disk Braking system reaction time t lag Initial braking velocity v0, dynamic load change ΔF during braking process (increased load on the front axle, decreased load on the rear axle) z The height of the vehicle's center of gravity when fully loaded (h) c Static full load, average vertical load F between front and rear axles z, static Equivalent wheelbase Leq The output signal is the maximum braking deceleration a. brake, max Full-load braking distance S brake Axle load transfer braking lock-up boundary LTR brake .

[0444] Step 3: Comprehensive Evaluation of Braking Performance and Stability. This step integrates the evaluation results from Step 1 and Step 2 to construct a comprehensive vehicle stability score. stab As shown in formula (20):

[0445] (20)

[0446] In the formula, ω1, ω2, ω3, ω4, and ω5 are weighting coefficients, and std is the industry standard reference value.

[0447] The input signal for this stage is the yaw rate gain G. yr Lateral acceleration gain G la Maximum braking deceleration a brake, max Full-load braking distance S brake Axle load transfer braking lock-up boundary LTR brake The output signal is the overall vehicle stability score. stab .

[0448] Step 4: Modular heavy-duty vehicle passability assessment. Step 4 consists of three parts.

[0449] Section 1: Multi-Axle Geometric Passability Assessment for Heavy-Duty Vehicles. This section, based on the multi-span wheelbase structure of n-axle heavy-duty vehicles, establishes a multi-axle coupled geometric passability evaluation system adapted to ultra-long vehicle bodies and multi-wheelset arrangements, considering the unique failure modes such as mid-section overhang, low-lying areas between axles, and interference from multiple wheel sets. It also adds longitudinal passability characteristics within multi-axle sections, with core indicators including minimum ground clearance h. min Approach angle α approach Departure angle α depart Define the longitudinal pass angle α of multi-axis coupling. ramp Instead of using the traditional front and rear single axle constraint method, the weighted solution is based on the maximum span of adjacent axle groups and the minimum ground clearance of each segment, which characterizes the bottoming resistance of multi-span vehicle bodies when passing through uneven roads and continuous potholes, as shown in formula (21):

[0450] (twenty one)

[0451] In the formula, L wb, max This represents the maximum longitudinal distance between adjacent wheelsets.

[0452] The input signals for this stage are the ground clearance of each section of the heavy-duty multi-axle vehicle, the geometric parameters of the front end, the set parameters of the rear end, and the maximum longitudinal distance L between adjacent wheelsets. wb, max The output signal is the minimum ground clearance h. min Approach angle α approach Departure angle α depart Multi-axis coupling longitudinal through angle α ramp .

[0453] Step Two: Performance Evaluation of Multi-Axle Load-Diverting Flexible Supports. This step, based on the multi-wheel-pair uniform load distribution, load diversion, and multi-point flexible grounding mechanism of ultra-multi-axle heavy-duty vehicles, as well as the significantly reduced single-wheel grounding pressure, significantly reduced road subsidence, and traction adhesion characteristics that are completely different from ordinary vehicles, establishes a performance evaluation model for soft road surface supports adapted to multi-axle heavy-duty vehicles, introducing a multi-axle load uniformity correction coefficient η. load This study introduces parameters such as the order of the number of axles and the dispersion of loads on each axle to reflect the coupled influence of multi-point pressure distribution and load uniformity of n-axle vehicles on road passability. It also constructs multi-axle coupled traction coefficient TC, multi-axle cooperative traction efficiency TE, and multi-axle heavy-duty fuel utilization index E. f Evaluation system. Multiaxial load uniform distribution correction factor η load As shown in formula (22):

[0454] (twenty two)

[0455] In the formula, This represents the average vertical load across all axles of the vehicle. The multi-axle coupling traction coefficient TC is shown in formula (23):

[0456] (twenty three)

[0457] In the formula, F net The total vehicle hook traction force is given by equation (24). The multi-axle cooperative traction efficiency TE is shown in equation (24):

[0458] (twenty four)

[0459] In the formula, P out For output power, P in Input power. Multi-axle heavy-duty fuel utilization index E f As shown in formula (25):

[0460] (25)

[0461] The input signal for this stage is the real-time vertical load F of each axle of the vehicle. zi Average vertical load on all axles of the vehicle The traction force F of the whole vehicle hook net Total vehicle weight M totalg, Output power P out Input power P in The output signal is the multi-axis load distribution correction coefficient η. load Multi-axle coupling traction coefficient TC, multi-axle cooperative traction efficiency TE, multi-axle heavy-duty fuel utilization index E f .

[0462] Step 3: Comprehensive Passability Rating. This step establishes a hierarchical comprehensive rating mechanism that combines hard thresholds for geometric constraints with soft evaluation of support performance. This aligns with the actual failure logic of heavy-duty vehicles. First, it determines whether multi-axle geometric bottoming interference exceeds limits, then integrates the support and passage capacity of soft road surfaces to achieve a high-precision comprehensive passability assessment. A multi-axle geometric passability level R is constructed. geo As shown in formula (26):

[0463] (26)

[0464] In the formula, β1, β2, β3, and β4 are weighting coefficients, and ref is the reference value. The passability rating R of the multi-axis support is constructed. support As shown in formula (27):

[0465] (27)

[0466] In the formula, ζ1, ζ2, and ζ3 are weighting coefficients, and ref is the reference benchmark value. Therefore, the overall passability rating R... mobility As shown in formula (28):

[0467] (28)

[0468] In the formula, λ1 and λ2 are weighting coefficients.

[0469] The input signal for this stage is the minimum ground clearance h. min Approach angle α approach Departure angle α depart Longitudinal passing angle α ramp Traction coefficient TC, traction efficiency TE, fuel utilization index E f The output signal is the geometric passability level R. geo Support passability rating R support Passability Comprehensive Rating mobility .

[0470] The process of building the whole vehicle model in the second step is as follows:

[0471] Step 1: Vehicle Dynamics Modeling. The model established in this step is particularly suitable for the dynamic response of heavy-duty vehicles with 6 axles or more. A whole-vehicle dynamics model with coupling in the longitudinal, lateral, and vertical directions is established. The number of wheelsets is n, and the whole vehicle has 6+2n degrees of freedom, namely longitudinal translational degrees of freedom, lateral translational degrees of freedom, vertical translational degrees of freedom, yaw rotational degrees of freedom, roll rotational degrees of freedom, pitch rotational degrees of freedom, and 2n degrees of freedom of the tires. The vehicle dynamics equation set is established as shown in formula (29):

[0472] (29)

[0473] In the formula, M v For the overall vehicle weight, a x a y These are the longitudinal and lateral accelerations, F, respectively. xi F yi F represents the sum of the longitudinal and lateral forces of each wheel on the i-th axis. areo For air resistance, F r R is the rolling resistance, r is the yaw rate, and I is the rolling resistance. z I x I y These are the yaw, roll, and pitch moments of inertia of the entire vehicle, respectively. i Let t be the longitudinal distance from the i-th axis to the centroid. i Let ϕ be the wheelbase of the i-th axis, φ be the roll angle, and c be the wheelbase of the yoke. ϕ k ϕ These are the roll damping coefficient and pitch stiffness coefficient, respectively, θ p c is the pitch angle. θp k θp These are the pitch damping coefficient and pitch stiffness coefficient, respectively. s F is the vertical displacement of the vehicle's center of gravity. zsi Let X be the vertical force exerted by the suspension on the vehicle body along the i-th axis. The resulting state vector of the entire vehicle system is X. body and its state derivative .

[0474] The input signal for this step is the vehicle structural parameter, the vehicle mass M. v The moment of inertia of the whole vehicle in terms of yaw, roll, and pitch I z I x I y Center of mass height h c Distance from each axis to the centroid l i Wheelbase t i The output signal is the vehicle system state vector X. body Vehicle system state derivative .

[0475] Step 2: Wheel dynamics modeling. Step 2 consists of three steps.

[0476] Step 1: Modeling the Vertical Characteristics of a Single-Wheel Tire. This step establishes a model of the vertical force-deformation relationship between the tires of each wheelset based on the tire's vertical stiffness and damping characteristics. The vertical force F of a single-wheel tire... zti As shown in formula (30):

[0477] (30)

[0478] In the formula, k ti For vertical stiffness, z ui For the vertical displacement of the wheel center, z ri For the vertical displacement of the road surface excitation, c ti This is the tire's vertical damping coefficient. The vertical velocity of the wheel center. The vertical excitation velocity of the road surface.

[0479] The input signal for this stage is the vertical stiffness k. ti Vertical displacement z of wheel center ui , road surface excitation vertical displacement z ri Tire vertical damping coefficient c ti Vertical velocity of wheel center Vertical excitation velocity of the road surface The output signal is the vertical force F of a single tire. zti .

[0480] Section 2: Tire Dynamics Integration Model for Multi-Axle Heavy-Duty Vehicles. This section, based on the single-wheel model in Section 1, systematically integrates the vertical, longitudinal, and lateral forces of the tires across all n wheelsets of the vehicle to establish a tire dynamics integration model for multi-axle heavy-duty vehicles. The core coupling characteristic parameters of multi-axle heavy-duty vehicles are defined, and the unique mechanical interference effects of ultra-multi-axle vehicles are quantified, as shown in formula (31):

[0481] (31)

[0482] In the formula, η Fz, i To account for the unevenness of load distribution across shafts, ξ ij σ represents the mechanical interference coefficients along the i-th and j-th axes. The closer the axial distance, the stronger the coupling interference. slip The non-synchronous distortion coefficient for multi-axle slip is a unique phenomenon of heavy-load vehicles with inconsistent wheel slip states. An integrated equation for tire dynamics of heavy-load vehicles is constructed, and full-domain coupling superposition is achieved through inter-axle interference characteristics, reflecting the cluster distortion characteristics of multi-axle heavy-load tire mechanics, as shown in equation (32):

[0483] (32)

[0484] The input signal for this stage is the vertical force F of a single tire. ztiUnevenness of load on each axis η Fz, i The mechanical interference coefficients ξ of the i-th and j-th axes ij Multi-axis slip asynchronous distortion coefficient σ slip Total longitudinal tire force F of each axle xi Total lateral tire force F of each axle yi The output signal is the total longitudinal tire force F of the multi-axis cluster coupling. x, sum Multi-axis cluster coupled total lateral tire force F y, sum .

[0485] Step 3: Multi-Axle Heavy-Duty Tire Temperature-Wear Coupling Modeling. This step addresses the unique characteristics of multi-axle heavy-duty vehicles, which operate under conditions of high load cluster pressure, continuous multi-wheel slippage, and prolonged continuous operation. These conditions result in significantly greater tire heat accumulation and wear degradation compared to ordinary vehicles. Furthermore, the coupled friction between the wheels of multiple axles exacerbates the differentiated evolution of temperature loss. To address this, a tire heat generation-dissipation balance equation adapted to ultra-multi-axle heavy-duty conditions is established. This equation couples tire temperature changes to tire peak force and stiffness parameters. Simultaneously, a wear rate calculation model is constructed to dynamically correct tire mechanical parameters based on temperature and wear degree. Tire heat generation power P heat As shown in formula (33):

[0486] (33)

[0487] In the formula, V slip, x V slip, y F represents the longitudinal and lateral slip velocities, respectively. xi For the total longitudinal tire force of each axle, F yi Total lateral tire force across all axles. Based on the characteristics of accumulated heat generation and uneven heat dissipation under heavy multi-axle loads, a dynamic temperature rise balance equation for multi-axle coupling is established. Tires in ultra-multi-axle vehicles do not experience independent temperature rises; clustered thermal radiation, close-range wheel-to-wheel thermal coupling, and continuous heat accumulation under heavy loads all contribute to the effect. Therefore, a multi-axle heat accumulation coefficient ε is introduced. c The dynamic change of tire temperature is shown in formula (34):

[0488] (34)

[0489] In the formula, C th Here, T represents the tire's heat capacity, and h represents the tire's real-time temperature. conv T is the convective heat transfer coefficient. amb For ambient temperature, A s This refers to the effective heat dissipation area of ​​the tire. Tire wear rate W rate As shown in formula (35):

[0490] (35)

[0491] In the formula, K wear E is the wear coefficient. aR is the activation energy, and R is the gas constant.

[0492] The input signals for this stage are the longitudinal and lateral slip velocities V. slip, x V slip, y Total longitudinal tire force F of each axle xi Total lateral tire force F of each axle yi Tire temperature T; the output signal is the tire heat generation power P. heat Tire heat capacity C th Ambient temperature T amb Tire wear rate W rate .

[0493] Step 3: Multi-round kinematic modeling. Step 3 consists of three parts.

[0494] Step 1: Establishing Multi-Axis Coupled Kinematic Constraints. This step establishes the kinematic constraint relationships between each wheelset and between the wheelsets and the vehicle body, providing constraint boundary conditions for subsequent multi-body motion state solutions. Each wheelset is connected to the vehicle body through the suspension system, and the suspension kinematic constraints are embodied in the wheel alignment parameter, kingpin inclination angle γ. kpi Kingpin inclination angle γ kpc Wheel camber angle γ camber γ front toe angle toe The dynamic relationship between suspension travel and suspension travel, let the vertical runout of the i-th axle be Δz. si The changes in the camber angle and toe angle of the wheel on this axle are shown in formula (36):

[0495] (36)

[0496] In the formula, γ camber, 0 γ toe, 0 To design the positioning angle, k camber k toe These are the suspension geometric kinematic parameters.

[0497] Adjacent wheelsets are coupled and constrained by the frame structure. For multi-axle rigid frame vehicles, the displacement of each axle relative to the frame is constrained by frame deformation, generating an inter-axle displacement deviation ΔP. i, i-1 As shown in formula (37):

[0498] (37)

[0499] In the formula, L i, i-1 Let e ​​be the initial axis distance between the i-th axis and the (i-1)-th axis. x For the vehicle's longitudinal unit vector, kinematic interference occurs when the inter-axle displacement deviation exceeds the allowable range, requiring constraint compensation. Based on the attitude constraints of each wheelset and the coupling relationship between the inter-axle pose, the instantaneous equivalent rotation angle δ of the entire vehicle with multiple wheelsets adapted to multi-axle vehicles is derived. insAs shown in formula (38):

[0500] (38)

[0501] The input signal for this stage is the kingpin inclination angle γ. kpi Kingpin inclination angle γ kpc Wheel camber angle γ camber γ front toe angle toe The vertical runout of the frame is Δz si The output signal is the inter-shaft displacement deviation ΔP. i, i-1 Multi-wheel instantaneous equivalent steering angle δ of the whole vehicle ins

[0502] Step Two: Kinematic Description of the Vehicle Multibody System. This step builds upon the suspension geometric constraints and multi-axle frame coupling constraints established in Step One. It abstracts the entire vehicle as a multibody system consisting of the vehicle body and multiple wheel pairs, achieving an accurate solution for the vehicle's multibody motion state. Considering the structural characteristics of ultra-multi-axle heavy-duty vehicles with 10 or more axles, the entire vehicle is viewed as a multibody system composed of the body and multiple wheel pairs. The vehicle's motion state is determined by the position of its center of mass P. G The yaw angle φ, pitch angle θp, roll angle ϕ, and the local motion state of each wheel pair are used to describe the vehicle. A vehicle coordinate system Oxyz is defined, with the origin O located at the projection of the vehicle's center of mass onto the horizontal plane. The x-axis points along the vehicle's longitudinal direction in the forward direction, the y-axis points to the driver's left, and the z-axis is perpendicular to the ground and points upwards. For the i-th axis and j-th wheel, the contact point position vector P... ij As shown in formula (39):

[0503] (39)

[0504] In the formula, R G Let r be the vehicle body attitude rotation matrix. ij The effective rolling radius is given by R. The vehicle body attitude rotation matrix is ​​also given by R. G The ZYX Euler angles are defined in the order shown in formula (40):

[0505] (40)

[0506] In the formula, R x R y R z These are the basic rotation matrices about the x, y, and z axes, respectively. Differentiating the contact point position vectors of each wheel yields the velocity vectors of each wheel's contact point. As shown in (41):

[0507] (41)

[0508] The input signal for this stage is the distance l from each axis to the centroid. i Wheelbase ti Location of the center of mass P G Yaw angle φ, pitch angle θp, roll angle ϕ, vehicle attitude rotation matrix R G Effective rolling radius r ij The output signal is the grounding point position vector P. ij Grounding point velocity vector .

[0509] Step 3: Integration of Augmented Kinematic Model of the Whole Vehicle. This step integrates the multi-axis coupled kinematic constraints of Step 1 with the multi-body system kinematic description of the whole vehicle in Step 2, and fuses the body dynamics state of Step 1 with the wheel dynamics state of Step 2 to construct a control-oriented multi-wheelset augmented kinematic model of the whole vehicle. Generalized coordinates q∈R are selected. 3n+6 , representing all kinematic degrees of freedom of the entire vehicle, based on the kinematic description of link one and the constraint relationship of link two, the augmented kinematic mapping matrix S(q) is constructed as shown in formula (42):

[0510] (42)

[0511] In the formula, S body Map the dynamic parameters output in step one to the derivative of the vehicle body's generalized coordinates, S susp S represents the suspension kinematic constraints of the mapping. steer The turning angular velocity is mapped to the derivative of the generalized coordinates. A null space basis V2 is designed, and a reduced-order kinematic mapping matrix A is constructed. red (q)=V2·S(q), and the reduced-order augmented kinematic model is obtained as shown in formula (43):

[0512] (43)

[0513] To embed the dynamic equations of step one and the tire force balance equations of step two into the kinematic model, a coupling constraint vector Ψ is defined as shown in formula (44):

[0514] (44)

[0515] In the formula, q body Let be the vehicle body pose vector. For vehicle speed, Let C be the vehicle body acceleration, C be the system damping matrix, K be the system stiffness matrix, and F be the system acceleration. step1 F is a generalized force vector. ij The actual tire force is denoted as . The constraint vector is differentiated with respect to time and transformed into a linear constraint at the generalized acceleration level. This constraint is then combined with the reduced-order kinematic model to obtain the standard discrete-time augmented kinematic model for control, as shown in equation (45):

[0516] (45)

[0517] In the formula, q(k+1) is the system generalized coordinate vector of the next discrete time step. This model fully describes the kinematic coupling relationship of all wheelsets of a multi-axle heavy-duty vehicle with more than 10 axles, and realizes bidirectional deep coupling with the vehicle dynamics state in step one and the wheel dynamics state in step two. Its output state vector q(k) provides all the necessary kinematic feedback for the subsequent multi-wheelset system trajectory following controller.

[0518] The input signals for this stage are the generalized coordinate q and the generalized force vector F. stepl Vertical displacement of the vehicle body z s Mass matrix M v Damping matrix C, stiffness matrix K, tire force F ij Tire slip angle α ij Slip ratio λ ij The output signals are the augmented kinematic mapping matrix S(q) and the reduced kinematic mapping matrix A. red (q), coupling constraint vector Ψ, and the system generalized coordinate vector q(k+1) for the next discrete time step.

[0519] Step 4: Steering Motor Modeling. Step 4 consists of three steps.

[0520] Step 1: Establishment of the Electromagnetic Model of the Steering Motor. Based on the performance specifications of multi-axle heavy-duty vehicles for fast response, low hysteresis, and large speed range of the steering system, this step establishes the electromagnetic dynamic mathematical model of the steering actuator motor. The mathematical model is adopted in the dq coordinate system, and the stator voltage equation is shown in formula (46):

[0521] (46)

[0522] In the formula, u d u q These are the stator voltages along the d-axis and q-axis, respectively, R s i is the equivalent resistance of the stator winding. d i q These are the stator currents along the d-axis and q-axis, respectively. d L q These are the equivalent inductances along the d-axis and q-axis, respectively, and ω e φ is the electric angular velocity of the motor. f The magnetic flux linkage is a permanent magnet. The electromagnetic torque T of the motor is... e As shown in formula (47):

[0523] (47)

[0524] In the formula, p n It is an extreme logarithm.

[0525] The input signals for this stage are the stator currents i along the d-axis and q-axis. d i q Stator winding equivalent resistance R s d-axis and q-axis equivalent inductance L d L q ω, the electric angular velocity of the motor e Permanent magnet flux linkage φ f The output signal is the d-axis and q-axis stator voltage u. d u q Motor electromagnetic torque T e .

[0526] Step 2: Mechanical Dynamics Modeling of the Steering Motor. This step establishes the mechanical dynamics equations of the steering motor, characterizing the dynamic relationship between the motor's output torque and its rotational angle and speed. The mechanical dynamics equilibrium equations of the motor are shown in formula (48):

[0527] (48)

[0528] In the formula, J m Let ω be the moment of inertia of the motor rotor. m B is the mechanical angular velocity of the motor. m T is the mechanical damping coefficient. L This represents the load torque.

[0529] The input signal for this stage is the rotor inertia J of the motor. m ω, mechanical angular velocity of the motor m Mechanical damping coefficient B m Load torque T L The output signal is the motor electromagnetic torque T. e .

[0530] Step 3: Dynamic Modeling of Steering System Based on Multi-Motor and Multi-Degree-of-Freedom Model. This step couples the steering motor model with the steering actuator to establish a complete transmission model from motor torque to wheel steering angle. The steering wheel dynamic equation is shown in formula (49):

[0531] (49)

[0532] In the formula, J sw δ is the equivalent rotational inertia of the steering wheel and steering column. sw B is the steering wheel angle. sw τ is the equivalent damping coefficient of the steering column. f, sw T is the coulomb friction torque of the steering column. driver The input torque applied to the driver, T column The torque transmitted from the steering column to the steering gear. For the i-th axis, the steering angles δ of the left and right wheels. i, L δi, R The Ackermann steering geometry constraints are satisfied as shown in formula (50):

[0533] (50)

[0534] In the formula, t i L represents the wheelbase of each axle. i, axle Let be the distance from the instantaneous center of rotation of each axle to the axle. For multi-axle steering vehicles, there is a definite distribution relationship between the steering angles of each steering axle, and the kinematic constraints of adjacent axles are shown in formula (51):

[0535] (51)

[0536] In the formula, δ i, i-1 The steering angle distribution coefficient is given. The dynamics of the steering actuator are determined by the motor driving torque, equivalent inertia, damping, friction, and tire self-centering torque. By combining the motor mechanical dynamics equations and the steering actuator dynamics balance equations, a dynamic model of the multi-motor, multi-degree-of-freedom steering system is obtained, as shown in formula (52):

[0537] (52)

[0538] In the formula, G r J is the steering gear reduction ratio. eq For the equivalent rotational inertia of the steering actuator, B eq τ is the equivalent damping coefficient of the steering actuator. f Let τ be the Coulomb friction torque. align δ is the tire return torque. w This refers to the wheel steering angle.

[0539] The input signal for this stage is the rotor inertia J of the motor. m Steering gear reduction ratio G r Equivalent moment of inertia J of steering actuator eq Wheel steering angle δ w Coulomb friction torque τ f Tire return torque τ align Mechanical damping coefficient B m The output signal is the motor electromagnetic torque T. e .

[0540] The process of designing the multi-round integrated steering control strategy in the third step is as follows:

[0541] Step 1: Determining the vehicle's steering intention. Step 1 consists of three steps.

[0542] Step 1: Traffic Situation Awareness and Scene Prediction. This step utilizes multi-source fusion information from cameras, millimeter-wave radar, lidar, and high-precision maps to construct a traffic scene state space adapted for ultra-long, multi-axle, heavy-duty vehicles. It extracts high-dimensional scene feature vectors containing vehicle motion, road geometry, and traffic density, and introduces a sliding window temporal prediction algorithm to achieve accurate rolling predictions of future short-term driving scenarios, providing a priori environmental temporal sequence for subsequent risk identification. The multi-axle heavy-duty scene state vector S is constructed. scene (t)=[V ego V target , d target , ρ road , λ traffic ] T A fixed-length sliding time window is used to extract continuous time series samples, and the scene time series matrix within the window is constructed as shown in formula (53):

[0543] (53)

[0544] In the formula, T w The duration is the sliding window length. Based on the temporal state matrix, a first-order temporal evolution prediction model is constructed, outputting a predicted sequence of future scenes in the time domain. As shown in formula (54):

[0545] (54)

[0546] In the formula, A1 and B1 are time-series prediction weight matrices, which are updated in real time through rolling iteration.

[0547] The input signal for this stage is the vehicle speed V. ego Target speed V target Relative distance d target Road curvature ρ load Traffic density λ traffic The output signal is the scene state vector S. scene (t), Road Environment Prediction Sequence .

[0548] Step Two: Driving Risk State Estimation. This step uses the scenario prediction sequence output from Step One as environmental input, integrating a recursive least squares system identification and a temporal fuzzy C-means artificial intelligence clustering dual-core algorithm to identify time-varying lateral dynamic parameters and road adhesion parameters of multi-axle vehicles online. Combined with the unique risk characteristics of multi-axle heavy loads, it achieves adaptive driving risk quantification without fixed thresholds. The least squares algorithm is used to identify key dynamic parameters such as road adhesion coefficient and lateral disturbance stiffness online, constructing a high-dimensional clustering feature vector X that includes environmental risk, stability risk, and multi-axle cooperative risk. risk (t), as shown in formula (55):

[0549] (55)

[0550] In the formula, μ est To identify the road surface adhesion coefficient using the least squares algorithm, δ des, i -δ act, i To address steering synchronization deviation in multi-axle wheelsets, a temporal fuzzy C-means artificial intelligence clustering objective function J is constructed. FCM Strengthen the weight of the current state and suppress the interference of overloaded state lag, as shown in formula (56);

[0551] (56)

[0552] In the formula, k is the cluster category number, α(t) is the temporal forgetting weight coefficient, and u kt Where m is the membership degree and v is the fuzzy coefficient. k Let be the cluster center vector of the k-th class. By iteratively updating the membership degree and cluster center, the driving conditions are adaptively divided into three categories: safe, warning, and high-risk, and the membership degree u of the high-risk condition is output. risk By combining cluster membership degree and three types of risk indicators, a dedicated driving risk cost function R for multi-axle heavy-load vehicles is constructed. risk As shown in formula (57):

[0553] (57)

[0554] The input signal for this step is the road environment prediction sequence. Real-time lateral load transfer rate (LTR), target steering angle δ of each wheelset des Actual steering angle δ of each wheel pair act Real-time minimum distance d to obstacles target, min The output signal is a high-dimensional clustering feature vector X. risk (t), membership degree of high-risk working conditions u risk Driving risk cost function R risk .

[0555] Step 3: Comprehensive Determination of Steering Intent. This step is based on the driving risk state estimation results from Step 2, combined with the driver's input commands and the planning commands δ from the autonomous driving decision module. cmd, auto The overall vehicle steering intention is determined by comprehensive analysis. The determination logic is as follows: when R... risk ≥R th, high Active steering intervention is triggered when an effective obstacle avoidance path exists; when |δ w |≥|δ th |and ≥ When δ is determined to be the driver actively steering; cmd, auto≠0 and when the autonomous driving mode is activated, the autonomous driving planning instruction is used, with the priority being safety emergency intervention, driver input, and autonomous driving instruction. The steering intention I is output based on the judgment logic and the actual input. steer .

[0556] The input signal for this stage is the driving risk cost function R. risk Driver inputs operation commands and planning commands. cmd, auto The output signal is the steering intention I. steer .

[0557] Step 2: Vehicle trajectory planning strategy. Step 2 consists of three parts.

[0558] Step 1: Initial Trajectory Generation. This step is based on the turning intention I determined in Step 1. steer With scene state vector S scene (t) To address the unique characteristics of multi-axle heavy-duty vehicles, such as long vehicle body, large turning radius, coupling constraints of multiple wheel pair turning angles, wide overall vehicle path envelope, and inconsistent motion trajectories of each axle's contact point, this paper improves the traditional RRT random tree algorithm to form an enhanced RRT planning algorithm adapted for heavy loads. This algorithm introduces multi-wheel contact point envelope constraints throughout the sampling, node expansion, and collision detection processes. Simultaneously, it verifies whether the contact points of all n axles' wheels intrude into obstacles and lane boundaries. The algorithm sequentially completes trajectory sampling, node expansion, cost evaluation, multi-wheel pair envelope collision detection, and path backtracking, generating multiple discrete initial candidate trajectory clusters that satisfy the kinematic constraints of the multi-axle vehicle. The enhanced RRT algorithm introduces an adaptive bias sampling strategy with a bias probability p. bias The dynamic adjustment with the number of iterations is shown in formula (58):

[0559] (58)

[0560] In the formula, k is the current iteration number, and τ bias Let C be the time constant for the bias rate growth. During the expansion phase, five candidate nodes are generated in a single sampling, and the cost function C(x) is selected as shown in formula (59):

[0561] (59)

[0562] In the formula, g(x) is the actual path cost from the starting point to the current node, h(x) is the heuristically estimated cost, and o(x) is the obstacle risk cost.

[0563] Add hard kinematic constraints to the boundaries of the whole vehicle, prohibit sampling beyond the maximum steering angle of a single wheel and the left and right limit lateral coordinates of the lane, and add lateral safety constraints to the envelope of the multi-axle whole vehicle. The set of constraint inequalities is shown in formula (60):

[0564] (60)

[0565] In the formula, δ max For the single-wheel limit steering angle, y lane, min y lane, max W represents the left and right lateral boundaries of the lane. ve For the width of the entire vehicle body, y obs Let d be the lateral position of the obstacle. target, y To minimize the lateral safety distance between trajectory points and obstacles, constraints are also imposed on the grounding point positions of all n axles to prevent scraping of the middle and rear wheels of ultra-long multi-axle vehicles. For the discrete node sequence obtained by RRT sampling, a fifth-order polynomial is used as the basis function for the trajectory curve to construct a continuous time-domain analytical expression for a single candidate trajectory, representing the longitudinal curve x... traj Horizontal curve y traj As shown in formula (61):

[0566] (61)

[0567] In the formula, t∈[0, T] p [ ] represents the planning time, and a0~a5 and b0~b5 are the polynomial coefficients of the longitudinal trajectory and the lateral trajectory, respectively. The coefficients are determined by the initial and final positions and the initial and final velocities of the boundary conditions output by RRT.

[0568] The input signal for this stage is steering intention I. steer Scene state vector S scene (t), cost function C(x); output signal is bias probability p bias Single-wheel limit steering angle δ max , the left and right lateral boundaries of the lane y lane, min y lane, max Trajectory longitudinal curve x traj Horizontal curve y traj .

[0569] Step 2: Trajectory Post-processing and Smoothing Optimization. This step performs post-processing and smoothing optimization on the initial trajectory generated in Step 1, eliminating unnecessary nodes and sharp corners, and generating a smooth trajectory that meets the kinematic and comfort requirements of heavy-duty vehicles. Due to the large suspension stiffness and mechanical limitations of the steering actuator in multi-axle heavy-duty vehicles, the severe fluctuations in trajectory curvature will directly cause frequent changes in the steering angle of each wheel pair, resulting in frequent impacts on the steering motor and an increased risk of vehicle rollover. Using the polynomial trajectory output in Step 1 as the initial value, a multi-objective cost function that takes into account both tracking accuracy and the smoothness of multi-wheel pair steering execution is constructed. The polynomial coefficients are iteratively corrected, and the smoothing optimization objective function is shown in Formula (62):

[0570] (62)

[0571] In the formula, ωa ω k ω δ These are the penalty weights for lateral acceleration, trajectory curvature change, and wheelset angle change, respectively. For the lateral acceleration of the trajectory, Let d be the time derivative of the trajectory curvature. δi, ref / dt represents the rate of change of the reference steering angle for the i-th wheel pair, and is a penalty term for multi-axle heavy-duty vehicles. The optimization solution updates the polynomial coefficients, outputting the corrected continuous and smooth trajectory curve T. smooth This ensures that position, velocity, and acceleration are continuously differentiable.

[0572] The input signal for this stage is the longitudinal curve x of the trajectory. traj Horizontal curve y traj The output signal is the corrected, continuous, smooth trajectory curve T. smooth .

[0573] Step 3: Trajectory Feasibility Verification and Safety Assessment. This step verifies the smooth trajectory curve T generated in Step 2. smooth Conduct multi-dimensional feasibility verification and safety assessment, and smooth the trajectory curve T. smooth Substituting the multi-axle vehicle dynamic coupling model built in the second step, we perform dynamic feasibility verification, kinematic feasibility verification, and full-domain safety constraint verification. If the verification finds that the trajectory has an over-limit risk, we locate the over-limit time period and call the enhanced RRT algorithm in step one to locally regenerate the trajectory until the safety constraints are met in all time domains. Finally, we output the unique optimal reference trajectory T with the highest safety score. final .

[0574] The input signal for this stage is the smooth trajectory curve T. smooth The output signal is the globally optimal vehicle reference trajectory T after verification and correction. final .

[0575] Step 3: Multi-round system trajectory following control strategy. Step 3 is completed in three stages.

[0576] Step 1: Augmented State Space Modeling of Multi-Wheelset Systems. This step builds upon the multi-wheelset vehicle model established in Step 2, considering the large inertia transient lag of multi-axle heavy-duty vehicles, the time-varying nature of heavy-load parameters, and the dynamic coupling disturbance characteristics of multi-wheelsets, to construct an augmented state space model adapted for transient trajectory following control. The state vector x = [Y, V, r, φ, ϕ, δ] is selected. i , …,δ n ] T All are inherited from the output of the augmented kinematics model in the second step, with the control input u=[δ 1, cmd , …, δ n, cmd , ΔT1, …,ΔT n ] TThis includes steering commands for each wheelset and compensation for transient torque under heavy load. The nonlinear state-space equation with transient disturbance correction is established as shown in equation (63):

[0577] (63)

[0578] In the formula, d tran (t) represents the transient coupling disturbance term under multi-axis heavy load. The globally optimal vehicle reference trajectory T output in step two is... final The trajectory parameters in the vector are used as reference inputs, and the augmented state vector x is used as the reference input. aug The integral term including the trajectory tracking error is used to eliminate the steady-state error, as shown in Equation (64):

[0579] (64)

[0580] The input signals for this stage are the output state variable x of the multi-wheel-pair augmented kinematic model of the whole vehicle, the control input u, and the globally optimal whole vehicle reference trajectory T. final The trajectory parameters are in the vector; the output signal is the augmented state vector x. aug .

[0581] Section Two: Layered Sliding Mode Variable Structure Trajectory Following Controller. This section, based on the augmented state-space model from Section One, designs a layered sliding mode variable structure controller to achieve robust following of the reference trajectory for the multi-wheelset system. First, a dynamic coordination layer—stability boundary delineation and priority weight generation—is designed. The stability boundary of the multi-axle heavy-duty vehicle is delineated using the phase plane method, defining the phase plane state variable as the centroid sideslip angle β. p angular velocity of the center of mass deflection The boundary equation of the phase plane stable domain is shown in equation (65):

[0582] (65)

[0583] In the formula, c1 and c2 are boundary coefficients related to vehicle speed and road surface adhesion coefficient. The zero-moment point roll instability judgment model position P is introduced. ZMP As shown in formula (66):

[0584] (66)

[0585] When P ZMP When the vehicle exceeds the boundary of the vehicle support polygon, it is determined to be a roll instability. Based on phase plane stability domain division and ZMP roll risk assessment, a priority weighting factor ρ is dynamically generated. prior As shown in formula (67):

[0586] (67)

[0587] In the formula, sat is the saturation function, and ρprior Constrained in the interval [0, 1], β crit , P ZMP, max These are the sideslip angle, sideslip velocity, and ZMP position critical threshold, respectively. A variable structure control layer—multi-constraint sliding mode control law design—is presented, defining the tracking error sliding surface s. track As shown in formula (68):

[0588] (68)

[0589] In the formula, λ1~λ3 are the strictly positive definite sliding surface parameters. The stability sliding surface s is defined. stab As shown in formula (69):

[0590] (69)

[0591] In the formula, λ4~λ7 are strictly positive definite sliding surface parameters, determined by the pole placement method to ensure fast convergence and good damping characteristics of the sliding mode dynamics. The composite sliding surface S is constructed as shown in formula (70):

[0592] (70)

[0593] The superspiral approach law is used to weaken chattering while ensuring finite-time convergence, as shown in formula (71):

[0594] (71)

[0595] In the formula, k1 and k2 are strictly positive definite reaching rate parameters, based on the synovial reachability condition. =0, combining the first step vehicle dynamics equations and the second step tire model, the inverse solution yields the desired generalized torque ν. des The design incorporates a constraint handling layer—multi-constraint boundary processing and multi-wheel pair cooperative error correction. For multiple constraint conditions, a combination of boundary layer and saturation function is used to handle actuator saturation and state constraints. The boundary thickness T is defined. b Adaptive adjustment, the saturation function is sat(s, T) b Define the comprehensive error E of multi-round transient cooperative tracking. syns As shown in formula (72):

[0596] (72)

[0597] The comprehensive error E syns As a feedback correction, it is added to the desired generalized torque ν des The closed-loop correction is formed as shown in formula (73):

[0598] (73)

[0599] In the formula, K sync For positive definite cooperative correction gain matrix, ν des, mod It is the desired generalized moment vector after co-correction.

[0600] The input signal for this stage is the augmented state vector x. aug Boundary coefficients c1 and c2 related to vehicle speed and road surface adhesion coefficient, and the position P of the roll instability judgment model. ZMP Tracking error sliding surface s track Stability of sliding surface s stab Comprehensive error E syn The output signal is the priority weighting factor ρ. prior Synthetic sliding surface S, desired generalized torque ν des The desired generalized moment vector ν after collaborative correction des, mod .

[0601] Step 3: Multi-actuator control allocation and follow-up execution. This step will assign the desired generalized torque ν output from Step 2. des The control is assigned to the steering actuators of each wheelset, and rolling time-domain updates and fault-tolerant reconfiguration are implemented. The control efficiency matrix M is defined. c Its elements are determined by the Jacobian matrix of the steering angle of each wheel pair and the tire force on the generalized force of the whole vehicle in the vehicle dynamics model of step one. The optimization problem J alloc As shown in formula (74):

[0602] (74)

[0603] In the formula, u act Let Q be the actual output vector of the actuator, R be the weight matrix for allocation accuracy, R be the weight matrix for load balancing, RΔ be the weight matrix for motion smoothing, and u be the weight matrix for motion smoothing. prev W is the actuator output vector at the previous time step, and W is the cumulative workload vector of each actuator. avg This represents the average load. For multi-axle heavy-duty vehicles, the vertical load F on each axle during transient steering is... zi Drastic dynamic changes directly affect the control efficiency of each actuator. Therefore, a transient load correction factor η is introduced into the control efficiency matrix. z, i (t), as shown in formula (75):

[0604] (75)

[0605] Corrected control efficiency matrix As shown in formula (76):

[0606] (76)

[0607] When a portion of the wheelset steering actuators fails, the failure detection and isolation module continuously monitors the residual error. i (k), when |ϵ i When (k)|≥ϵth and continues for more than 5 sampling periods, it is considered a failure, and the fault tolerance allocation problem J fault As shown in formula (77):

[0608] (77)

[0609] In the formula, M fail To reduce the order of the control efficiency matrix after removing the column corresponding to the failed actuator, u act, fail F represents the actual output vector of the remaining healthy actuators under failure conditions. act For the set of failed actuators, U fail This represents the feasible constraint set for the remaining healthy actuators under failure conditions. Each sampling time T... s Based on the latest state measurements, the sliding mode control law and control assignment problem are resolved to achieve rolling updates, and the desired steering angle command δ for each wheelset is re-output. i, cmd This forms a complete closed loop of "perception-decision-allocation-execution-feedback", enabling multi-round pair systems to achieve high-precision collaborative following of the reference trajectory.

[0610] The input signal for this stage is the desired generalized torque ν. des Control efficiency matrix M c The actual output vector u of the actuator act The executor output vector u at the previous time step prev The output signal is the actual output vector u of the remaining healthy actuator under failure conditions. act, fail Failed actuator set F act The feasible constraint set U for the remaining healthy actuators under effective operating conditions fail δ, the desired steering angle command for each wheelset i, cmd .

[0611] Step 4: Multi-target following control strategy for the motor. Step 4 consists of three parts.

[0612] Stage 1: Steering Angle Loop Following Control. This stage is the outermost closed loop of the motor actuator system, precisely tracking the desired steering angle command δ assigned to each wheel pair by the upper-level multi-wheel pair system. i, cmd A closed-loop tracking controller for the upper-level rotation angle position is designed to achieve precise tracking of the motor rotor position to the target rotation angle. The target rotation angle θ of the motor is... m, ref =δ i, cmd / Gr, the position loop adopts a proportional-integral-derivative control framework, and introduces model-based feedforward compensation and a reduced-order extended state observer to improve dynamic response speed and steady-state accuracy. The control law is shown in equation (78):

[0613] (78)

[0614] In the formula, ω m, ref e is the target angular velocity output by the position loop. θ K represents the position tracking error. p, pos K i, pos K d, pos These are the position loop proportional, integral, and differential gain coefficients, respectively. The rate of change of error, For feedforward velocity compensation term, This is the total perturbation compensation for the position loop estimated by the reduced-order extended state observer. The reduced-order extended state observer is of second-order form, as shown in equation (79):

[0615] (79)

[0616] In the formula, This is a location estimate. For the total disturbance estimate, β z1 β z2 For observer gain, θ m This represents the actual electrical angle of the motor rotor.

[0617] The input signal for this stage is the desired steering angle command δ for each wheelset. i, cmd Steering gear reduction ratio Gr, position loop proportional, integral, and differential gain coefficient K p, pos K i, pos K d, pos Actual rotor electrical angle θ of the motor m The output signal is the target angular velocity ω output by the position loop. m, ref Position tracking error e θ Error change rate .

[0618] Step 2: Smooth Closed-Loop Control of Motor Speed. This step is based on the target angular velocity ω output from the position loop. m, ref A mid-level speed smoothing closed-loop controller is designed to achieve rapid and smooth tracking of the motor's actual speed to the target speed. The speed loop employs a proportional-integral control law and introduces a first-order low-pass filter ω. LPF / (s+ω LPF ), filtering out velocity measurement noise, filter cutoff angular frequency ω LPF The selection should balance noise suppression effect and phase lag, and should be 5 to 10 times the bandwidth of the velocity loop. The phase lag of the filter at the velocity loop crossing frequency can be offset by acceleration feedforward compensation.

[0619] Introducing acceleration feedforward term As shown in formula (80):

[0620] (80)

[0621] In the formula, ω m, prev T represents the target angular velocity at the previous moment. vel For the speed loop control cycle. Feedforward term. The current command directly reflects the changing trend of the target speed and is superimposed on the PI controller output in an open-loop manner, so that the current command receives a corresponding increment at the beginning of the speed change, thus shortening the acceleration response time. The overall control law is shown in formula (81):

[0622] (81)

[0623] In the formula, K p, vel K i, vel For the speed loop proportional and integral gain coefficients, sat AW To resist integral saturation, the q-axis target current i output by the velocity loop is... q, ref With the underlying torque limit value T e, limit The combined effect, after online saturation limiting, is shown in formula (82):

[0624] (82)

[0625] In the formula, p n This represents the number of pole pairs of the motor.

[0626] The input signal for this stage is the target angular velocity ω output by the position loop. m, ref , Filter cutoff angular frequency ω LPF The target angular velocity ω at the previous moment m, prev The bottom torque limit value T e, limit Number of pole pairs p of the motor n The output signal is the trend of the target velocity. The target q-axis current i output by the speed loop q, ref .

[0627] Step 3: Adaptive Closed-Loop Control of Motor Torque Loop. This step is based on the q-axis target current i output by the speed loop. q, ref A bottom-level torque adaptive closed-loop controller is designed to achieve accurate and rapid response of the motor's electromagnetic torque to the target torque. The current loop adopts a control architecture that includes feedforward decoupling and PI control, and superimposes current prediction compensation, adaptive field weakening control, torque-current adaptive calibration and active damping injection modules. The current loop control law is shown in formula (83):

[0628] (83)

[0629] In the formula, K p, cur K i, curThese are the current loop proportional gain and integral gain coefficient, u q, comp u d, comp This is the current prediction compensation term. When the motor terminal voltage approaches the DC bus voltage limit, adaptive field weakening control is activated, and the d-axis target current i output by the speed loop is... d, ref As shown in formula (84):

[0630] (84)

[0631] In the formula, φ f (T motor The permanent magnet flux linkage is temperature-corrected. The torque constant deviation is identified online using recursive least squares with a forgetting factor, and the identification model is shown in formula (85):

[0632] (85)

[0633] In the formula, Let be the estimated torque constant to be identified at the k-th sampling time, and ε(k) be the identification residual at the k-th sampling time. The recursive process is shown in formula (86):

[0634] (86)

[0635] In the formula, P(k-1) is the covariance matrix at the (k-1)th step, and λ FF Let P(k) be the forgetting factor. The covariance matrix P(k) for the k-th step is constructed as shown in formula (87):

[0636] (87)

[0637] Corrected q-axis target current i q, ref, mod As shown in formula (88):

[0638] (88)

[0639] The transfer function of the active damped injection bandpass filter is shown in equation (89):

[0640] (89)

[0641] In the formula, G BPF (s) is the transfer function of the bandpass filter, s is the complex frequency domain variable, ζ is the damping ratio of the bandpass filter, and ω is the signal strength of the bandpass filter. n The center angular frequency of the bandpass filter. Active damping injection current compensation Δi q, damp As shown in formula (90):

[0642] (90)

[0643] In the formula, K damp The active damping gain is dynamically adjusted according to the steering load as shown in formula (91):

[0644] (91)

[0645] In the formula, K damp, 0 K is the base-valued damping gain. damp, 1 T is the load-related damping gain coefficient. L For steering load torque. Corrected total q-axis current command i q, ref, total As shown in formula (92):

[0646] (92)

[0647] Add a torque overload graded protection and safety monitoring module, current RMS value I rms As shown in formula (93):

[0648] (93)

[0649] To suppress temperature rise and ensure safe operation of the motor by linearly reducing the torque limit when the motor temperature exceeds the rated value but does not reach the limit value, the first-level protection torque derating is activated, as shown in formula (94):

[0650] (94)

[0651] In the formula, T e, limit T is the torque limiting value after the first-level protection is activated. nom T is the rated operating temperature of the motor. max This is the maximum allowable temperature for the motor.

[0652] The input signal for this stage is the q-axis target current i output by the velocity loop. q, ref Current loop proportional gain and integral gain coefficient K p, cur K i, cur The estimated value of the torque constant to be identified at the k-th sampling time. The transfer function G of a bandpass filter BPF (s), Active damping injection current compensation Δi q, damp Rated operating temperature of the motor (T) nom Maximum allowable temperature T of the motor max The output signal is the d-axis target current i output by the speed loop. d, ref Corrected q-axis target current i q, ref, mod Corrected total q-axis current command i q, ref, total Torque limiting value T after first-level protection starts e, limit .

[0653] The process of constructing the multi-round integrated steering evaluation method in the fourth step is as follows:

[0654] Step 1: Minimum number of wheel pairs verification calculation method. Step 1 consists of three steps.

[0655] Step 1: Single Wheelset Capacity Boundary Calculation. This step takes the capacity requirements output from the first step, the overall vehicle performance specification design, as input, and outputs the lower limit of the number of wheelsets configuration in the second step, multi-wheelset vehicle modeling, forming a two-way verification closed loop of "positive demand setting and negative lower limit setting." This is combined with the rated load capacity m of the single wheelset from the first step. rated Wheelset dynamic load correction factor k v (V, R), dimensionless comprehensive slip variable σ of the second-step tire mechanics model, tire curvature factor E1, nonlinear correction coefficient σ tire (E1, σ), comprehensively considering four core influencing factors—road speed coupled steady-state load, tire curvature stiffness nonlinear degradation, tire slip mechanical attenuation, and heavy-load transient dynamic impact—a quantitative model of the single-wheelset ultimate load capacity adapted to complex dynamic conditions is constructed. This model integrates multi-dimensional degradation and transient response characteristics to establish a quantitative formula for the single-wheelset performance boundary. The dynamic ultimate load capacity m of the single-wheelset is... lim The calculation formula is shown in formula (95):

[0656] (95)

[0657] Single wheel pair ultimate lateral adhesion force F y, lim The calculation formula is shown in formula (96):

[0658] (96)

[0659] A tire lateral force wheel-pair rotation angle coupling equation is established to represent the dynamic relationship between rotation angle δ and the load-bearing and adhesion performance of a single wheel pair, as shown in formula (97):

[0660] (97)

[0661] In the formula, σ(δ) is the dimensionless composite slip variable that varies with the steering angle, and k δ δ is the steering slip correlation coefficient, and δ0 is the zero-slip reference steering angle of the wheelset.

[0662] The input signal for this stage is the rated load capacity (m) of a single wheelset. rated Wheelset dynamic load correction factor k v (V,R), dimensionless comprehensive slip variable σ of tire mechanics model, tire curvature factor E1, nonlinear correction coefficient σ tire (E1, σ), load unevenness distribution coefficient η loadThe output signal is the dynamic ultimate bearing capacity m of a single wheelset. lim Single wheel pair ultimate lateral adhesion force F y, lim , Angle δ.

[0663] Step Two: Calculation of Vehicle Capability Boundary Lines under Specific Loads. This step, based on multi-condition carrying capacity, stability, and adhesion boundaries, calculates the basic vehicle capability boundary lines required to meet the load requirements under a specific load condition. The total driving force F required by the vehicle is... req From the highest vehicle speed V in the first step of the dynamics assessment max The maximum gradient α is determined as shown in formula (98):

[0664] (98)

[0665] In the formula, M cargo For cargo mass, the total braking force F required for the entire vehicle b, req The maximum braking deceleration 'a' given by the first step of the braking stability assessment x, max As shown in formula (99):

[0666] (99)

[0667] Total lateral force required for the vehicle F y, req The maximum safe lateral acceleration a given by the first step of the handling stability assessment y, max As shown in formula (100):

[0668] (100)

[0669] The input signal for this stage is the maximum vehicle speed V. max Maximum gradeability α, maximum braking deceleration a x, max Maximum safe lateral acceleration a y, max Vehicle weight M v Goods quality M cargo The output signal is the total driving force F required by the entire vehicle. req Total braking force required for the entire vehicle F b, req Total lateral force F required for the entire vehicle y, req .

[0670] Step 3: Reverse Derivation of Minimum Wheelset Count. This step, based on the microscopic performance boundaries of a single wheelset in Step 1 and the macroscopic boundaries of the entire vehicle in Step 2, employs a reverse derivation logic using performance parameters. Constrained by the compliance requirements of the first step (transportation safety) and the third step (steering stability), it independently derives the minimum wheelset count from four dimensions: load-bearing constraints, power constraints, braking constraints, and lateral stability constraints. This reverse derivation achieves the minimum wheelset count required for safe driving under all operating conditions. The minimum wheelset count n based on load-bearing capacity constraints is... load As shown in formula (101):

[0671] (101)

[0672] In the formula, F0 is the rated load of a single wheelset, and λ z The bearing capacity degradation coefficient is the road surface. The variance of the multi-axis load is given. The minimum number of wheelsets n based on driving force constraints is also given. drive As shown in formula (102);

[0673] (102)

[0674] In the formula, λ d To drive the dimensional coupling weight coefficients, ΔF dyn (n) represents the dynamic axle load transfer coupling term under accelerated conditions, indicating the degree of degradation of drive adhesion conditions caused by axle load redistribution during acceleration under different wheelset numbers. The minimum number of wheelsets n based on braking force constraints. brake As shown in formula (103):

[0675] (103)

[0676] In the formula, λ b Φ is the coupling weight coefficient for the braking dimension. brake (n) is the axle load mismatch function under braking conditions, characterizing the risk of wheel lock-up caused by axle load shifting during emergency braking. The minimum number of wheelsets n based on lateral stability constraints. lat As shown in formula (104):

[0677] (104)

[0678] In the formula, λ y , λ s For the lateral stability dimension coupling weight coefficients, Φ roll (n) is the multi-axis roll stiffness attenuation coupling term, characterizing the effect of wheelset number variation on the vehicle's equivalent roll stiffness and ultimate roll angle. P s (n) is the multi-axis steering geometry interference penalty function. The minimum number of wheelsets in the global envelope is n. min As shown in formula (105):

[0679] (105)

[0680] Take the maximum value of the required number of wheelsets for each boundary, ensuring that a single wheelset simultaneously meets all requirements for driving force, braking force, lateral force, and load-bearing capacity. Compare the calculated result with the actual number of wheelsets n configured in the second step. If n ≥ n min The current configuration meets the carrying requirements if n < n minIf the error persists, the "insufficient wheel pair configuration" signal needs to be returned to the second step to reconfigure the number of wheel pairs.

[0681] The input signal for this stage is the total driving force F required by the entire vehicle. req Total braking force required for the entire vehicle F b, req Total lateral force F required for the entire vehicle y, req Theoretical maximum carrying capacity of the whole vehicle M total The output signal is the minimum number of wheelsets n based on the bearing constraint. min .

[0682] Step 2: Wheelset Failure Performance Assessment Method. Step 2 consists of three steps.

[0683] Step 1: Establishment of Wheelset Failure Mode Identification Mechanism. This step identifies the failure modes of each wheelset based on the multi-wheelset steering system execution response, motor operating status, tire mechanical feedback, vehicle motion parameters, and wheelset status monitoring data. Model-based residual analysis is used for failure detection and isolation. The residual ϵ... i (k)=u cmd, i (k)-u fb, i (k) represents the instruction value u cmd, i (k) and actual feedback value u fb, i The difference (k) when |ϵ i (k)|≥ϵ th When the failure persists for more than 5 sampling periods, it is determined that a failure has occurred. A quantitative classification criterion is established for the failure mode, as shown in formula (106):

[0684] (106)

[0685] In the formula, l i For the failure mode indication vector, ΔT lag To respond to the delay time, u max, sensor Δu is the upper limit of the sensor's measurement range. jump, max N represents the maximum allowable value for signal transitions. total, fail The cumulative number of judgment cycles for complete failure.

[0686] The input signal for this stage is the command value u. cmd, i (k) Actual feedback value u fb, i (k), Response delay time ΔT lag Sensor range upper limit u max, sensor Maximum allowable value of signal transition Δu jump, max The cumulative number of complete failure judgment cycles N total, fail The output signal is the failure mode indicator vector l. i .

[0687] Step 2: Quantification of Performance Degradation of Failed Wheelsets. This step is based on the failure mode indicator vector l determined in Step 1. i This quantifies the performance degradation degree of each failed wheelset. The discrete failure type of each wheelset is mapped to a continuous performance degradation factor ρ. deg, i , representing the degree of performance loss of each failed wheelset relative to the healthy state, as shown in formula (107):

[0688] (107)

[0689] In the formula, T avail T represents the currently available torque. req For the target torque requirement, l i =0 indicates a normal state. After obtaining the degradation factors for each wheelset, the performance degradation level at the whole vehicle level is further evaluated. Different wheelsets bear different axle loads in the vehicle, and their impact on the overall vehicle performance has different weights. Therefore, the comprehensive index of overall vehicle performance degradation ρ deg, total With the axle load F of each wheelset z, i The weighted average is calculated as shown in formula (108):

[0690] (108)

[0691] The input signal for this stage is the performance degradation factor ρ. deg, i Failure Mode Indication Vector l i Axle load F of each wheelset z, i The output signal is the overall vehicle performance degradation index ρ. deg, total .

[0692] Step 3: Steering performance verification under failure conditions. This step uses the overall vehicle performance degradation index ρ output from Step 2 as the basis for verification. deg, total As input, it is embedded into the vehicle dynamics model established in the second step. By modifying the steering angle and tire force output boundary of the corresponding failed wheelset, the three key steering performance indicators under the failure condition are recalculated: minimum steering radius R min Maximum safe steering speed V safe Yaw rate tracking error e r The performance degradation index is used to correct the steering angle and tire force output constraints of the failed wheelset, as shown in formula (109):

[0693] (109)

[0694] In the formula, δ i, max, fail F y, i, max, fail These represent the maximum steering angle and maximum lateral tire force after correction of the failed wheelset.

[0695] When a steering actuator fails, the dynamic response of a multi-axle heavy-duty vehicle exhibits strong nonlinearity, multi-axis coupling, and multi-timescale characteristics. A physical information neural network (PIN) is introduced to perform a high-precision evaluation of steering performance under failure conditions using a data model fusion-driven approach. The PIN model consists of four parts: an input layer, a feature extraction layer, a physical constraint layer, and an output layer. The input layer receives the current vehicle state vector X. fail As shown in formula (110):

[0696] (110)

[0697] In the formula, failID is the identifier code for the failed wheel pair. The feature extraction layer uses a bidirectional long short-term memory network to extract temporal features and embeds a self-attention mechanism to automatically identify the key features most influential on the performance evaluation of the failure condition. The high-dimensional temporal input is mapped to the latent feature space, and the enhanced feature vector H is obtained after self-attention weighting. att As shown in formula (111):

[0698] (111)

[0699] In the formula, Att is the self-attention operator, BIL is the dynamic feature extracted by the bidirectional long short-term memory network, and H(t-1) is the historical hidden state vector of the bidirectional long short-term memory network at time t-1. The physical constraint layer uses prior constraints of yaw and roll dynamics of the multi-axle vehicle to correct the network output. The physical constraint loss term Loss phys As shown in formula (112):

[0700] (112)

[0701] In the formula, f dyn This is a multi-wheelset vehicle dynamics mapping function. The corrected failed wheelset constraints and real-time vehicle state are input into a trained physical information neural network, and the output layer directly maps the predicted values ​​of three key steering performance parameters under the failure condition. , , As shown in formula (113):

[0702] (113)

[0703] The input signal for this stage is the overall vehicle performance degradation index ρ. deg, total Maximum steering angle and maximum lateral tire force δ after correction of the failed wheelset i, max, fail F y, i, max, fail The current vehicle state vector X fail Physical constraint loss term Loss physThe output signal is the enhanced feature vector H after self-attention weighting. att Physical constraint loss term Loss phys Predicted minimum turning radius under failure conditions Predicted maximum safe steering speed under failure conditions yaw rate tracking error prediction value .

[0704] Step 3: Low-speed steering wheel pair coordination evaluation method. Step 3 consists of three steps.

[0705] Step 1: Low-Speed ​​Steering Wheelset Status Acquisition. This step involves acquiring real-time status parameters of each wheelset under low-speed steering conditions. At low speeds, the tire slip angle is small, and the wheelset kinematics are closer to the Ackerman steering geometry. Therefore, the data acquired under this condition more accurately reflects the steering coordination between multiple wheelsets, and the theoretical Ackerman steering angle δ of each wheel is [data missing]. i, theory As shown in formula (114):

[0706] (114)

[0707] In the formula, R ack δ1 is the turning radius of the first axis, and δ1 is the actual turning angle of the first axis.

[0708] Based on this, real-time status parameters of each wheelset are collected and calculated, including the actual steering angle δ of each wheel. i Rate of change of rotation angle of each wheel Tire slip angle α of each wheel i Tire wear rate W for each wheel rate, i After aligning the four types of data by timestamp, they are organized into a low-speed steering wheelset state matrix S, with wheelsets as the unit. low As shown in formula (115):

[0709] (115)

[0710] In the formula, each row in the matrix corresponds to a wheelset, the row number is the axle number, from the 1st axle to the nth axle, and the four columns are the actual steering angle, the rate of change of steering angle, the tire slip angle and the tire wear rate of each wheel, respectively.

[0711] The input signal for this stage is the first axle steering radius R. ack 1. Actual steering angle δ1 of the first axis; 2. Longitudinal distance L from the first axis to the i-th axis. i Actual steering angle δ of each wheel i Rate of change of rotation angle of each wheel Tire slip angle α of each wheel i Tire wear rate W for each wheel rate, i The output signal is the theoretical Ackermann angle δ for each wheel. i, theoryLow-speed steering wheel pair state matrix S low .

[0712] Step 2: Calculation of Wheelset Wear Degree. This step is based on the low-speed steering wheelset state matrix S collected in Step 1. low The degree of uneven wear among wheelsets is assessed from the perspective of wear rate. The wear rate of each wheelset reflects its wear rate per unit time. By comparing the wear rates of all wheelsets collected in step one, the reasonableness of the wear pattern is determined. The average wear rate W of all wheelsets is calculated. rate, avg As shown in formula (116):

[0713] (116)

[0714] The wear index W of the i-th pair uneven, i Defined as the relative degree to which the wear rate of the wheelset deviates from the average wear rate, as shown in formula (117):

[0715] (117)

[0716] In the formula, ε W For custom small quantities, to prevent the denominator from being zero. W uneven, i A larger value indicates a more severe deviation of the wheelset's wear from the average level, and a higher risk of uneven wear. Simultaneously, the wheelset number i with the most severe uneven wear is identified. max =argmax i W uneven, i and its wear index W uneven, max The overall vehicle wear uniformity index U wear Defined as the average compensation value of the wear index of all wheelsets, as shown in formula (118):

[0717] (118)

[0718] U wear The closer the value is to 1, the more uniform the wear of each wheel pair; the closer it is to 0, the more severe the uneven wear. When the result indicates severe uneven wear, the steering control strategy needs to be optimized and adjusted, and the weight of the steering angle distribution of each axle needs to be re-optimized to reduce the load on the high-wear wheelsets.

[0719] The input signal for this stage is the low-speed steering wheel pair state matrix S. low The output signal is the average wear rate W of the wheelset. rate, avg The wear index W of each wheelset uneven, i The wheelset with the most severe uneven wear is numbered i max Overall vehicle wear uniformity index U wear .

[0720] Step 3: Comprehensive Calculation of Wheelset Coordination. This step comprehensively calculates the low-speed steering wheelset coordination based on three dimensions: the deviation of each wheelset's steering angle from the theoretical Ackermann steering angle, the consistency of each wheelset's sideslip angle, and the uniformity of uneven wear. The coordination degree C of each wheelset is calculated. sync, i As shown in formula (119):

[0721] (119)

[0722] In the formula, ω δ ω α ω W α is the weighting coefficient. max The critical sideslip angle. The low-speed steering wheel pair coordination degree C of the entire vehicle. sync, total The arithmetic mean of the coordination of all wheelsets is given by formula (120):

[0723] (120)

[0724] The input signal for this stage is the actual steering angle δ. i Theoretical Ackermann rotation δ i, theory Actual sideslip angle α i Mean sideslip angle α avg Critical sideslip angle α max The wear index W of each wheelset uneven, i The output signal is the coordination degree C of each wheelset. sync, i Low-speed steering wheel pair coordination degree C of the whole vehicle sync, total .

[0725] Step 4: High-speed steering stability assessment method. Step 4 consists of three steps.

[0726] Step 1: Extraction of Dynamic Response Characteristics for High-Speed ​​Steering. This step focuses on the high-speed steering conditions of heavy-duty vehicles, constructing a standardized test case set S that includes constant speed lane change, high-speed step steering, and high-speed sinusoidal steering. c The constant speed lane change test is used to evaluate the vehicle's lateral stability and yaw response quality in highway scenarios; the high-speed step steering test is used to obtain the transient response characteristics of the vehicle's yaw rate, including response time and overshoot; the high-speed sinusoidal steering test is used for frequency domain identification, extracting amplitude attenuation and phase lag characteristics at different excitation frequencies. The extracted time-domain response signal includes the steering wheel angle δ. i Yaw angular velocity r(t), roll angle ϕ(t), lateral acceleration a y (t), Vertical force F of each wheel zi (t) , The collected data is filtered and denoised to remove outliers, forming a standardized test dataset S. d .

[0727] The input signal for this stage is the steering wheel angle δ. i Yaw angular velocity r(t), roll angle ϕ(t), lateral acceleration a y (t), Vertical force F of each wheel zi (t); the output signal is the standardized test condition set S. c Standardized test dataset S d .

[0728] Step Two: Calculation of High-Speed ​​Steering Stability Indicators. This step is based on the standardized test dataset S extracted in Step One. d Calculate the core stability evaluation indicators. Solve for the yaw oscillation attenuation coefficient, high-speed trajectory tracking accuracy, and high-speed lateral stability margin respectively. Yaw oscillation attenuation rate ζ yaw The calculation is shown in formula (121):

[0729] (121)

[0730] In the formula, n is the total number of oscillation peaks counted in the yaw rate decay oscillation waveform, and A r1 A represents the amplitude of the first peak of the yaw oscillation. rn This represents the amplitude of the nth peak of the yaw oscillation. High-speed trajectory tracking accuracy error E y, high The calculation is shown in formula (122):

[0731] (122)

[0732] In the formula, N is the total number of sampling points, Y ref Y represents the desired lateral trajectory coordinates. act These are the actual lateral trajectory coordinates. High-speed lateral stability margin K. stab, high The calculation is shown in formula (123):

[0733] (123)

[0734] The input signal for this step is the standardized test dataset S. d The output signal is the yaw oscillation decay rate ζ. yaw High-speed trajectory tracking accuracy error E y, high High-speed lateral stability margin K stab, high .

[0735] Step 3: Comprehensive Rating of High-Speed ​​Steering Stability. This step combines the three core indicators calculated in Step 2: yaw rate decay ζ. yaw High-speed trajectory tracking accuracy error E y, high High-speed lateral stability margin K stab, highA weighted fusion method is used to establish a comprehensive rating for high-speed steering stability. Three indicators characterize high-speed steering stability from three dimensions: transient response quality, trajectory holding capability, and roll safety margin. yaw E reflects the vehicle's damping characteristics in response to steering input; a larger value indicates faster yaw rate convergence. y, high This reflects the vehicle's ability to track the desired path; a smaller value indicates higher tracking accuracy. K stab, high Reflects a vehicle's ability to resist rollover; a higher value indicates a greater lateral stability margin. High-speed cornering stability overall rating R high As shown in formula (124):

[0736] (124)

[0737] In the formula, λ h1 , λ h2 , λ h3 ζ is the weighting coefficient. yaw, ref For reference to the yaw oscillation decay rate, E y, allow To allow for the maximum trajectory tracking error, K stab, ref For reference to lateral stability margin, the rating criteria are as follows: R high A value ≥0.85 is considered excellent, indicating high safety during high-speed steering, superior handling quality, rapid yaw convergence, accurate trajectory tracking, and ample roll margin; 0.70≤R high A value <0.85 is considered good, meeting basic safety requirements for high-speed steering; however, it is recommended to focus on optimizing the transient response quality. A value ≤0.55 is considered good. high A value <0.70 is acceptable, but this indicates a risk of roll or trajectory deviation, requiring limitation of the maximum steering speed or optimization of tire lateral stiffness matching; R high A value less than 0.55 is considered unacceptable, indicating a safety hazard during high-speed steering. The suspension roll stiffness, wheel set steering angle distribution strategy, or tire selection need to be reassessed.

[0738] The input signal for this stage is the yaw rate decay ζ. yaw High-speed trajectory tracking accuracy error E y, high High-speed lateral stability margin K stab, high Reference yaw oscillation decay rate ζ yaw, ref Maximum allowable trajectory tracking error E y, allow Reference lateral stability margin K stab, ref The output signal is the comprehensive rating of high-speed steering stability, R. high Rating results.

Claims

1. A multi-wheelset, multi-actuator integrated steering method for heavy-duty vehicles, characterized in that: The method includes the following steps: Step 1: Modular heavy-duty vehicle performance specification design; The second step is to establish a multi-wheel vehicle model. Step 3: Design of multi-wheel integrated steering control strategy; Step 4: Construction of a multi-round integrated steering evaluation method.

2. The integrated steering method for multi-wheelset multi-actuator vehicles for heavy-duty vehicles according to claim 1, characterized in that: The process of designing the performance specifications for modular heavy-duty vehicles in the first step is as follows: Step 1: Assessment of the carrying capacity of modular heavy-duty vehicles. Step 1 consists of three parts, as follows: Step 1: Calculation of the mapping relationship between the number of wheelsets and the carrying capacity. This step establishes a basic mapping relationship between the number of wheelsets and the carrying capacity of the entire vehicle based on the load-bearing characteristics of multi-wheelset configurations in heavy-duty vehicles. Let the rated carrying capacity of a single wheelset be m. rated With n wheelsets, considering the coupled effect of tire vertical stiffness and road surface unevenness, the actual load-bearing capacity of a single wheelset dynamically changes with vehicle speed V and road surface grade R. A dynamic load-bearing correction factor k is established. v (V, R), then the theoretical maximum carrying capacity M of the whole vehicle total As shown in formula (1): (1); In the formula, η load Let be the load uneven distribution coefficient, where k is the dynamic load correction factor. v The expression for calculating (V, R) is shown in formula (2): (2); In the formula, G q (r) represents the displacement power spectral density of road surface unevenness, α R This is the gain coefficient of the tire suspension coupled dynamic load; The input signals for this stage are the number of wheelsets n and the rated load capacity m of a single wheelset. rated Vehicle speed V, road surface grade R; the output signal is the theoretical maximum carrying capacity M of the vehicle. total ; Section Two: Multi-condition Carrying Capacity Calculation. This section calculates the carrying capacity under various typical operating conditions faced by heavy-duty vehicles in actual operation, including constant speed on a flat road, acceleration on a flat road, starting on an incline, crawling on an incline, and driving on curves. It establishes the maximum permissible carrying mass under constraints of traction force, resistance balance and adhesion, power, and stability. The carrying capacity M for constant speed driving on a flat road is calculated. flat, const As shown in formula (3): (3); In the formula, μ is the road surface adhesion coefficient, ψ is the static load distribution ratio of the drive axle, and ρ a C is the density of air. D Where A is the air resistance coefficient, f is the frontal area, g is the rolling resistance coefficient, and M is the acceleration due to gravity. The carrying capacity M is the acceleration on a straight road. flat, acc As shown in formula (4): (4); In the formula, ψ' is the dynamic load coefficient of the drive shaft during acceleration, δ is the rotational mass conversion coefficient, a is the target acceleration, and M is the carrying capacity for starting on an incline. ramp, start As shown in formula (5); (5); In the formula, F max The maximum driving force of the drive wheels is given by θ, where θ is the road gradient angle, and M represents the carrying capacity for crawling on the slope. ramp, climb As shown in formula (6): (6); In the formula, P max η is the maximum output power of the engine motor. t For the overall efficiency of the transmission system, the carrying capacity M when cornering curve As shown in formula (7): (7); In the formula, F tire Δlat represents the maximum allowable vertical load on a single wheel, and Δlat is the percentage increase in wheel load during curves. The input signals for this stage are the road surface adhesion coefficient μ, the static load distribution ratio of the drive shaft ψ, and the air density ρ. a Air drag coefficient C D Frontal area A, rolling resistance coefficient f, dynamic load coefficient of drive shaft during acceleration ψ', rotational mass conversion coefficient δ, target acceleration a, maximum driving force of drive wheel F max Road slope angle θ, maximum output power of engine motor P max Overall efficiency η of the transmission system t Maximum allowable vertical load F per wheel tire The wheel load increment ratio Δlat in curves; the output signal is the carrying capacity M for the following driving conditions: constant speed on a straight road, acceleration on a straight road, hill start, hill crawl, and curve driving. flat, const M flat, acc M ramp, start M ramp, climb M curve ; Step 3: Determining the carrying capacity boundary and safety margin. This step, based on the carrying capacity calculations for the five operating conditions in Step 2, takes the minimum allowable carrying mass for each operating condition as the lower limit boundary of the vehicle's carrying capacity, and takes the flat static benchmark value as the upper limit boundary, thus determining the safe carrying range of the entire vehicle [M]. min, safe M max, safe The carrying capacity safety margin SM is defined as shown in formula (8): (8); In the formula, M actual The actual carrying capacity is determined by the SM (Shipment Scale) threshold. When the SM reaches a certain threshold, an overload warning is triggered. The input signal for this stage is the carrying capacity M under the following conditions: constant speed on a flat road, acceleration on a flat road, starting on a slope, crawling on a slope, and driving on a curve. flat, const M flat, acc M ramp, start M ramp, climb M curve The output signal represents the safe carrying range of the entire vehicle [M]. min, safe M max, safe Safety margin of carrying capacity (SM); Step 2: Modular Heavy-Duty Vehicle Dynamics Assessment. Step 2 consists of three parts, as follows: Step 1: Multi-wheel drive power matching calculation. This step calculates the total drive power required for three driving conditions—high-speed cruising, climbing, and acceleration—based on the vehicle's total mass, design maximum speed, and maximum climbing resistance. Then, it allocates the output power of each wheel pair according to the axle load ratio of each drive wheel axle. The total power demand P at the vehicle's steady-state maximum speed is calculated. total, max As shown in formula (9): (9); In the formula, M v V represents the unloaded mass. max At the maximum vehicle speed, the power distribution P of a single wheelset i As shown in formula (10): (10); In the formula, F i For vertical load on a single axle; The input signal for this stage is the theoretical maximum carrying capacity M of the vehicle. total Unloaded mass M v Maximum speed V max air density ρ a Air drag coefficient C D Frontal area A, rolling resistance coefficient f, vertical load on a single axle F i The output signal is the total power demand P at the vehicle's steady-state maximum speed. total, max Power distribution P of a single wheelset i ; Section Two: Acceleration and Climbing Ability Assessment. This section assesses the vehicle's acceleration and climbing ability based on its drive power and curb weight, with acceleration time t... acc The calculation is shown in formula (11): (11); Maximum stable gradient α max The calculation is shown in formula (12): (12); In the formula, V c To maintain a stable speed while climbing the hill; The input signal for this stage is the theoretical maximum carrying capacity M of the vehicle. total Unloaded mass M0, vehicle speed V, air density ρ a Air drag coefficient C D Frontal area A, rolling resistance coefficient f, rotational mass conversion factor δ, maximum output power of engine / motor P max Overall efficiency η of the transmission system t Stable climbing speed V c The output signal is the acceleration time t. acc Maximum stable gradient α max ; Step 3, Comprehensive Evaluation and Output of Dynamic Performance: This step integrates the calculation results from Step 1 and Step 2 to construct a performance evaluation system that includes the maximum vehicle speed V. max Maximum stable gradient α max Acceleration time t acc The dynamic performance evaluation system comprises three core indicators, and a weighted comprehensive score is applied based on five common operating conditions of heavy-duty vehicles. The analytic hierarchy process (AHP) is used, with the weight coefficient for each operating condition set as ω. i Finally, the integrated calculation yields the comprehensive power score S. power As shown in formula (13): (13); In the formula, λ1, λ2, and λ3 are weighting coefficients; The input signal for this stage is the maximum vehicle speed V. max Maximum stable gradient α max Acceleration time t acc Weighting coefficient ω for operating conditions i The output signal is the comprehensive power score S. power ; Step 3: Modular heavy-duty vehicle stability assessment. Step 3 consists of three parts, as follows: Step 1: Handling Stability Assessment. This step focuses on the steering characteristics of multi-axle heavy-duty vehicles, establishing a handling stability evaluation index system, using understeer angle Us and yaw rate gain G as the main metrics. yr Lateral acceleration gain G la As a core evaluation indicator, for an n-axis vehicle, l is defined as i Let be the longitudinal distance of the i-th axis relative to the sprung center of mass of the entire vehicle, and let l be the distance between the axles in front of the center of mass. i <0, the axle behind the center of mass satisfies l i >0, the first axis is the equivalent front axle, serving as the main steering input axis of the vehicle, and the set of rear axles is defined as Ωr={i|l i For all load-bearing axes and active steering driven axes behind the center of mass, ≥0, 2≤i≤n}, a unique equivalent rear axle is constructed. The understeer angle Us is defined as the equivalent front axle tire slip angle α. f Equivalent rear axle tire slip angle α r The difference between the two and the lateral acceleration a y The ratio of the two values ​​is used to calculate the steady-state yaw rate gain G. yr As shown in formula (14): (14); In the formula, L eq For the equivalent wheelbase, K Us λ is the understeering coefficient. i Here are the steering weighting coefficients for each axis, and the lateral acceleration gain G. la As shown in formula (15): (15); The input signal for this stage is the front axle tire slip angle α. f Rear axle tire slip angle α r Lateral acceleration a y Equivalent wheelbase L eq Insufficient turning coefficient K Us Vehicle speed V; output signals are understeer angle Us and yaw rate gain G. yr Lateral acceleration gain G la ; Step Two: Braking Longitudinal Stability Assessment. This step addresses the longitudinal stability issue of multi-axle heavy-duty vehicles during braking, establishing evaluation indicators for braking performance and braking directional stability, with the maximum braking deceleration 'a' as the criterion. brake, max Full-load braking distance S brake Axle load transfer braking lock-up boundary LTR brake The maximum braking deceleration a is used as an evaluation metric. brake, max As shown in formula (16): (16); In the formula, μ brake η is the peak adhesion coefficient of the braking surface. disk The multi-axis braking force distribution balance coefficient is constructed as shown in formula (17): (17); In the formula, F bi The braking force corresponding to the real-time braking pressure of each axle. The average braking force of all brake axles in the vehicle, and the fully loaded braking distance S. brake As shown in formula (18): (18); In the formula, t lag v0 is the initial braking velocity, v0 is the braking reaction time, and LTR is the axle load transfer braking lock-up boundary. brake As shown in formula (19): (19); In the formula, ΔF z F represents the dynamic load change during braking, which is equivalent to an increase in load on the front axle and a decrease in load on the rear axle. z, static h represents the average vertical load on the front and rear axles under static full load. c The height of the vehicle's center of gravity when fully loaded; The input signal for this stage is the peak adhesion coefficient μ of the braking surface. brake Multi-axis braking force distribution balance coefficient η disk Braking system reaction time t lag Initial braking velocity v0, dynamic load change ΔF during braking process (increased load on the front axle, decreased load on the rear axle) z The height of the vehicle's center of gravity when fully loaded (h) c Static full load, average vertical load F between front and rear axles z, static Equivalent wheelbase L eq The output signal is the maximum braking deceleration a. brake, max Full-load braking distance S brake Axle load transfer braking lock-up boundary LTR brake ; Step 3, Comprehensive Evaluation of Braking Performance and Stability: This step integrates the evaluation results from Step 1 and Step 2 to construct a comprehensive vehicle stability score. stab As shown in formula (20): (20); In the formula, ω1, ω2, ω3, ω4, and ω5 are weighting coefficients, and std is the industry standard reference value; The input signal for this stage is the yaw rate gain G. yr Lateral acceleration gain G la Maximum braking deceleration a brake, max Full-load braking distance S brake Axle load transfer braking lock-up boundary LTR brake The output signal is the overall vehicle stability score. stab ; Step 4: Modular heavy-duty vehicle passability assessment. Step 4 consists of three parts, as follows: Section 1: Multi-Axle Geometric Passability Assessment for Heavy-Duty Vehicles. This section is based on the multi-span wheelbase structure of n-axle heavy-duty vehicles, and the unique failure modes of mid-section overhang, low-lying areas between multiple axles, and interference from multiple wheel sets. It establishes a multi-axle coupled geometric passability evaluation system adapted to ultra-long vehicle bodies and multi-wheelset arrangements, adding longitudinal passability characteristics within multi-axle sections. Core indicators include minimum ground clearance h. min Approach angle α approach Departure angle α depart Define the longitudinal pass angle α of multi-axis coupling. ramp Instead of using the traditional front and rear single axle constraint method, the weighted solution is based on the maximum span of adjacent axle groups and the minimum ground clearance of each segment, which characterizes the bottoming resistance of multi-span vehicle bodies when passing through uneven roads and continuous potholes, as shown in formula (21): (21); In the formula, L wb, max This represents the maximum longitudinal distance between adjacent wheelsets. The input signals for this stage are the ground clearance of each section of the heavy-duty multi-axle vehicle, the geometric parameters of the front end, the set parameters of the rear end, and the maximum longitudinal distance L between adjacent wheelsets. wb, max The output signal is the minimum ground clearance h. min Approach angle α approach Departure angle α depart Multi-axis coupling longitudinal through angle α ramp ; Step Two: Assessment of the Passability of Flexible Supports with Multi-Axle Load Diversion. This step is based on the multi-wheel pair uniform load distribution, load diversion, and multi-point flexible grounding mechanism of ultra-multi-axle heavy-duty vehicles, as well as the issues of significantly reduced single-wheel grounding pressure, significantly reduced road subsidence, and traction adhesion characteristics that are completely different from ordinary vehicles. A passability evaluation model for soft road surface supports adapted to multi-axle heavy-duty vehicles is established, introducing a multi-axle load uniformity correction coefficient η. load This study introduces parameters such as the order of the number of axles and the dispersion of loads on each axle to reflect the coupled influence of multi-point pressure distribution and load uniformity of n-axle vehicles on road passability. It also constructs multi-axle coupled traction coefficient TC, multi-axle cooperative traction efficiency TE, and multi-axle heavy-duty fuel utilization index E. f Evaluation system, multiaxial load uniform distribution correction factor η load As shown in formula (22): (22); In the formula, The average vertical load of all axles of the vehicle is given by the multi-axle coupling traction coefficient TC, as shown in formula (23): (23); In the formula, F net The total vehicle hook traction force and the multi-axle cooperative traction efficiency TE are shown in formula (24): (24); In the formula, P out For output power, P in For input power, the multi-axle heavy-duty fuel utilization index E f As shown in formula (25): (25); The input signal for this stage is the real-time vertical load F of each axle of the vehicle. zi Average vertical load on all axles of the vehicle The traction force F of the whole vehicle hook net Total vehicle weight M total g, Output power P out Input power P in The output signal is the multi-axis load distribution correction coefficient η. load Multi-axle coupling traction coefficient TC, multi-axle cooperative traction efficiency TE, multi-axle heavy-duty fuel utilization index E f ; Step 3, Comprehensive Passability Rating: This step establishes a hierarchical comprehensive rating mechanism that combines hard thresholds for geometric constraints with soft evaluation of support performance. This aligns with the actual failure logic of heavy-duty vehicles. First, it determines whether multi-axle geometric underrun interference exceeds limits. Then, it integrates the support and passage capacity of soft road surfaces to achieve a high-precision comprehensive passability assessment and construct a multi-axle geometric passability level R. geo As shown in formula (26): (26); In the formula, β1, β2, β3, and β4 are weighting coefficients, and ref is the reference value, which constructs the multi-axis support passability level R. support As shown in formula (27): (27); In the formula, ζ1, ζ2, and ζ3 are weighting coefficients, and ref is the reference benchmark value. The overall passability rating R is then calculated. mobility As shown in formula (28): (28); In the formula, λ1 and λ2 are weighting coefficients; The input signal for this stage is the minimum ground clearance h. min Approach angle α approach Departure angle α depart Longitudinal passing angle α ramp Traction coefficient TC, traction efficiency TE, fuel utilization index E f The output signal is the geometric passability level R. geo Support passability rating R support Passability Comprehensive Rating mobility .

3. The integrated steering method for multi-wheel pair multi-actuator vehicles for heavy-duty vehicles according to claim 1, characterized in that: The process of establishing the multi-wheel vehicle model in the second step is as follows: Step 1: Vehicle dynamics modeling. The model established in this step is applicable to the dynamic response of heavy-duty vehicles with 6 axles or more. A whole vehicle dynamics model with coupling in the longitudinal, lateral, and vertical directions is established. The number of wheelsets is n, and the whole vehicle's degrees of freedom is 6+2n, namely, longitudinal translational degrees of freedom, lateral translational degrees of freedom, vertical translational degrees of freedom, yaw rotational degrees of freedom, roll rotational degrees of freedom, pitch rotational degrees of freedom, and 2n degrees of freedom of the tires. The vehicle dynamics equation set is established as shown in formula (29): (29); In the formula, M v For the overall vehicle weight, a x a y These are the longitudinal and lateral accelerations, F, respectively. xi F yi F represents the sum of the longitudinal and lateral forces of each wheel on the i-th axis. areo For air resistance, F r R is the rolling resistance, r is the yaw rate, and I is the rolling resistance. z I x I y These are the yaw, roll, and pitch moments of inertia of the entire vehicle, respectively. i Let t be the longitudinal distance from the i-th axis to the centroid. i Let ϕ be the wheelbase of the i-th axis, φ be the roll angle, and c be the wheelbase of the yoke. ϕ k ϕ These are the roll damping coefficient and pitch stiffness coefficient, respectively, θ p c is the pitch angle. θp k θp These are the pitch damping coefficient and pitch stiffness coefficient, respectively. s F is the vertical displacement of the vehicle's center of gravity. zsi Let X be the vertical force exerted by the suspension on the vehicle body along the i-th axis. This is then rearranged to obtain the overall vehicle system state vector X. body and its state derivative ; The input signal for this step is the vehicle structural parameter, the vehicle mass M. v The moment of inertia of the whole vehicle in terms of yaw, roll, and pitch I z I x I y Center of mass height h c Distance from each axis to the centroid l i Wheelbase t i The output signal is the vehicle system state vector X. body Vehicle system state derivative ; Step 2, Wheel Dynamics Modeling. Step 2 consists of three parts, as follows: Step 1: Modeling the Vertical Characteristics of a Single-Wheel Tire. This step establishes a model of the vertical force-deformation relationship between the tires of each wheelset based on the tire's vertical stiffness and damping characteristics. The vertical force F of a single-wheel tire is... zti As shown in formula (30): (30); In the formula, k ti For vertical stiffness, z ui For the vertical displacement of the wheel center, z ri For the vertical displacement of the road surface excitation, c ti This is the tire's vertical damping coefficient. The vertical velocity of the wheel center. The vertical excitation velocity of the road surface; The input signal for this stage is the vertical stiffness k. ti Vertical displacement z of wheel center ui , road surface excitation vertical displacement z ri Tire vertical damping coefficient c ti Vertical velocity of wheel center Vertical excitation velocity of the road surface The output signal is the vertical force F of a single tire. zti ; Section 2: Tire Dynamics Integration Model for Multi-Axle Heavy-Duty Vehicles. Based on the single-wheel model in Section 1, this section integrates the vertical, longitudinal, and lateral forces of the tires for all n wheelsets of the vehicle to establish a tire dynamics integration model for multi-axle heavy-duty vehicles. It defines the core coupling characteristic parameters of multi-axle heavy-duty vehicles and quantifies the unique mechanical interference effects of ultra-multi-axle vehicles, as shown in formula (31). (31); In the formula, η Fz, i To account for the unevenness of load distribution across shafts, ξ ij σ represents the mechanical interference coefficients along the i-th and j-th axes. The closer the axial distance, the stronger the coupling interference. slip The non-synchronous distortion coefficient of multi-axle slip is a unique phenomenon of heavy-load vehicles with inconsistent wheel slip states. The tire dynamics integrated equation of heavy-load vehicles is constructed, and the full-domain coupling superposition is achieved through the inter-axle interference characteristics to reflect the mechanical cluster distortion characteristics of multi-axle heavy-load tires, as shown in formula (32): (32); The input signal for this stage is the vertical force F of a single tire. zti Unevenness of load on each axis η Fz, i The mechanical interference coefficients ξ of the i-th and j-th axes ij Multi-axis slip asynchronous distortion coefficient σ slip Total longitudinal tire force F of each axle xi Total lateral tire force F of each axle yi The output signal is the total longitudinal tire force F of the multi-axis cluster coupling. x, sum Multi-axis cluster coupled total lateral tire force F y, sum ; Step 3: Multi-axle heavy-duty tire temperature-wear coupling modeling. This step addresses the unique characteristics of multi-axle heavy-duty vehicles, which operate under conditions of high load cluster pressure, continuous multi-wheel slippage, and prolonged continuous operation. These conditions result in significantly greater tire heat accumulation and wear degradation compared to ordinary vehicles. Furthermore, the coupling friction between multiple axles exacerbates the differentiated evolution of temperature loss. A tire heat generation-dissipation balance equation adapted to ultra-multi-axle heavy-duty conditions is established, coupling tire temperature changes to tire peak force and stiffness parameters. Simultaneously, a wear rate calculation model is constructed to dynamically correct tire mechanical parameters with temperature and wear degree, and to calculate the tire heat generation power P. heat As shown in formula (33): (33); In the formula, V slip, x V slip, y F represents the longitudinal and lateral slip velocities, respectively. xi For the total longitudinal tire force of each axle, F yi Based on the characteristics of accumulated heat generation and uneven heat dissipation under heavy multi-axle loads, a multi-axle coupled dynamic temperature rise balance equation is established for the total lateral tire force on each axle. Since the tires of ultra-multi-axle vehicles do not experience independent temperature rises, but rather clustered thermal radiation, close-range wheel-to-wheel thermal coupling, and continuous heat accumulation under heavy loads, a multi-axle heat accumulation coefficient ε is introduced. c The dynamic change of tire temperature is shown in formula (34): (34); In the formula, C th Here, T represents the tire's heat capacity, and h represents the tire's real-time temperature. conv T is the convective heat transfer coefficient. amb For ambient temperature, A s The effective heat dissipation area of ​​the tire, the tire wear rate W rate As shown in formula (35): (35); In the formula, K wear E is the wear coefficient. a The activation energy is R, and the gas constant is R. The input signals for this stage are the longitudinal and lateral slip velocities V. slip, x V slip, y Total longitudinal tire force F of each axle xi Total lateral tire force F of each axle yi Tire temperature T; the output signal is the tire heat generation power P. heat Tire heat capacity C th Ambient temperature T amb Tire wear rate W rate ; Step 3: Multi-round kinematic modeling. Step 3 consists of three parts, as follows: Step 1: Establishing Multi-Axis Coupled Kinematic Constraints. This step establishes the kinematic constraint relationships between each wheelset and between the wheelsets and the vehicle body, providing constraint boundary conditions for subsequent multi-body motion state solutions. Each wheelset is connected to the vehicle body through the suspension system, and the suspension kinematic constraints are embodied in the wheel alignment parameter kingpin inclination angle γ. kpi Kingpin inclination angle γ kpc Wheel camber angle γ camber γ front toe angle toe The dynamic relationship between suspension travel and suspension travel, let the vertical runout of the i-th axle be Δz. si The changes in the camber angle and toe angle of the wheel on this axle are shown in formula (36): (36); In the formula, γ camber, 0 γ toe, 0 To design the positioning angle, k camber k toe These are the suspension geometric kinematic parameters; Adjacent wheelsets are coupled and constrained by the frame structure. For multi-axle rigid frame vehicles, the displacement of each axle relative to the frame is constrained by the frame deformation, generating an inter-axle displacement deviation ΔP. i, i-1 As shown in formula (37): (37); In the formula, L i, i-1 Let e ​​be the initial axis distance between the i-th axis and the (i-1)-th axis. x For the longitudinal unit vector of the vehicle, kinematic interference occurs when the inter-axle displacement deviation exceeds the allowable range, and constraint compensation is performed. Based on the attitude constraints of each wheelset and the coupling relationship between the inter-axle pose, the instantaneous equivalent steering angle δ of the entire vehicle with multiple wheelsets adapted to multi-axle vehicles is derived. ins As shown in formula (38): (38); The input signal for this stage is the kingpin inclination angle γ. kpi Kingpin inclination angle γ kpc Wheel camber angle γ camber γ front toe angle toe The vertical runout of the frame is Δz si The output signal is the inter-shaft displacement deviation ΔP. i, i-1 Multi-wheel instantaneous equivalent steering angle δ of the whole vehicle ins ; Step Two: Kinematic Description of the Vehicle Multibody System. This step builds upon the suspension geometric constraints and multi-axle frame coupling constraints established in Step One. The entire vehicle is abstracted as a multibody system consisting of the vehicle body and multiple wheel pairs, achieving an accurate solution for the vehicle's multibody motion state. Considering the structural characteristics of ultra-multi-axle heavy-duty vehicles with 10 or more axles, the entire vehicle is viewed as a multibody system composed of the body and multiple wheel pairs. The vehicle's motion state is determined by the position of the center of mass P. G The yaw angle φ, pitch angle θp, roll angle ϕ, and the local motion state of each wheel pair are used to describe the vehicle's motion. A vehicle coordinate system Oxyz is defined, with the origin O located at the projection of the vehicle's center of mass onto the horizontal plane. The x-axis points along the vehicle's longitudinal direction in the forward direction, the y-axis points to the driver's left, and the z-axis is perpendicular to the ground and points upwards. For the i-th axis and j-th wheel, the contact point position vector P... ij As shown in formula (39): (39); In the formula, R G Let r be the vehicle body attitude rotation matrix. ij The effective rolling radius is given by R, where R is the vehicle body attitude rotation matrix. G The ZYX Euler angles are defined in the order shown in formula (40): (40); In the formula, R x R y R z Let x, y, and z be the fundamental rotation matrices about the x, y, and z axes, respectively. Differentiating these matrices with respect to the position vectors of each wheel's contact point yields the velocity vectors of each wheel's contact point. As shown in (41): (41); The input signal for this stage is the distance l from each axis to the centroid. i Wheelbase t i Location of the center of mass P G Yaw angle φ, pitch angle θp, roll angle ϕ, vehicle attitude rotation matrix R G Effective rolling radius r ij The output signal is the grounding point position vector P. ij Grounding point velocity vector ; Step 3: Integration of the Augmented Kinematic Model of the Whole Vehicle. This step integrates the multi-axis coupled kinematic constraints of Step 1 with the multi-body system kinematic description of the whole vehicle in Step 2, and fuses the vehicle body dynamics state from Step 1 with the wheel dynamics state from Step 2 to construct a control-oriented multi-wheelset augmented kinematic model of the whole vehicle, selecting generalized coordinates q∈R. 3n+6 , representing all kinematic degrees of freedom of the entire vehicle, based on the kinematic description of link one and the constraint relationship of link two, the augmented kinematic mapping matrix S(q) is constructed as shown in formula (42): (42); In the formula, S body Map the dynamic parameters output in step one to the derivative of the vehicle body's generalized coordinates, S susp S represents the suspension kinematic constraints of the mapping. steer To map the turning angular velocity to the derivative of the generalized coordinates, a null space basis V2 is designed, and a reduced-order kinematic mapping matrix A is constructed. red (q)=V2·S(q), and the reduced-order augmented kinematic model is obtained as shown in formula (43): (43); To embed the dynamic equations of step one and the tire force balance equations of step two into the kinematic model, a coupling constraint vector Ψ is defined as shown in formula (44): (44); In the formula, q body Let be the vehicle body pose vector. For vehicle speed, Let C be the vehicle body acceleration, C be the system damping matrix, K be the system stiffness matrix, and F be the system acceleration. step1 F is a generalized force vector. ij For the actual tire force, the constraint vector is differentiated with respect to time and transformed into a linear constraint at the generalized acceleration level. Combined with the reduced-order kinematic model, the standard discrete-time augmented kinematic model for control is obtained as shown in formula (45): (45); In the formula, q(k+1) is the system generalized coordinate vector of the next discrete time. This model fully describes the kinematic coupling relationship of all wheelsets of a multi-axle heavy-duty vehicle with more than 10 axles. It realizes bidirectional deep coupling with the vehicle dynamics state in step one and the wheel dynamics state in step two. Its output state vector q(k) provides all the necessary kinematic feedback for the subsequent multi-wheelset system trajectory following controller. The input signals for this stage are the generalized coordinate q and the generalized force vector F. stepl Vertical displacement of the vehicle body z s Mass matrix M v Damping matrix C, stiffness matrix K, tire force F ij Tire slip angle α ij Slip ratio λ ij The output signals are the augmented kinematic mapping matrix S(q) and the reduced kinematic mapping matrix A. red (q), coupling constraint vector Ψ, and the system generalized coordinate vector q(k+1) for the next discrete time step; Step 4: Steering motor modeling. Step 4 consists of three parts, as follows: Step 1: Establishment of the electromagnetic model of the steering motor. This step is based on the performance specifications of multi-axle heavy-duty vehicles for fast response, low hysteresis and large speed range of the steering system. The electromagnetic dynamic mathematical model of the steering motor is established, and the mathematical model in the dq coordinate system is adopted. The stator voltage equation is shown in formula (46): (46); In the formula, u d u q These are the stator voltages along the d-axis and q-axis, respectively, R s i is the equivalent resistance of the stator winding. d i q These are the stator currents along the d-axis and q-axis, respectively. d L q These are the equivalent inductances along the d-axis and q-axis, respectively, and ω e φ is the electric angular velocity of the motor. f For permanent magnet flux linkage, the electromagnetic torque T of the motor e As shown in formula (47): (47); In the formula, p n It is the extreme logarithm; The input signals for this stage are the stator currents i along the d-axis and q-axis. d i q Stator winding equivalent resistance R s d-axis and q-axis equivalent inductance L d L q ω, the electric angular velocity of the motor e Permanent magnet flux linkage φ f The output signal is the d-axis and q-axis stator voltage u. d u q Motor electromagnetic torque T e ; Step 2: Modeling the mechanical dynamics of the steering motor. This step establishes the mechanical dynamics equations of the steering motor, describing the dynamic relationship between the motor's output torque and the rotation angle and speed. The mechanical dynamics balance equation of the motor is shown in formula (48): (48); In the formula, J m Let ω be the moment of inertia of the motor rotor. m B is the mechanical angular velocity of the motor. m T is the mechanical damping coefficient. L This is the load torque; The input signal for this stage is the rotor inertia J of the motor. m ω, mechanical angular velocity of the motor m Mechanical damping coefficient B m Load torque T L The output signal is the motor electromagnetic torque T. e ; Step 3: Dynamic modeling of the steering system based on multiple motors and multiple degrees of freedom. This step couples the steering motor model with the steering actuator to establish a complete transmission model from motor torque to wheel steering angle; the steering wheel dynamic equation is shown in formula (49): (49); In the formula, J sw δ is the equivalent rotational inertia of the steering wheel and steering column. sw B is the steering wheel angle. sw τ is the equivalent damping coefficient of the steering column. f, sw T is the coulomb friction torque of the steering column. driver The input torque applied to the driver, T column The torque transmitted from the steering column to the steering gear, for the i-th axis, is the steering angle δ of the left and right wheels. i, L δ i, R The Ackermann steering geometry constraints are satisfied as shown in formula (50): (50); In the formula, t i L represents the wheelbase of each axle. i, axle Let be the distance from the instantaneous center of rotation of each axle to the axle. For multi-axle steering vehicles, there is a definite distribution relationship between the steering angles of each steering axle, and the kinematic constraints of adjacent axles are shown in formula (51): (51); In the formula, δ i, i-1 The steering actuator dynamics are determined by the motor driving torque, equivalent inertia, damping, friction, and tire return torque, with the steering angle distribution coefficient being the factor. By combining the motor mechanical dynamics equations and the steering actuator dynamics balance equations, a multi-motor, multi-degree-of-freedom steering system dynamic model is obtained, as shown in formula (52): (52); In the formula, G r J is the steering gear reduction ratio. eq For the equivalent rotational inertia of the steering actuator, B eq τ is the equivalent damping coefficient of the steering actuator. f Let τ be the Coulomb friction torque. align δ is the tire return torque. w This refers to the wheel steering angle; The input signal for this stage is the rotor inertia J of the motor. m Steering gear reduction ratio G r Equivalent moment of inertia J of steering actuator eq Wheel steering angle δ w Coulomb friction torque τ f Tire return torque τ align Mechanical damping coefficient B m The output signal is the motor electromagnetic torque T. e .

4. The multi-wheelset multi-actuator integrated steering method for heavy-duty vehicles according to claim 1, characterized in that: The process of designing the multi-wheel integrated steering control strategy in the third step is as follows: Step 1: Determining the vehicle's steering intention. Step 1 consists of three steps, as follows: Step 1, Traffic Situation Awareness and Scene Prediction, utilizes multi-source fusion information from cameras, millimeter-wave radar, lidar, and high-precision maps to construct a traffic scene state space adapted for ultra-long, multi-axle, heavy-duty vehicles. It extracts high-dimensional scene feature vectors containing vehicle motion, road geometry, and traffic density, and introduces a sliding window temporal prediction algorithm to achieve accurate rolling predictions of future short-term driving scenarios. This provides a priori environmental temporal sequence for subsequent risk identification, constructing a multi-axle heavy-duty scene state vector S. scene (t)=[V ego V target , d target , ρ road , λ traffic ] T A fixed-length sliding time window is used to extract continuous time series samples, and the scene time series matrix within the window is constructed as shown in formula (53): (53); In the formula, T w Given a sliding window duration, a first-order temporal evolution prediction model is constructed based on the temporal state matrix to output a predicted sequence of future scenes in the time domain. As shown in formula (54): (54); In the formula, A1 and B1 are time-series prediction weight matrices, which are updated in real time through rolling iteration; The input signal for this stage is the vehicle speed V. ego Target speed V target Relative distance d target Road curvature ρ load Traffic density λ traffic The output signal is the scene state vector S. scene (t), Road Environment Prediction Sequence ; Step 2, Driving Risk State Estimation, uses the scenario prediction sequence output from Step 1 as environmental input. It integrates a recursive least squares system identification and a temporal fuzzy C-means AI clustering dual-core algorithm to identify time-varying lateral dynamic parameters and road adhesion parameters of multi-axle vehicles online. Combined with the unique risk characteristics of multi-axle heavy loads, it achieves adaptive driving risk quantification without fixed thresholds. The least squares algorithm is used to identify key dynamic parameters such as road adhesion coefficient and lateral disturbance stiffness online, constructing a high-dimensional clustering feature vector X that includes environmental risk, stability risk, and multi-axle cooperative risk. risk (t), as shown in formula (55): (55); In the formula, μ est To identify the road surface adhesion coefficient using the least squares algorithm, δ des, i -δ act, i To address the steering synchronization deviation of multi-axle wheelsets, a temporal fuzzy C-means artificial intelligence clustering objective function J is constructed. FCM Strengthen the weight of the current state and suppress the interference of overloaded state lag, as shown in formula (56): (56); In the formula, k is the cluster category number, α(t) is the temporal forgetting weight coefficient, and u kt Where m is the membership degree and v is the fuzzy coefficient. k Given the cluster center vector of the k-th class, the driving conditions are adaptively divided into three categories—safe, warning, and high-risk—by iteratively updating the membership degree and cluster center. The membership degree u of the high-risk condition is then output. risk By combining cluster membership degree and three types of risk indicators, a dedicated driving risk cost function R for multi-axle heavy-load vehicles is constructed. risk As shown in formula (57): (57); The input signal for this step is the road environment prediction sequence. Real-time lateral load transfer rate (LTR), target steering angle δ of each wheelset des Actual steering angle δ of each wheel pair act Real-time minimum distance d to obstacles target, min The output signal is a high-dimensional clustering feature vector X. risk (t), membership degree of high-risk working conditions u risk Driving risk cost function R risk ; Step 3: Comprehensive Determination of Steering Intent. This step is based on the driving risk state estimation results from Step 2, combined with the driver's input operation commands and the planning commands δ from the autonomous driving decision module. cmd, auto The overall vehicle steering intention is determined by a comprehensive assessment. The determination logic is as follows: when R... risk ≥R th, high Active steering intervention is triggered when an effective obstacle avoidance path exists; when |δ w |≥|δ th |and ≥ When δ is determined to be the driver actively steering; cmd, auto ≠0 and when the autonomous driving mode is activated, the autonomous driving planning instruction is used, with the priority being safety emergency intervention, driver input, and autonomous driving instruction. The steering intention I is output based on the judgment logic and the actual input. steer ; The input signal for this stage is the driving risk cost function R. risk Driver inputs operation commands and planning commands. cmd, auto The output signal is the steering intention I. steer ; Step 2, Vehicle Trajectory Planning Strategy. Step 2 consists of three parts, as follows: Step 1: Initial trajectory generation. This step is based on the turning intention I determined in Step 1. steer With scene state vector S scene (t) To address the unique characteristics of multi-axle heavy-duty vehicles, such as long vehicle body, large turning radius, coupling constraints between multiple wheel pairs' turning angles, wide overall vehicle path envelope, and inconsistent motion trajectories of each axle's contact points, this paper improves the traditional RRT random tree algorithm to form an enhanced RRT planning algorithm adapted for heavy-duty applications. This algorithm introduces multi-wheel contact point envelope constraints throughout the sampling, node expansion, and collision detection processes. Simultaneously, it verifies whether the contact points of all n axles' wheels intrude into obstacles and lane boundaries. The algorithm sequentially completes trajectory sampling, node expansion, cost evaluation, multi-wheel pair envelope collision detection, and path backtracking, generating multiple discrete initial candidate trajectory clusters that satisfy the kinematic constraints of the multi-axle vehicle. The enhanced RRT algorithm introduces an adaptive bias sampling strategy with a bias probability p. bias The dynamic adjustment with the number of iterations is shown in formula (58): (58); In the formula, k is the current iteration number, and τ bias As the bias rate growth time constant, during the expansion phase, five candidate nodes are generated in a single sampling, and the cost function C(x) is selected as shown in formula (59): (59); In the formula, g(x) is the actual path cost from the starting point to the current node, h(x) is the heuristically estimated cost, and o(x) is the obstacle risk cost; Add hard kinematic constraints to the boundaries of the whole vehicle, prohibit sampling beyond the maximum steering angle of a single wheel and the left and right limit lateral coordinates of the lane, and add lateral safety constraints to the envelope of the multi-axle whole vehicle. The set of constraint inequalities is shown in formula (60): (60); In the formula, δ max For the single-wheel limit steering angle, y lane, min y lane, max W represents the left and right lateral boundaries of the lane. ve For the width of the entire vehicle body, y obs Let d be the lateral position of the obstacle. target, y To minimize the lateral safety distance between the trajectory point and obstacles, constraints are also imposed on the grounding point positions of all n axes to prevent scraping of the middle and rear wheels of the ultra-long multi-axle vehicle. For the discrete node sequence obtained by RRT sampling, a fifth-order polynomial is used as the basis function for the trajectory curve to construct a continuous time-domain analytical expression for a single candidate trajectory. The longitudinal curve x of the trajectory... traj Horizontal curve y traj As shown in formula (61): (61); In the formula, t∈[0, T] p [ ] represents the planning time, and a0~a5 and b0~b5 represent the polynomial coefficients of the longitudinal trajectory and the lateral trajectory, respectively. The coefficients are determined by the initial and final positions and the initial and final velocities of the boundary conditions output by RRT. The input signal for this stage is steering intention I. steer Scene state vector S scene (t), cost function C(x); output signal is bias probability p bias Single-wheel limit steering angle δ max , the left and right lateral boundaries of the lane y lane, min y lane, max Trajectory longitudinal curve x traj Horizontal curve y traj ; Step 2, Trajectory Post-processing and Smoothing Optimization: This step performs post-processing and smoothing optimization on the initial trajectory generated in Step 1, eliminating unnecessary nodes and sharp corners, and generating a smooth trajectory that meets the kinematic and comfort requirements of heavy-duty vehicles. Due to the large suspension stiffness and mechanical limitations of the steering actuator in multi-axle heavy-duty vehicles, the severe fluctuations in trajectory curvature will directly cause frequent changes in the steering angle of each wheel pair, resulting in frequent impacts on the steering motor and an increased risk of vehicle rollover. Using the polynomial trajectory output in Step 1 as the initial value, a multi-objective cost function that balances tracking accuracy and the smoothness of multi-wheel pair steering execution is constructed. The polynomial coefficients are iteratively corrected, and the smoothing optimization objective function is shown in Formula (62): (62); In the formula, ω a ω k ω δ These are the penalty weights for lateral acceleration, trajectory curvature change, and wheelset angle change, respectively. For the lateral acceleration of the trajectory, Let d be the time derivative of the trajectory curvature. δi, ref / dt represents the rate of change of the reference steering angle of the i-th wheel pair, and is a penalty term for multi-axle heavy-duty vehicles. The polynomial coefficients are updated through optimization, and the corrected continuous smooth trajectory curve T is output. smooth This ensures that position, velocity, and acceleration are continuously differentiable. The input signal for this stage is the longitudinal curve x of the trajectory. traj Horizontal curve y traj The output signal is the corrected, continuous, smooth trajectory curve T. smooth ; Step 3: Trajectory Feasibility Verification and Safety Assessment. This step involves verifying the smooth trajectory curve T generated in Step 2. smooth Conduct multi-dimensional feasibility verification and safety assessment, and smooth the trajectory curve T. smooth Substituting the multi-axle vehicle dynamic coupling model built in the second step, we perform dynamic feasibility verification, kinematic feasibility verification, and full-domain safety constraint verification. If the verification finds that the trajectory has an over-limit risk, we locate the over-limit time period and call the enhanced RRT algorithm in step one to locally regenerate the trajectory until the safety constraints are met in all time domains. Finally, we output the unique optimal reference trajectory T with the highest safety score. final ; The input signal for this stage is the smooth trajectory curve T. smooth The output signal is the globally optimal vehicle reference trajectory T after verification and correction. final ; Step 3: Multi-round system trajectory following control strategy. Step 3 is completed in three stages, as follows: Step 1: Augmented State Space Modeling of the Multi-Wheelset System. This step builds upon the multi-wheelset vehicle model established in Step 2, considering the large inertia transient lag of multi-axle heavy-duty vehicles, the time-varying nature of heavy-load parameters, and the dynamic coupling disturbance characteristics of multi-wheelsets. An augmented state space model adapted for transient trajectory following control is constructed, selecting the state vector x=[Y, V, r, φ, ϕ, δ]. i , …, δ n ] T All are inherited from the output of the augmented kinematics model in the second step, with the control input u=[δ 1, cmd , …, δ n, cmd , ΔT1, …, ΔT n ] T The nonlinear state-space equation containing the steering commands of each wheelset and the compensation for the transient torque under heavy load is established as shown in formula (63): (63); In the formula, d tran (t) represents the transient coupling disturbance term under multi-axis heavy load, and the globally optimal vehicle reference trajectory T output in step two is used. final The trajectory parameters in the vector are used as reference inputs, and the augmented state vector x is used as the reference input. aug The integral term including the trajectory tracking error is used to eliminate the steady-state error, as shown in Equation (64): (64); The input signals for this stage are the output state variable x of the multi-wheel-pair augmented kinematic model of the whole vehicle, the control input u, and the globally optimal whole vehicle reference trajectory T. final The trajectory parameters are in the vector; the output signal is the augmented state vector x. aug ; Section Two: Layered Sliding Mode Variable Structure Trajectory Following Controller. This section, based on the augmented state-space model of Section One, designs a layered sliding mode variable structure controller to achieve robust following of the reference trajectory for the multi-wheelset system. First, a dynamic coordination layer—stability boundary partitioning and priority weight generation—is designed. The stability boundary of the multi-axle heavy-duty vehicle is partitioned based on the phase plane method, and the phase plane state variable is defined as the centroid sideslip angle β. p angular velocity of the center of mass deflection The boundary equation of the phase plane stable domain is shown in equation (65): (65); In the formula, c1 and c2 are boundary coefficients related to vehicle speed and road surface adhesion coefficient. The zero-moment point roll instability judgment model position P is introduced. ZMP As shown in formula (66): (66); When P ZMP When the vehicle exceeds the boundary of the vehicle support polygon, it is determined to be a roll instability. Based on the phase plane stability domain division and ZMP roll risk assessment, a priority weighting factor ρ is dynamically generated. prior As shown in formula (67): (67); In the formula, sat is the saturation function, and ρ prior Constrained in the interval [0, 1], β crit , P ZMP, max Given the sideslip angle, sideslip angular velocity, and ZMP position critical threshold, design a variable structure control layer—a multi-constraint sliding mode control law—and define the tracking error sliding surface s. track As shown in formula (68): (68); In the formula, λ1~λ3 are strictly positive definite sliding surface parameters, and the stable sliding surface s is defined. stab As shown in formula (69): (69); In the formula, λ4~λ7 are strictly positive definite sliding surface parameters, which are determined by the pole placement method to ensure the rapid convergence of the sliding mode dynamics and good damping characteristics. The composite sliding surface S is constructed as shown in formula (70): (70); The superspiral approach law is used to weaken chattering while ensuring finite-time convergence, as shown in formula (71): (71); In the formula, k1 and k2 are strictly positive definite reaching rate parameters, based on the synovial reachability condition. =0, combining the first step vehicle dynamics equations and the second step tire model, the inverse solution yields the desired generalized torque ν. des The design incorporates a constraint processing layer—multi-constraint boundary handling and multi-wheel pair collaborative error correction. For multiple constraint conditions, a combination of boundary layer and saturation function is used to handle actuator saturation and state constraints, defining a boundary thickness T. b Adaptive adjustment, the saturation function is sat(s, T) b Define the comprehensive error E of multi-round transient cooperative tracking. syns As shown in formula (72): (72); The comprehensive error E syns As a feedback correction, it is added to the desired generalized torque ν des The closed-loop correction is formed as shown in formula (73): (73); In the formula, K sync For positive definite cooperative correction gain matrix, ν des, mod This is the desired generalized moment vector after collaborative correction; The input signal for this stage is the augmented state vector x. aug Boundary coefficients c1 and c2 related to vehicle speed and road surface adhesion coefficient, and the position P of the roll instability judgment model. ZMP Tracking error sliding surface s track Stability of sliding surface s stab Comprehensive error E syn The output signal is the priority weighting factor ρ. prior Synthetic sliding surface S, desired generalized torque ν des The desired generalized moment vector ν after collaborative correction des, mod ; Step 3: Multi-actuator control allocation and follow-up execution. This step will distribute the desired generalized torque ν output from Step 2. des The control is assigned to the steering actuators of each wheelset, and rolling time-domain updates and fault-tolerant reconfiguration are implemented. The control efficiency matrix M is defined. c Its elements are determined by the Jacobian matrix of the steering angle of each wheel pair and the tire force on the generalized force of the whole vehicle in the vehicle dynamics model of step one. The optimization problem J alloc As shown in formula (74): (74); In the formula, u act Let Q be the actual output vector of the actuator, R be the weight matrix for allocation accuracy, R be the weight matrix for load balancing, RΔ be the weight matrix for motion smoothing, and u be the weight matrix for motion smoothing. prev W is the actuator output vector at the previous time step, and W is the cumulative workload vector of each actuator. avg For multi-axle heavy-duty vehicles, the vertical load F of each axle during transient steering is the average load. zi Drastic dynamic changes directly affect the control efficiency of each actuator. Therefore, a transient load correction factor η is introduced into the control efficiency matrix. z, i (t), as shown in formula (75): (75); Corrected control efficiency matrix As shown in formula (76): (76); When a portion of the wheelset steering actuators fails, the failure detection and isolation module continuously monitors the residual error. i (k), when |ϵ i When (k)|≥ϵth and continues for more than 5 sampling periods, it is considered a failure, and the fault tolerance allocation problem J fault As shown in formula (77): (77); In the formula, M fail To reduce the order of the control efficiency matrix after removing the column corresponding to the failed actuator, u act, fail F represents the actual output vector of the remaining healthy actuators under failure conditions. act For the set of failed actuators, U fail For each sampling time T, there is a feasible set of constraints for the remaining healthy actuators under failure conditions. s Based on the latest state measurements, the sliding mode control law and control assignment problem are resolved to achieve rolling updates, and the desired steering angle command δ for each wheelset is re-output. i, cmd This forms a complete closed loop of "perception-decision-allocation-execution-feedback," enabling multi-round systems to achieve high-precision collaborative following of the reference trajectory. The input signal for this stage is the desired generalized torque ν. des Control efficiency matrix M c The actual output vector u of the actuator act The executor output vector u at the previous time step prev The output signal is the actual output vector u of the remaining healthy actuator under failure conditions. act, fail Failed actuator set F act The feasible constraint set U for the remaining healthy actuators under effective operating conditions fail δ, the desired steering angle command for each wheelset i, cmd ; Step 4: Multi-target following control strategy for motors. Step 4 consists of three parts, as follows: Stage 1: Steering angle loop following control. This stage is the outermost closed loop of the motor execution system, accurately tracking the desired steering angle command δ assigned to each wheel pair by the upper-level multi-wheel pair system. i, cmd A closed-loop tracking controller for the upper-level rotation angle position is designed to achieve precise tracking of the motor rotor position to the target rotation angle θ. m, ref =δ i, cmd / Gr, the position loop adopts a proportional-integral-derivative control framework, and introduces a model-based feedforward compensation and a reduced-order extended state observer to improve the dynamic response speed and steady-state accuracy. The control law is shown in formula (78): (78); In the formula, ω m, ref e is the target angular velocity output by the position loop. θ K represents the position tracking error. p, pos K i, pos K d, pos These are the position loop proportional, integral, and differential gain coefficients, respectively. The rate of change of error, For feedforward velocity compensation term, The total perturbation compensation for the position loop is estimated by the reduced-order extended state observer, which is in second-order form, as shown in formula (79): (79); In the formula, This is a location estimate. For the total disturbance estimate, β z1 β z2 For observer gain, θ m This represents the actual electrical angle of the motor rotor. The input signal for this stage is the desired steering angle command δ for each wheelset. i, cmd Steering gear reduction ratio Gr, position loop proportional, integral, and differential gain coefficient K p, pos K i, pos K d, pos Actual rotor electrical angle θ of the motor m The output signal is the target angular velocity ω output by the position loop. m, ref Position tracking error e θ Error change rate ; Step 2: Smooth closed-loop control of motor speed loop. This step is based on the target angular velocity ω output by the position loop. m, ref A mid-level speed smoothing closed-loop controller is designed to achieve fast and smooth tracking of the motor's actual speed to the target speed. The speed loop adopts a proportional-integral control law and introduces a first-order low-pass filter ω. LPF / (s+ω LPF ), filtering out velocity measurement noise, filter cutoff angular frequency ω LPF The selection should balance noise suppression effect and phase lag, and should be 5 to 10 times the speed loop bandwidth. The phase lag of the filter at the speed loop crossing frequency can be offset by acceleration feedforward compensation. Introducing acceleration feedforward term As shown in formula (80): (80); In the formula, ω m, prev T represents the target angular velocity at the previous moment. vel For the speed loop control cycle, the feedforward term The current command directly reflects the changing trend of the target speed and is superimposed on the output of the PI controller in an open-loop manner, so that the current command obtains a corresponding increment at the beginning of the speed change, shortening the acceleration response time. The overall control law is shown in formula (81): (81); In the formula, K p, vel K i, vel For the speed loop proportional and integral gain coefficients, sat AW To resist integral saturation, the q-axis target current i output by the velocity loop is... q, ref With the underlying torque limit value T e, limit The combined effect, after online saturation limiting, is shown in formula (82): (82); In the formula, p n This represents the number of pole pairs of the motor. The input signal for this stage is the target angular velocity ω output by the position loop. m, ref , Filter cutoff angular frequency ω LPF The target angular velocity ω at the previous moment m, prev The bottom torque limit value T e, limit Number of pole pairs p of the motor n The output signal is the trend of the target velocity. The target q-axis current i output by the speed loop q, ref ; Section 3: Adaptive closed-loop control of the motor torque loop. This section is based on the q-axis target current i output by the speed loop. q, ref A bottom-level torque adaptive closed-loop controller is designed to achieve accurate and rapid response of the motor's electromagnetic torque to the target torque. The current loop adopts a control architecture that includes feedforward decoupling and PI control, and superimposes current prediction compensation, adaptive field weakening control, torque-current adaptive calibration and active damping injection modules. The current loop control law is shown in formula (83): (83); In the formula, K p, cur K i, cur These are the current loop proportional gain and integral gain coefficient, u q, comp u d, comp As a current prediction compensation term, when the motor terminal voltage approaches the DC bus voltage limit, adaptive field weakening control is activated, and the d-axis target current i output by the speed loop is... d, ref As shown in formula (84): (84); In the formula, φ f (T motor The permanent magnet flux linkage is temperature-corrected. The torque constant deviation is identified online using recursive least squares with a forgetting factor. The identification model is shown in formula (85). (85); In the formula, Let be the estimated torque constant to be identified at the k-th sampling time, and ε(k) be the identification residual at the k-th sampling time, where The recursive process is shown in formula (86): (86); In the formula, P(k-1) is the covariance matrix at the (k-1)th step, and λ FF As the forgetting factor, the covariance matrix P(k) at the k-th step is constructed as shown in formula (87): (87); Corrected q-axis target current i q, ref, mod As shown in formula (88): (88); The transfer function of the active damped injection bandpass filter is shown in equation (89): (89); In the formula, G BPF (s) is the transfer function of the bandpass filter, s is the complex frequency domain variable, ζ is the damping ratio of the bandpass filter, and ω is the signal strength of the bandpass filter. n The center angular frequency of the bandpass filter is Δi, and the active damping injection current compensation is Δi. q, damp As shown in formula (90): (90); In the formula, K damp The active damping gain is dynamically adjusted according to the steering load as shown in formula (91): (91); In the formula, K damp, 0 K is the base-valued damping gain. damp, 1 T is the load-related damping gain coefficient. L For steering load torque, the corrected total q-axis current command i q, ref, total As shown in formula (92): (92); Add a torque overload graded protection and safety monitoring module, current RMS value I rms As shown in formula (93): (93); To suppress temperature rise and ensure safe operation of the motor by linearly reducing the torque limit when the motor temperature exceeds the rated value but does not reach the limit value, the first-level protection torque derating is activated, as shown in formula (94): (94); In the formula, T e, limit T is the torque limiting value after the first-level protection is activated. nom T is the rated operating temperature of the motor. max This refers to the maximum allowable temperature of the motor. The input signal for this stage is the q-axis target current i output by the velocity loop. q, ref Current loop proportional gain and integral gain coefficient K p, cur K i, cur The estimated value of the torque constant to be identified at the k-th sampling time. The transfer function G of a bandpass filter BPF (s), Active damping injection current compensation Δi q, damp Rated operating temperature of the motor (T) nom Maximum allowable temperature T of the motor max The output signal is the d-axis target current i output by the speed loop. d, ref Corrected q-axis target current i q, ref, mod Corrected total q-axis current command i q, ref, total Torque limiting value T after first-level protection starts e, limit .

5. The multi-wheelset multi-actuator integrated steering method for heavy-duty vehicles according to claim 1, characterized in that: The process of constructing the multi-round integrated steering evaluation method in the fourth step is as follows: Step 1: Minimum number of rounds verification calculation method. Step 1 consists of three steps, as follows: Step 1: Single-wheelset capacity boundary calculation. This step takes the capacity requirements output from the first step of the vehicle performance specification design as input, and outputs the lower limit of the number of wheelsets configuration in the second step of multi-wheelset vehicle modeling, forming a two-way verification closed loop of "positive demand setting and negative lower limit setting". This is combined with the rated load capacity m of the single-wheelset in the first step. rated Wheelset dynamic load correction factor k v (V, R), dimensionless comprehensive slip variable σ of the second-step tire mechanics model, tire curvature factor E1, nonlinear correction coefficient σ tire (E1,σ), comprehensively considering four core influencing factors—road speed coupled steady-state load, tire curvature stiffness nonlinear degradation, tire slip mechanical attenuation, and heavy-load transient dynamic impact—a quantitative model of the single-wheelset ultimate load capacity adapted to complex dynamic conditions is constructed. This model integrates multi-dimensional degradation and transient response characteristics to construct a quantitative formula for the single-wheelset performance boundary, representing the dynamic ultimate load capacity m of the single-wheelset. lim The calculation formula is shown in formula (95): (95); Single wheel pair ultimate lateral adhesion force F y, lim The calculation formula is shown in formula (96): (96); A tire lateral force wheel-pair rotation angle coupling equation is established to represent the dynamic relationship between rotation angle δ and the load-bearing and adhesion performance of a single wheel pair, as shown in formula (97): (97); In the formula, σ(δ) is the dimensionless composite slip variable that varies with the steering angle, and k δ δ0 is the steering slip correlation coefficient, where δ0 is the zero-slip reference steering angle of the wheelset; The input signal for this stage is the rated load capacity (m) of a single wheelset. rated Wheelset dynamic load correction factor k v (V, R), dimensionless comprehensive slip variable σ of tire mechanics model, tire curvature factor E1, nonlinear correction coefficient σ tire (E1, σ), load unevenness distribution coefficient η load The output signal is the dynamic ultimate bearing capacity m of a single wheelset. lim Single wheel pair ultimate lateral adhesion force F y, lim , rotation angle δ; Step Two: Calculation of Vehicle Capability Boundary Conditions under Specific Loads. This step, based on multi-condition carrying capacity, stability, and adhesion boundaries, calculates the basic vehicle capability boundary required to meet the load requirements under a specific load condition, including the total driving force F required by the vehicle. req From the highest vehicle speed V in the first step of the dynamics assessment max The maximum gradient α is determined as shown in formula (98): (98); In the formula, M cargo For cargo mass, the total braking force F required for the entire vehicle b, req The maximum braking deceleration 'a' given by the first step of the braking stability assessment x, max As shown in formula (99): (99); Total lateral force required for the vehicle F y, req The maximum safe lateral acceleration a given by the first step of the handling stability assessment y, max As shown in formula (100): (100); The input signal for this stage is the maximum vehicle speed V. max Maximum gradeability α, maximum braking deceleration a x, max Maximum safe lateral acceleration a y, max Vehicle weight M v Goods quality M cargo The output signal is the total driving force F required by the entire vehicle. req Total braking force required for the entire vehicle F b, req Total lateral force F required for the entire vehicle y, req ; Step 3: Reverse Derivation of Minimum Wheelset Count. This step, based on the microscopic performance boundaries of a single wheelset in Step 1 and the macroscopic boundaries of the entire vehicle in Step 2, employs a reverse derivation logic using performance parameters. Constrained by the compliance requirements of the first step (transportation safety) and the third step (steering stability), it independently derives the minimum wheelset count from four dimensions: load-bearing constraints, power constraints, braking constraints, and lateral stability constraints. This reverse derivation determines the minimum wheelset count n that satisfies safe driving under all operating conditions. load As shown in formula (101): (101); In the formula, F0 is the rated load of a single wheelset, and λ z The bearing capacity degradation coefficient is the road surface. Given the discrete variance of multi-axis loads, find the minimum number of wheelsets n based on driving force constraints. drive As shown in formula (102); (102); In the formula, λ d To drive the dimensional coupling weight coefficients, ΔF dyn (n) represents the dynamic transfer coupling term of axle load under accelerated conditions, indicating the degree of degradation of drive adhesion conditions caused by axle load redistribution during acceleration under different wheelset numbers, and is the minimum number of wheelsets n based on braking force constraints. brake As shown in formula (103): (103); In the formula, λ b Φ is the coupling weight coefficient for the braking dimension. brake (n) is the axle load mismatch function under braking conditions, characterizing the risk of wheel lock-up caused by axle load shifting during emergency braking, and is the minimum number of wheelsets n based on lateral stability constraints. lat As shown in formula (104): (104); In the formula, λ y , λ s For the lateral stability dimension coupling weight coefficients, Φ roll (n) is the multi-axis roll stiffness attenuation coupling term, characterizing the effect of wheelset number variation on the vehicle's equivalent roll stiffness and ultimate roll angle. P s (n) is the multi-axis steering geometric interference penalty function, and the minimum number of wheelsets in the global envelope is n. min As shown in formula (105): (105); Take the maximum value of the required number of wheelsets for each boundary, ensuring that the number of individual wheelsets simultaneously meets all requirements for driving force, braking force, lateral force, and load-bearing capacity. Compare the calculated result with the actual number of wheelsets n configured in the second step. If n ≥ n min The current configuration meets the carrying requirements if n < n min If so, the "insufficient wheel pair configuration" signal needs to be returned to the second step again, and the wheel pair configuration needs to be increased again; The input signal for this stage is the total driving force F required by the entire vehicle. req Total braking force required for the entire vehicle F b, req Total lateral force F required for the entire vehicle y, req Theoretical maximum carrying capacity of the whole vehicle M total The output signal is the minimum number of wheelsets n based on the bearing constraint. min ; Step 2, Wheelset Failure Performance Assessment Method. Step 2 consists of three steps, as follows: Step 1: Establishment of a wheelset failure mode discrimination mechanism. This step, based on the multi-wheelset steering system execution response, motor operating status, tire mechanical feedback, vehicle motion state parameters, and wheelset status monitoring data, identifies the failure modes of each wheelset. Model-based residual analysis is used for failure detection and isolation. The residual ϵ i (k)=u cmd, i (k)-u fb, i (k) represents the instruction value u cmd, i (k) and actual feedback value u fb, i The difference (k) when |ϵ i (k)|≥ϵ th If the failure continues for more than 5 sampling periods, it is determined that a failure has occurred. A quantitative classification criterion is established for the failure mode, as shown in formula (106): (106); In the formula, l i For the failure mode indication vector, ΔT lag To respond to the delay time, u max, sensor Δu is the upper limit of the sensor's measurement range. jump, max N represents the maximum allowable value for signal transitions. total, fail The cumulative number of judgment cycles for complete failure; The input signal for this stage is the command value u. cmd, i (k) Actual feedback value u fb, i (k), Response delay time ΔT lag Sensor range upper limit u max, sensor Maximum allowable value of signal transition Δu jump, max The cumulative number of complete failure judgment cycles N total, fail The output signal is the failure mode indicator vector l. i ; Step 2: Quantification of performance degradation of failed wheelsets. This step is based on the failure mode indicator vector l determined in Step 1. i The performance degradation degree of each failed wheelset is quantified, and the discrete failure type of each wheelset is mapped to a continuous performance degradation factor ρ. deg, i , representing the degree of performance loss of each failed wheelset relative to the healthy state, as shown in formula (107): (107); In the formula, T avail T represents the currently available torque. req For the target torque requirement, l i =0 indicates a normal state. After obtaining the degradation factors of each wheelset, the performance degradation level at the whole vehicle level is further evaluated. Different wheelsets bear different axle loads in the vehicle, and their impact on the overall vehicle performance has different weights. The comprehensive index of overall vehicle performance degradation ρ deg, total With the axle load F of each wheelset z, i The weighted average is calculated as shown in formula (108): (108); The input signal for this stage is the performance degradation factor ρ. deg, i Failure Mode Indication Vector l i Axle load F of each wheelset z, i The output signal is the overall vehicle performance degradation index ρ. deg, total ; Step 3: Steering performance verification under failure conditions. This step uses the overall vehicle performance degradation index ρ output from Step 2 as the basis for verification. deg, total As input, it is embedded into the vehicle dynamics model established in the second step. By modifying the steering angle and tire force output boundary of the corresponding failed wheelset, the three key steering performance indicators under the failure condition are recalculated: minimum steering radius R min Maximum safe steering speed V safe Yaw rate tracking error e r The performance degradation index is used to correct the steering angle and tire force output constraints of the failed wheelset, as shown in formula (109): (109); In the formula, δ i, max, fail F y, i, max, fail The maximum steering angle and maximum lateral tire force after correction for the failed wheelset; When a steering actuator fails, the dynamic response of a multi-axle heavy-duty vehicle exhibits strong nonlinearity, multi-axis coupling, and multi-timescale characteristics. A physical information neural network (PIN) is introduced to perform a high-precision evaluation of steering performance under failure conditions using a data model fusion-driven approach. The PIN model consists of four parts: an input layer, a feature extraction layer, a physical constraint layer, and an output layer. The input layer receives the current vehicle state vector X. fail As shown in formula (110): (110); In the formula, failID is the identifier encoding of the failed wheel pair. The feature extraction layer uses a bidirectional long short-term memory network to extract temporal features and embeds a self-attention mechanism to automatically identify the key features that have the greatest impact on the performance evaluation of the failure condition. The high-dimensional temporal input is mapped to the latent feature space, and the enhanced feature vector H is obtained after self-attention weighting. att As shown in formula (111): (111); In the formula, Att is the self-attention operator, BIL is the dynamic feature extracted by the bidirectional long short-term memory network, H(t-1) is the historical hidden state vector of the bidirectional long short-term memory network at time t-1, the physical constraint layer uses the prior constraints of the yaw and roll dynamics of the multi-axle vehicle to correct the network output, and the physical constraint loss term Loss phys As shown in formula (112): (112); In the formula, f dyn The multi-wheelset vehicle dynamics mapping function is obtained by inputting the corrected failed wheelset constraints and real-time vehicle state into a trained physical information neural network. The output layer directly maps the predicted values ​​of three key steering performance parameters under the failure condition. , , As shown in formula (113): (113); The input signal for this stage is the overall vehicle performance degradation index ρ. deg, total Maximum steering angle and maximum lateral tire force δ after correction of the failed wheelset i, max, fail F y, i, max, fail The current vehicle state vector X fail Physical constraint loss term Loss phys The output signal is the enhanced feature vector H after self-attention weighting. att Physical constraint loss term Loss phys Predicted minimum turning radius under failure conditions Predicted maximum safe steering speed under failure conditions yaw rate tracking error prediction value ; Step 3: Low-speed steering wheel pair coordination evaluation method. Step 3 consists of three parts, as follows: Step 1: Low-Speed ​​Steering Wheelset Status Acquisition. This step collects real-time status parameters of each wheelset under low-speed steering conditions. At low speeds, the tire slip angle is small, and the wheelset kinematics are closer to the Ackerman steering geometry. Therefore, the data collected under this condition more accurately reflects the steering coordination between multiple wheelsets. The theoretical Ackerman angle δ of each wheel... i, theory As shown in formula (114): (114); In the formula, R ack δ1 is the turning radius of the first axis, and δ1 is the actual turning angle of the first axis; Based on this, real-time status parameters of each wheelset are collected and calculated, including the actual steering angle δ of each wheel. i Rate of change of rotation angle of each wheel Tire slip angle α of each wheel i Tire wear rate W for each wheel rate, i After aligning the four types of data by timestamp, they are organized into a low-speed steering wheelset state matrix S, with wheelsets as the unit. low As shown in formula (115): (115); In the formula, each row in the matrix corresponds to a wheelset, the row number is the axle number, the first axle to the nth axle, and the four columns are the actual steering angle, the rate of change of steering angle, the tire slip angle and the tire wear rate of each wheel, respectively. The input signal for this stage is the first axle steering radius R. ack 1. Actual steering angle δ1 of the first axis; 2. Longitudinal distance L from the first axis to the i-th axis. i Actual steering angle δ of each wheel i Rate of change of rotation angle of each wheel Tire slip angle α of each wheel i Tire wear rate W for each wheel rate, i The output signal is the theoretical Ackermann angle δ for each wheel. i, theory Low-speed steering wheel pair state matrix S low ; Step 2: Calculation of wheel set wear degree. This step is based on the low-speed steering wheel set state matrix S collected in Step 1. low The wear rate is used to assess the unevenness of wear among wheelsets. The wear rate of each wheelset reflects the wear rate of each wheelset per unit time. The wear rates of each wheelset collected in step one are compared to determine whether the wear is reasonable. The average wear rate W of all wheelsets is then calculated. rate, avg As shown in formula (116): (116); The wear index W of the i-th pair uneven, i Defined as the relative degree to which the wear rate of the wheelset deviates from the average wear rate, as shown in formula (117): (117); In the formula, ε W For custom small quantities, to prevent the denominator from being zero, W uneven, i A larger value indicates a more severe deviation of the wheelset's wear from the average level, and a higher risk of uneven wear. The wheelet number i with the most severe uneven wear is also identified. max =argmax i W uneven, i and its wear index W uneven, max The overall vehicle wear uniformity index U wear Defined as the average compensation value of the wear index of all wheelsets, as shown in formula (118): (118); U wear The closer to 1, the more uniform the wear of each wheel pair; the closer to 0, the more severe the uneven wear. When the result indicates severe uneven wear, the steering control strategy needs to be optimized and adjusted, and the weight of the steering angle distribution of each axle needs to be re-optimized to reduce the load on the high-wear wheelsets. The input signal for this stage is the low-speed steering wheel pair state matrix S. low The output signal is the average wear rate W of the wheelset. rate, avg The wear index W of each wheelset uneven, i The wheelset with the most severe uneven wear is numbered i max Overall vehicle wear uniformity index U wear ; Step 3: Comprehensive Calculation of Wheelset Coordination. This step comprehensively calculates the low-speed steering wheelset coordination based on three dimensions: the deviation of each wheelset's steering angle from the theoretical Ackermann steering angle, the consistency of each wheelset's slip angle, and the uniformity of wear. The coordination degree C of each wheelset is calculated. sync, i As shown in formula (119): (119); In the formula, ω δ ω α ω W α is the weighting coefficient. max The critical sideslip angle is the low-speed steering wheel pair coordination degree C of the entire vehicle. sync, total The arithmetic mean of the coordination of all wheelsets is given by formula (120): (120); The input signal for this stage is the actual steering angle δ. i Theoretical Ackermann rotation δ i, theory Actual sideslip angle α i Mean sideslip angle α avg Critical sideslip angle α max The wear index W of each wheelset uneven, i The output signal is the coordination degree C of each wheelset. sync, i Low-speed steering wheel pair coordination degree C of the whole vehicle sync, total ; Step 4: High-speed steering stability assessment method. Step 4 consists of three parts, as follows: Step 1: Extraction of Dynamic Response Characteristics for High-Speed ​​Steering. This step focuses on the high-speed steering conditions of heavy-duty vehicles, constructing a standardized test case set S that includes constant speed lane change, high-speed step steering, and high-speed sinusoidal steering. c The constant speed lane change test is used to evaluate the lateral stability and yaw response quality of a vehicle in a highway scenario; the high-speed step steering test is used to obtain the transient response characteristics of the vehicle's yaw rate, including response time and overshoot; the high-speed sinusoidal steering test is used for frequency domain identification, extracting the amplitude attenuation and phase lag characteristics at different excitation frequencies, and the extracted time-domain response signal includes the steering wheel angle δ. i Yaw angular velocity r(t), roll angle ϕ(t), lateral acceleration a y (t), Vertical force F of each wheel zi (t) , The collected data is filtered and denoised to remove outliers, forming a standardized test dataset S. d ; The input signal for this stage is the steering wheel angle δ. i Yaw angular velocity r(t), roll angle ϕ(t), lateral acceleration a y (t), Vertical force F of each wheel zi (t); the output signal is the standardized test condition set S. c Standardized test dataset S d ; Step 2: Calculation of high-speed steering stability index. This step is based on the standardized test dataset S extracted in Step 1. d The core stability evaluation indices are calculated, and the yaw oscillation attenuation coefficient, high-speed trajectory tracking accuracy, high-speed lateral stability margin, and yaw oscillation attenuation rate ζ are solved respectively. yaw The calculation is shown in formula (121): (121); In the formula, n is the total number of oscillation peaks counted in the yaw rate decay oscillation waveform, and A r1 A represents the amplitude of the first peak of the yaw oscillation. rn The amplitude of the nth peak of the yaw oscillation is given by E, representing the high-speed trajectory tracking accuracy error. y, high The calculation is shown in formula (122): (122); In the formula, N is the total number of sampling points, Y ref Y represents the desired lateral trajectory coordinates. act The actual lateral trajectory coordinates; high-speed lateral stability margin K stab, high The calculation is shown in formula (123): (123); The input signal for this step is the standardized test dataset S. d The output signal is the yaw oscillation decay rate ζ. yaw High-speed trajectory tracking accuracy error E y, high High-speed lateral stability margin K stab, high ; Section 3: Comprehensive Rating of High-Speed ​​Steering Stability. This section will assess the stability of the three core indicators calculated in Section 2: yaw rate attenuation ζ. yaw High-speed trajectory tracking accuracy error E y, high High-speed lateral stability margin K stab, high A weighted fusion method is used to establish a comprehensive rating for high-speed steering stability. The three indicators characterize high-speed steering stability from three dimensions: transient response quality, trajectory holding capability, and roll safety margin. yaw E reflects the vehicle's damping characteristics in response to steering input; a larger value indicates faster yaw rate convergence. y, high This reflects the vehicle's ability to track the desired path; a smaller value indicates higher tracking accuracy. K stab, high Reflecting a vehicle's ability to resist rollover, a higher value indicates a greater lateral stability margin. The overall rating for high-speed cornering stability is R. high As shown in formula (124): (124); In the formula, λ h1 , λ h2 , λ h3 ζ is the weighting coefficient. yaw, ref For reference to the yaw oscillation decay rate, E y, allow To allow for the maximum trajectory tracking error, K stab, ref For reference to lateral stability margin, the rating criteria are as follows: R high A value of ≥0.85 is considered excellent, indicating high safety during high-speed steering, superior handling quality, fast yaw convergence, accurate trajectory tracking, and ample roll margin. 0.70≤R high A value <0.85 is considered good, meeting the basic safety requirements for high-speed steering; 0.55 ≤ R high A value <0.70 is acceptable, but this indicates a risk of roll or trajectory deviation, requiring limitation of the maximum steering speed or optimization of tire lateral stiffness matching; R high A value <0.55 is considered unacceptable, indicating a safety hazard during high-speed steering. The suspension roll stiffness, wheelset steering angle distribution strategy, or tire selection need to be reassessed. The input signal for this stage is the yaw rate decay ζ. yaw High-speed trajectory tracking accuracy error E y, high High-speed lateral stability margin K stab, high Reference yaw oscillation decay rate ζ yaw, ref Maximum allowable trajectory tracking error E y, allow Reference lateral stability margin K stab, ref The output signal is the comprehensive rating of high-speed steering stability, R. high Rating results.

Citation Information

Patent Citations

  • Multi-axial vehicle all-wheel steering control method and multi-axial vehicle all-wheel steering control system

    CN102514619A

  • System and method for improving safety of wheels of heavy multi-axle vehicle during steering jamming

    CN114560011A

  • Steering control method of heavy-duty vehicle capable of being dragged

    CN116001910A