A self-verifying estimation system and method for longitudinal force at the main hanger hinge point

By constructing redundant pseudo-measurements and physical consistency verification methods, the problems of timing differences and parameter identification of master and slave information are solved, and reliable estimation and credible output of longitudinal force at articulation points are achieved without relying on dedicated sensors.

CN123116609APending Publication Date: 2026-11-17JILIN UNIVERSITY
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202611397992.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-09
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies struggle to effectively address the temporal differences in tractor-trailer information, uncertainties in longitudinal force reconstruction, and the identifiability of slow time-varying parameters without relying on dedicated articulated force sensors. Furthermore, they lack redundant verification of dynamic results on both the tractor and trailer sides, making it difficult to distinguish between bilateral physical conflicts and unilateral inconsistencies, resulting in insufficient reliability of the estimation results.

Method used

By constructing redundant pseudo-measurements of longitudinal dynamic information on the tractor side and trailer side, adaptive observation and physical consistency verification are performed. Anomaly judgment and graded degradation output are performed using the two-sided physical feasible set. The system includes modules such as comprehensive information acquisition, time-series synchronization, axle longitudinal force reconstruction, selective identification of slow time-varying parameters, redundancy estimation of dynamics on both the tractor and trailer sides, and common difference mode consistency observation.

Benefits of technology

It enables reliable identification of longitudinal force at hinge points when there is information conflict on both sides or partial failure of estimation channels, reduces synchronization error, reduces mutual compensation of parameters, improves the reliability and continuity of estimation results, and ensures the interpretability and credibility of the output.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123116609A_ABST
    Figure CN123116609A_ABST
Patent Text Reader

Abstract

This invention discloses a self-verifying estimation system and method for longitudinal force at the main-trailer articulation point, belonging to the field of vehicle dynamics state estimation and intelligent control technology. The system acquires driving, braking, longitudinal motion, vehicle parameters, and communication time information from both the tractor and trailer sides, completing unified time-base synchronization and longitudinal force reconstruction. It selectively identifies trailer mass, slope correction, and driving resistance parameters based on the longitudinal dynamic relationship between the main and trailer sides, and constructs bilateral pseudo-measurements of the longitudinal force at the same articulation point. It adaptively adjusts the observation uncertainty using common-mode and differential-mode residuals, and estimates the results based on the intersection constraints of the bilateral physical feasible sets. When the intersection is empty, it does not force projection and identifies candidate anomaly sources. After trailer swapping, it selectively reinitializes the trailer's relevant states and implements bilateral fusion, healthy single-side, short-term prediction maintenance, or failure mode output based on the channel health status. This invention is used for main-trailer cooperative driving, cooperative braking, energy management, and connection load monitoring.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of vehicle dynamics state estimation and intelligent control technology, specifically relating to a self-verifying estimation system and method for the longitudinal force of the main-trailer articulation point, which uses longitudinal dynamic information from the tractor side and trailer side to estimate, evaluate the reliability of, verify the physical consistency of, and handle anomaly degradation of the longitudinal force at the main-trailer articulation point. Background Technology

[0002] A tractor and trailer form a combined vehicle via a drawbar and draw pin, a drawbar and trailer clasp, or other articulated connection structures. During vehicle driving, coasting, braking, and hill-climbing, longitudinal tension or thrust is transmitted between the tractor and trailer through the articulation points. These longitudinal forces at the articulation points are crucial state variables for monitoring tractor-trailer coordinated driving, coordinated braking, energy management, and load monitoring of connecting components. Existing methods for obtaining articulation force mainly include direct measurement and indirect determination based on dynamic models. For example, Chinese patent CN111458058B discloses a measurement scheme for obtaining articulation force by setting strain gauges on the drawbar clasp of a center-axle trailer and using calibration curves; US patent application US20240010036A1 discloses a scheme for determining articulation force components based on vehicle and trailer dynamic models.

[0003] Direct measurement methods typically require the installation and calibration of dedicated force-measuring elements along the load-bearing path of the main-trailer connection components. Under conditions of impact loads, long-term vibration, temperature variations, structural wear, and frequent trailer replacements, sensor durability, zero-point drift, installation consistency, and maintenance costs must also be considered. While indirect estimation methods based on dynamic models can utilize existing vehicle information, their estimation results are susceptible to incomplete motor torque and mechanical braking information, wheel system dynamic compensation deviations, wheel slippage, asynchronous sampling of main-trailer information, and changes in trailer mass, road gradient, and driving resistance parameters. Furthermore, insufficient longitudinal excitation can lead to mutual compensation or erroneous convergence of trailer mass, gradient, and resistance parameters. Simultaneously, existing solutions often lack redundant verification of the dynamic results on both the tractor and trailer sides, making it difficult to distinguish between shared offsets and unilateral inconsistencies. When the physical allowable ranges on both sides are inconsistent, the trailer is replaced, or some information becomes invalid, it is also difficult to simultaneously address anomaly identification, estimation continuity, and output reliability.

[0004] Therefore, a longitudinal force estimation system for the main-trailer articulation point is needed. Without relying on a dedicated articulation force sensor, it should be able to comprehensively handle the time-series differences in main-trailer information, the uncertainty of longitudinal force reconstruction, and the identifiability of slow time-varying parameters. It should utilize the dynamic information of the tractor side and the trailer side to perform redundant estimation and physical consistency verification of the longitudinal force at the articulation point. In the event of information conflict between the two sides, trailer replacement, or failure of some estimation channels, it should determine the credible and effective states of the estimation results and implement corresponding anomaly handling and graded degradation output. Summary of the Invention

[0005] The purpose of this invention is to provide a self-verifying estimation system and method for longitudinal force at the main-trailer articulation point, in order to solve the problems of time-series differences in main-trailer information, the influence of multi-source errors on wheel longitudinal force reconstruction results, the difficulty in simultaneously and reliably identifying trailer mass, road slope and driving resistance parameters, the lack of redundant verification in unilateral dynamic estimation, the difficulty in attributing bilateral physical conflicts, and the insufficient reliability of output after trailer swapping or partial information failure.

[0006] To achieve the above objectives, this invention utilizes longitudinal dynamic information from both the tractor side and the trailer side to construct redundant pseudo-measurements of the longitudinal force at the same articulation point. The redundant pseudo-measurements are adaptively observed using temporal reliability, longitudinal force reconstruction reliability, parameter identifiability, and bilateral innovative information. Furthermore, the observation results are physically consistent with the bilateral physical feasible set. When bilateral physical conflicts, trailer replacements, or partial channel failures occur, further steps are taken to determine candidate anomaly sources, perform selective re-initialization, and implement graded degradation output.

[0007] The self-verification estimation system and method for longitudinal force at the main-trailer articulation point described in this invention includes a comprehensive information acquisition and time-series synchronization module, an axle longitudinal force reconstruction module, a slow time-varying parameter selective identification module, a main-trailer dual-side dynamic redundancy estimation module, a common difference mode consistency adaptive observation module, and a dual-side physical feasible domain self-verification and error attribution module; it also includes a trailer replacement identification and selective re-initialization module, and a fault confirmation and graded degradation output module.

[0008] The integrated information acquisition and timing synchronization module provides synchronization information and timing status corresponding to a unified time base to the axle longitudinal force reconstruction module, the slow time-varying parameter selective identification module, the main-trailer dual-side dynamic redundancy estimation module, and the common-mode consistency adaptive observation module. The axle longitudinal force reconstruction module and the slow time-varying parameter selective identification module provide the main-trailer dual-side dynamic redundancy estimation module with longitudinal forces on both sides, vehicle parameters, and their uncertainties. The main-trailer dual-side dynamic redundancy estimation module provides pseudo-measurements on the tractor side and trailer side to the common-mode consistency adaptive observation module. The common-mode consistency adaptive observation module forms a longitudinal force correction estimate at the articulation point, and the dual-side physical feasible region self-verification and error attribution module performs physical constraints and conflict attribution.

[0009] The trailer replacement identification and selective re-initialization module is used to update trailer-related parameters and observation status after confirming a trailer replacement, while retaining the still valid tractor parameters and road status. The fault confirmation and graded degradation output module selects between two-sided fusion estimation, healthy one-sided estimation, short-term prediction maintenance, and failure mode based on the health status of both sides of the channel, the health status of the common input, the physical feasible set status, and the estimation uncertainty. It outputs the longitudinal force estimate of the articulation point, the estimation uncertainty, the confidence status, the effective status, the current estimation mode, and the candidate anomaly sources.

[0010] The integrated information acquisition and timing synchronization module acquires the actual motor torque, motor speed or wheel speed, and mechanical braking information of each electric drive shaft on the tractor side and trailer side; the longitudinal speed, longitudinal acceleration, attitude, and articulation status of the tractor and trailer; the road gradient; vehicle structure and transmission parameters; the main-trailer connection status; trailer identification information; and the source timestamp, frame number, receiving timestamp, and communication status corresponding to each piece of information; in the... Establish a unified estimation time for each estimation period. It performs filtering rate estimation, time delay compensation or interpolation on continuously changing signals, timestamp matching or recent valid state preservation on discrete state information, and prohibits continuous extrapolation across discrete state switching moments.

[0011] For any information channel involved in the estimation Based on the valid sampling time of this information The effective delay and the coupling amount between the effective delay and the signal rate of change are determined according to the following formula: Delay-Dynamic Coupling Mismatch

[0012]

[0013] In the formula, For information channels The time delay reference scale, The filtered rate of change of the corresponding signal. As a reference scale for the amplitude of the corresponding signal, It is a dimensionless quantity.

[0014] The estimation period The time delay reference scale for rapidly changing information such as motor torque, wheel speed, longitudinal speed, and longitudinal acceleration is set at 0.008–0.015 s, with a typical value of 0.010 s. The time delay reference scale for determining information with slow changing trends is set at 0.5–0.75 s. Timing normal threshold. Set the timing failure threshold to 0.8–1.5. Set between 4 and 9, recovery threshold Set the value to be between 0.36 and 0.64, and satisfy the following conditions: When the mismatch is not higher than the normal timing threshold, the corresponding information is used normally. When the mismatch is between the normal timing threshold and the timing failure threshold, the corresponding information is marked as timing-limited, and the relevant uncertainty is increased or the usage weight is reduced. When the mismatch reaches the timing failure threshold, the corresponding information stops participating in parameter updates or pseudo-measurement correction. Timing anomalies are confirmed after 3 to 5 consecutive estimation cycles, and recovery is achieved after 5 to 10 consecutive estimation cycles that meet the recovery threshold. Data ages of rapidly changing information exceeding 0.05s are directly judged as timeouts.

[0015] The axle longitudinal force reconstruction module establishes index sets for the electric drive axles on the tractor side and trailer side, respectively. and And denote the union of the two as For any electric drive shaft The electric drive wheel end torque is determined based on the actual motor torque (with directional markings), transmission ratio of the transmission stage, transmission ratio of the main reducer, and torque transmission coefficient corresponding to the power flow direction of driving, coasting, or energy recovery. The actual mechanical braking torque (with directional markings) is then superimposed, and the wheel train inertial torque and internal transmission system loss torque are deducted. The longitudinal force foundation reconstruction value is then determined according to the following formula:

[0016]

[0017] In the formula, This represents the actual motor torque. and These are the transmission ratio of the transmission stage and the transmission ratio of the main reducer, respectively. The torque transmission coefficient, This is the actual mechanical braking torque. To calculate the equivalent moment of inertia at the wheel end, This is the filtered wheel angular acceleration. Torque loss within the transmission system The effective rolling radius.

[0018] The upper limit of longitudinal mechanical braking force for a single tire is 9.0 × 10³ N; the upper limit of mechanical braking torque at the wheel end on one side is determined based on the number of tires at that wheel end, the upper limit of longitudinal mechanical braking force for a single tire, and the effective rolling radius; for non-electrically driven brake axles, the mechanical braking torque at the marked wheel ends of each brake wheel on that axle is converted into additional longitudinal force according to the effective rolling radius, and the corresponding longitudinal resultant force and its uncertainty on the tractor side or trailer side are included.

[0019] The uncertainty of motor torque, mechanical braking torque, torque transmission coefficient, effective rolling radius, wheel angular acceleration, and internal loss torque is determined based on the following: Longitudinal force reconstruction variance of electric drive shaft The wheel slip index is determined based on the wheel circumferential speed and the vehicle longitudinal speed. The mismatch is reconstructed according to the following formula:

[0020]

[0021] In the formula, As a longitudinal force reference scale, As a sliding reference scale, This represents the effect coefficient of timing mismatch.

[0022] The longitudinal force reference scale Set at N, Slip reference scale Set to 0.08–0.12, the timing mismatch effect coefficient The thresholds are set to 0.1–0.3; the slip warning threshold is set to 0.08–0.12; the slip invalid threshold is set to 0.18–0.25; the normal reconstruction threshold is set to 1.2–2.0; and the reconstruction failure threshold is set to 4.0–6.0. ​​Based on these thresholds, the reconstruction state of each axis is divided into normal, restricted, and invalid. In the restricted state, the reconstruction uncertainty is increased, and the corresponding results are prohibited or restricted from being used for slow time-varying parameter updates. If the restricted state fails to recover after 10–30 consecutive estimation cycles, it is converted to invalid. When the actual mechanical braking force is missing under composite braking conditions, the mechanical braking torque is not treated as zero; instead, the corresponding uncertainty is increased, or pseudo-measurements containing the corresponding axis information are stopped from participating in observation correction.

[0023] The slow time-varying parameter selective identification module converts the longitudinal driving force, longitudinal acceleration, and driving resistance of the tractor side and trailer side into a unified road longitudinal coordinate system, and adds the longitudinal dynamic equations of both sides to eliminate the longitudinal forces at the hinge points with opposite directions in the equations, forming the following longitudinal dynamic relationship of the whole vehicle:

[0024]

[0025] In the formula, and These are the equivalent mass of the tractor and the equivalent mass of the trailer, respectively. and These are the longitudinal accelerations on both sides, and These are the longitudinal driving forces on both sides. and These represent the equivalent driving resistance on both sides.

[0026] The slow time-varying parameter selective identification module determines the road slope benchmark using at least one of the following: vehicle-mounted slope sensing information, inertial measurement information, satellite positioning information, electronic map information, or existing slope observation results, and estimates the slope correction only within a bounded range. Based on the vehicle speed range, longitudinal acceleration excitation, road slope change, driving force change, normalized sensitivity energy of candidate parameters, and parameter correlation, it selects a subset of currently active parameters from trailer equivalent mass, slope correction, rolling resistance parameters, air resistance parameters, and unmodeled resistance correction. Parameters outside the active parameter subset remain at the previous valid estimate or prior value.

[0027] An information matrix is ​​constructed using the normalized sensitivity of a subset of active parameters within a preset sliding time window. The input comprehensive credibility factor is determined based on the credibility of driving force reconstruction, timing synchronization, and longitudinal motion state. The identifiable incentive index can be calculated using the following formula:

[0028]

[0029] In the formula, and These are the minimum and maximum eigenvalues ​​of the matrix, respectively. This is the denominator protection quantity.

[0030] The lower threshold for incentive is set at 0.02–0.05, the upper threshold at 0.08–0.15, and the lower limit of the input comprehensive confidence factor is set at 0.60–0.80. When the identifiable incentive index is not higher than the lower threshold, the subset of active parameters is frozen; when it is between the lower and upper thresholds, the single-cycle update amplitude of the parameters is limited; when it reaches the upper threshold, normal updates are allowed. The information matrix is ​​constructed using a sliding time window of 1–3 seconds, with a normal upper limit for the condition number set at 20–50 and a restricted update upper limit set at 100–500; the upper limit for parameter correlation is set at 0.90–0.97. When the correlation between candidate parameters exceeds the upper limit or the information matrix does not meet the rank-sum conditional requirement, the subset of active parameters is reduced and re-evaluated.

[0031] The equivalent trailer mass is constrained to 7810–42210 kg, the absolute upper limit of the road slope correction is set at 0.0035–0.0087 rad, and the rolling resistance coefficient is constrained to 0.004–0.009. The upper limit of the single-cycle update range for the equivalent trailer mass, slope correction, and rolling resistance coefficient is set at 0.05%–0.2% of the current value of the equivalent trailer mass, respectively. rad and In an unidentifiable state, the slope correction, trailer equivalent mass, and all driving resistance parameters are not updated simultaneously.

[0032] The dual-side dynamic redundancy estimation module for the main trailer establishes a unified road longitudinal coordinate axis based on the road tangential direction at the articulation point location. The longitudinal force at the articulation point is defined as the component of the force exerted by the main trailer at the articulation point on this unified road longitudinal coordinate axis, with the direction of the tractor's pulling force on the trailer as positive and the direction of the trailer pushing the tractor as negative. Based on the longitudinal dynamic relationship between the tractor side and the trailer side, pseudo-measurements for the longitudinal force at the same articulation point are constructed respectively.

[0033]

[0034] In the formula, and The longitudinal force at the articulation points on the tractor side and trailer side were measured separately.

[0035] Based on the joint uncertainty of the driving force, equivalent mass, longitudinal acceleration, road slope, driving resistance, and timing synchronization information used in the pseudo-measurements on both sides, the variance, basic reliable state, and effective state of the pseudo-measurements on both sides are determined, while retaining the correlation of errors on both sides caused by road slope, air density, articulation state, and common motion state information. The upper limit of the allowable standard deviation of the pseudo-measurement on the tractor side is set at [value missing]. N, the upper limit of the permissible standard deviation for trailer side pseudo-measurement is set at... N; Valid pseudo-measurements are generated only when the corresponding key input is valid and the pseudo-measurement variance does not exceed the allowable upper limit. Decreased credibility is represented by increasing the measurement uncertainty or reducing the fusion weight, not by directly reducing the pseudo-measurement force value.

[0036] The common-mode consistency adaptive observation module obtains the predicted value of the longitudinal force at the hinge point for the current estimation period based on the estimated value of the longitudinal force at the hinge point in the previous estimation period, the current change in vehicle driving force, and the longitudinal motion state. And prediction uncertainty, based on the deviation of the two sides of the pseudo-measurement relative to the predicted value, construct the tractor side innovation. Trailer side innovation Common mode residual Sum and difference residuals :

[0037]

[0038] In the formula, and The weights are non-negative weights determined based on the basic credible states, pseudo-measurement uncertainties, and effective states on both sides, and the sum of the two is 1.

[0039] Determining the common-mode residual variance while considering the cross-correlation of pseudo-measurements on both sides. Sum and difference model residual variance The dimensionless normalized common modulus statistic and the normalized difference modulus statistic are formed according to the following formula:

[0040]

[0041] In the formula, It is the denominator protection quantity with the dimension of force variance.

[0042] Based on the respective basic innovation variances of both sides Determining the inconsistency of the differential mode in the first Innovation contribution coefficient of the side And form the driving force for adjusting process uncertainty. and the Side measurement uncertainty adjustment drive quantity :

[0043]

[0044] In the formula, For dimensionless denominator protection quantities, Indicates the side of the tractor unit. Indicates the trailer side.

[0045] The entry thresholds for both the common-mode and differential-mode statistics are set between 6.63 and 10.83, and the recovery thresholds are set between 2.71 and 3.84. When neither the common-mode nor differential-mode statistics exceed the entry thresholds, the basic process uncertainty and measurement uncertainty are maintained. When the common-mode statistics exceed the entry thresholds but the differential-mode statistics do not, the process uncertainty is adjusted only if there is shared healthy input and the common offset is supported by changes in driving force or acceleration. When the common-mode statistics do not exceed the entry thresholds but the differential-mode statistics do, the measurement uncertainty of the side with the larger contribution is increased or its fusion weight is decreased based on the innovation contribution and independent health information. When both statistics exceed the entry thresholds, updates to relevant slow time-varying parameters are paused, and the process transitions to physical feasibility domain verification.

[0046] Common-mode anomalies or differential-mode anomalies are confirmed after exceeding the entry threshold for 3–5 estimation cycles consecutively, and recover after not exceeding the recovery threshold for 8–20 estimation cycles consecutively. The absolute change in fusion weights within a single estimation cycle should not exceed 0.05–0.10, and the upper limit of the single-cycle amplification factor for process uncertainty and measurement uncertainty is set at 1.2–1.5. Differential-mode anomalies are only used to measure the degree of inconsistency between the two sides and are not used as sufficient evidence to determine sensor failure, longitudinal force reconstruction failure, or parameter failure. Predictive-innovative correction is performed using adjusted process uncertainty, measurement uncertainty, or fusion weights to obtain the estimated value of longitudinal force correction at the articulation point. And its estimated uncertainty.

[0047] The dual-sided physical feasible region self-verification and error attribution module outputs physical self-verification estimates or transient estimates, estimation uncertainties, physical consistency states, physical conflict states, and candidate anomaly sources, forming a basic estimation result of the longitudinal force at the articulation point. The fault confirmation and graded degradation output module selects an estimation mode based on the basic estimation result, the health status of both sides of the channel, and the health status of the common input, determines the output credibility and output validity flag, and forms the final system output. For the tractor side and trailer side, it combines the physical boundaries, estimation uncertainties, inter-correlation relationships, and valid states of the longitudinal driving force, equivalent mass, longitudinal acceleration, road slope, driving resistance, articulation state, and timing compensation amount into a joint permissible set. And through the first Side dynamics mapping Form the physically feasible set of longitudinal forces at the hinge point on this side, and determine the intersection of the physically feasible sets on both sides:

[0048]

[0049] In the formula, Indicates the first Side-joint allows for the combination of dynamic inputs and vehicle parameters in the set. and These are the physically feasible sets for the tractor side and the trailer side, respectively. It is a common physically feasible set.

[0050] When the common physical feasible set is not empty, the physical self-verification estimate with the smallest deviation from the corrected estimate is determined within this set; when the common physical feasible set is empty, forced projection is not performed, and the degree of physical conflict is characterized by the minimum distance between the two physical feasible sets.

[0051]

[0052] In the formula, Represents the projection of the common physically feasible set. This is a physically self-verified estimate. It is the minimum distance between the two physically feasible sets.

[0053] The physical boundary for vehicle longitudinal acceleration is set to an upper limit of 0.81 m / s² for the driving condition, depending on the operating conditions. 2 The lower limit of normal braking conditions is -1.04 m / s. 2 And the lower limit of emergency braking condition is -4.5m / s 2 Longitudinal force reference scale at hinge point Set at N, with As a normalized conflict level, the conflict alert threshold is set to... ,correspond The severe conflict threshold is set to... ,correspond .

[0054] When the common physical feasible set is empty and the set distance is not higher than the conflict warning threshold, maintain transient estimation, increase estimation uncertainty, and continue observation; when the set distance is between the conflict warning threshold and the severe conflict threshold, it is judged as a restricted conflict and the adjustment range of fusion weights is limited; when the set distance reaches the severe conflict threshold and lasts for 3 to 5 consecutive estimation cycles, candidate anomaly attribution and downgrading processing are initiated. Bounded relaxation is performed on the joint allowable subsets corresponding to longitudinal force reconstruction, vehicle parameters, longitudinal motion state, road and driving resistance, temporal synchronization, and articulated state, with the maximum bounded relaxation amount set at [value missing]. ,correspond If the common intersection cannot be restored after the bounded relaxation, the process will proceed to the common input anomaly or multi-source anomaly handling, without forcing the selection of either side.

[0055] Based on the minimum relaxation category and relaxation degree required to restore the intersection of the two physically feasible sets, and combined with the temporal health status, longitudinal force reconstruction health status, motion state health status, parameter identifiability status, and conflict duration of the corresponding channels, candidate anomaly sources and their contribution order are determined. Only after continuous confirmation are the corresponding channels' bounded weight reduction, corresponding parameter update freeze, resynchronization, reconstruction restriction, or observation mode switching triggered. The common physically feasible set is restored to non-empty, and the set distance is no higher than [a certain value] for 5–10 consecutive estimation periods. Afterwards, the restricted conflict state is lifted.

[0056] The trailer replacement identification and selective re-initialization module identifies trailer replacement events in the following order: trailer unique identity information, reconnection event after main trailer disconnection, trailer structure or drive fingerprint, and mass or resistance auxiliary features. When a change in trailer unique identity exists, a replacement candidate is formed by combining the connection status; when unique identity information is missing, a reconnection event must be detected simultaneously, and the difference must be a normalized structure or drive fingerprint value ranging from 0 to 1. Exceeding the fingerprint difference threshold set between 0.10 and 0.25 Changes in trailer mass or driving resistance are only supplementary information. A trailer replacement is confirmed if the candidate trailer meets the confirmation criteria for 20–50 consecutive estimation periods.

[0057] After confirming the trailer replacement, the slow time-varying parameters are divided into tractor parameter subvectors, trailer parameter subvectors, and current road parameter subvectors. Only the trailer parameter subvector and its corresponding covariance are replaced with the new trailer's prior values ​​and their prior uncertainties, while the still valid tractor and road parameters are retained.

[0058]

[0059] In the formula, and These are the initial values ​​of the new trailer parameters and the initial covariance, respectively. The parameters and covariance marked with "-" are the corresponding estimates that are still valid before the trailer replacement.

[0060] After the trailer swap, clear or reinitialize the longitudinal force reconstruction history, parameter recursion history, bilateral innovation statistical state, bilateral measurement uncertainty adaptive state, physical feasible set, physical conflict state, and error attribution history related to the original trailer. Use the basic fusion results of the effective bilateral pseudo-measurements of the new trailer or the current vehicle dynamics prediction results to set the initial value of the longitudinal force observation at the articulation point, and initialize the observation uncertainty according to the prior uncertainty of the new trailer. The trailer swap recovery must continuously meet the following conditions for 2–10 seconds: the identifiable excitation index is not lower than the recovery threshold set between 0.08 and 0.15, the differential modulus statistic is not higher than the recovery threshold set between 2.71 and 3.84, the bilateral common physical feasible set is non-empty, and the variance of the longitudinal force estimation at the articulation point is not higher than the recovery upper limit. The recovery upper limit is set at N 2 And below the upper limit of estimated variance failure.

[0061] The fault confirmation and degradation output module determines the health reliability of both estimated channels based on the temporal reliability status, longitudinal force reconstruction reliability status, motion state validity, parameter estimation validity, pseudo-measurement innovation and its uncertainty, physical feasible set status, and candidate anomaly sources of both the tractor and trailer sides. Anomalies caused by road gradient, common velocity or acceleration, unified time base, and other shared information are marked as shared input anomalies and not directly assigned to any single channel. The non-intersection of the two sides' physical feasible sets does not automatically indicate simultaneous failure of both channels; only after combining candidate anomaly sources, independent health information, and conflict persistence status does it determine the isolated anomaly side.

[0062] No. Side channel health reliability and output credibility The values ​​range from 0 to 1. Channel fault threshold. Set between 0.20 and 0.40, the channel recovery threshold. Set to 0.65–0.80; lower limit of output confidence. Given a range of 0.50 to 0.70, estimate the upper limit of variance. Set at N 2 Maximum holding time The timeframe is set between 0.2 and 0.4 seconds. The abnormal state is confirmed after 3 to 5 consecutive estimation cycles, and the hysteresis interval is defined by the fault entry threshold and a higher recovery threshold.

[0063] When the health confidence of both channels is not lower than the recovery threshold, the common input is healthy, and the common physical feasible set is not empty, the two-sided fusion estimation mode is adopted; when the health confidence of one channel is not higher than the fault entry threshold and the other side is not lower than the recovery threshold, or when the anomaly can be isolated to one side through continuous confirmation, the corresponding health one-sided estimation mode is adopted; when both sides are temporarily invalid, the common input is temporarily abnormal, or the physical conflict cannot be reliably attributed, but the hold time, estimation variance, and output confidence still meet the corresponding boundaries and the information required for state prediction is valid, the short-term prediction hold mode is adopted; when any of the hold time, estimation variance, or output confidence exceeds the failure boundary, or the information required for state prediction is lacking, the failure mode is entered.

[0064] Upon recovery from a failure mode or short-term prediction-maintained mode, the system first satisfies the recovery threshold for at least one channel, the unilateral physical feasibility condition, and the output boundary condition for 10–30 consecutive estimation cycles to recover to a healthy unilateral estimation mode. Only after the conditions for both channels, shared input, and common physical feasibility set are continuously satisfied for 20–50 estimation cycles does the system recover to a unilateral fusion estimation mode. The output valid state is set to valid only when the current mode is not a failure mode, the output confidence is not lower than the lower limit of output confidence, and the estimation variance is not higher than the upper limit of estimation variance.

[0065] The present invention also provides a self-verification estimation method for the longitudinal force at the main hanger hinge point corresponding to the above system, the method comprising the following steps:

[0066] S1. Acquire the actual motor torque, motor speed or wheel speed, mechanical braking information, longitudinal speed, longitudinal acceleration, vehicle structure and transmission parameters, road conditions, articulation status, main-trailer connection status, trailer identification information, and communication time information of the tractor side and trailer side; establish a unified estimated time, perform time delay compensation or interpolation on continuously changing signals, perform timestamp matching or maintain the most recent valid state for discrete state information, and determine the time-series reliable state and time-series valid state of each information channel based on data age, signal change degree during time delay, time delay fluctuation, data integrity, and communication status.

[0067] S2. For each electric drive shaft on the tractor side and trailer side, reconstruct the longitudinal force of each electric drive shaft based on the actual motor torque with direction symbols, mechanical braking torque, power flow direction, transmission parameters, equivalent rotational inertia of the wheel system, wheel angular acceleration, internal loss torque of the transmission system, and effective rolling radius. Convert the mechanical braking force of the non-electric drive braking shaft into an additional longitudinal force and synthesize the longitudinal resultant force on the tractor side and trailer side respectively. Determine the reconstruction uncertainty, credible state, and effective state based on the reconstruction input uncertainty, mechanical braking information integrity, power flow direction, wheel slippage, and timing state.

[0068] S3. Transform the longitudinal resultant force, longitudinal acceleration, and driving resistance on both sides to a unified road longitudinal coordinate system, and add the longitudinal dynamic equations of the tractor side and trailer side to eliminate the longitudinal force at the articulation point; using the external road slope information as the slope reference, and based on the longitudinal excitation, input confidence state, candidate parameter normalization sensitivity, and parameter correlation, select a subset of active parameters that meet the identifiable conditions from the trailer equivalent mass, slope correction amount, and driving resistance parameters for updating, and freeze or restrict updates for parameters that do not meet the identifiable conditions.

[0069] S4. Establish a unified road longitudinal coordinate axis with the road tangential direction at the articulation point location. Define the direction of the pulling force generated by the tractor on the trailer as the positive direction of the longitudinal force at the articulation point. Construct pseudo-measurements for the longitudinal force at the same articulation point based on the longitudinal dynamics of the tractor side and the trailer side respectively. Determine the uncertainty, basic credible state, and effective state of the pseudo-measurements on both sides based on the joint uncertainty of the longitudinal force, mass, acceleration, road slope, driving resistance, and time sequence information used on both sides respectively, and retain the correlation of errors on both sides caused by the common input.

[0070] S5. Based on the previous valid estimate and the current vehicle dynamics change, obtain the predicted value of the longitudinal force at the articulation point. Calculate the innovation of the pseudo-measurements on both sides relative to the predicted value, construct the common mode residual, the differential mode residual, and the corresponding normalized statistics. Determine the innovation contribution coefficient of each side based on the basic innovation variance and independent health information on both sides, form the adjustment direction of process uncertainty and measurement uncertainty, and perform prediction-innovation correction using the adjusted process uncertainty, measurement uncertainty, or fusion weight to obtain the corrected estimate of the longitudinal force at the articulation point and its estimated uncertainty.

[0071] S6. Establish joint allowable sets for both the tractor side and the trailer side, including dynamic inputs, vehicle parameters, input correlations, and effective states. Form physically feasible sets on both sides through corresponding dynamic mapping and determine the common intersection. When the common intersection is not empty, apply common physical feasible set constraints to the corrected estimates. When the common intersection is empty, do not perform forced projection, and determine the degree of physical conflict based on the distance between the two sides' physical feasible sets. By performing bounded relaxation on the joint allowable subsets corresponding to different error sources, and combining independent health information and conflict persistence states, determine candidate anomaly sources and their contribution order.

[0072] S7. Based on changes in the trailer's unique identification information or the reconnection event after the main trailer is disconnected, a replacement candidate is formed. When the unique identification information is missing, the replacement is confirmed by combining the reconnection event with the trailer structure, axle configuration, drive system configuration, or wheel system parameter fingerprint. After confirming that a replacement has occurred, only the trailer-related parameters and their uncertainties, the trailer-related longitudinal force reconstruction and parameter recursion history, the bilateral innovation statistical status, the physical feasible set, and the error attribution history are re-initialized. The still valid tractor parameters and the current road status are retained, and the replacement recovery state is exited after the recovery conditions are continuously met.

[0073] S8. Based on the temporal states of both sides, longitudinal force reconstruction state, pseudo-measurement innovation, independent health information, common physical feasible set state, candidate anomaly sources, and common input health state, the health state of the estimated channels on both sides is determined by continuous periodic confirmation and hysteresis recovery. Based on the health state of each side, estimation variance, output confidence and short-term holding time, the two-sided fusion estimation, single-sided health estimation, short-term prediction holding or failure mode is selected, and the longitudinal force estimate or transient estimate of the articulation point, estimation uncertainty, output confidence state, output effective state, current estimation mode and candidate anomaly source category are output.

[0074] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0075] 1. This invention synchronizes and hierarchically uses multi-source information by unifying the estimated time, delay compensation, and delay-dynamic coupling mismatch, so that each dynamic input corresponds to a unified time base, which can reduce the synchronization error caused by judging the validity of information based solely on fixed communication delay or a single data age.

[0076] 2. This invention reconstructs the longitudinal force of the main trailer based on motor torque, mechanical braking torque, power flow direction, wheel system inertia, transmission loss, effective rolling radius, wheel slippage and timing state, and incorporates the mechanical braking force of the non-electric drive brake axle into the resultant longitudinal force on both sides, which can reduce the impact of compound braking, rapid wheel speed changes and incomplete mechanical braking information on the reliability of dynamic input.

[0077] 3. This invention eliminates the internal forces at the hinge point by adding the longitudinal dynamic equations on both sides, and selects a subset of active parameters based on normalized sensitivity, information matrix conditionality, input confidence, and parameter correlation, so that the parameter update matches the current longitudinal excitation. This can reduce the situation where trailer mass, slope correction, and driving resistance parameters compensate for each other or converge incorrectly when the excitation is insufficient.

[0078] 4. This invention constructs redundant pseudo-measurements of the longitudinal force at the same hinge point based on the longitudinal dynamic relationship between the tractor side and the trailer side, and adjusts the uncertainty by using common mode residuals, differential mode residuals, and innovative contributions from both sides. This can characterize the common offset and inconsistency between the two sides respectively, reducing the possibility of misjudging real dynamic changes as unilateral anomalies or incorrectly attributing common input anomalies to a certain side channel.

[0079] 5. This invention constructs a joint permissible set from the dynamic inputs, vehicle parameters, correlations, and effective states of each side, and performs physical self-verification of the estimation results through the common intersection of the two-sided physically feasible sets. When the common intersection is empty, forced projection is not performed, but candidate anomaly sources are determined based on the set distance and the minimum bounded relaxation result. This can avoid outputting estimation results that do not meet the dynamic constraints of the main trailer and provides a basis for subsequent parameter freezing, resynchronization, and channel degradation.

[0080] 6. After trailer replacement, this invention only reinitializes the trailer-related parameters, covariance, and historical status, while retaining the still valid tractor parameters and road conditions. At the same time, it implements graded degradation and hysteresis recovery based on the estimated channel health status, physical consistency, output reliability, estimation variance, and holding time on both sides. This can improve the continuity, reliability, and interpretability of the estimated output under conditions of trailer replacement, unilateral anomalies, short-term anomalies of shared inputs, and no effective pseudo-measurements. Attached Figure Description

[0081] The present invention will be further described below with reference to the accompanying drawings:

[0082] Figure 1 This is a schematic diagram showing the composition of the self-verification estimation system for the longitudinal force at the main hinge point of the present invention and the information interaction relationship between each module.

[0083] Figure 2 This is a flowchart of the overall process for the self-verification estimation method of longitudinal force at the main hinge point described in this invention.

[0084] Figure 3 This is a schematic diagram of three typical states of self-verification of the two-sided physical feasible region described in this invention, wherein, Figure 3 (a) is a state in which the physical feasible set on the tractor side and the physical feasible set on the trailer side intersect and the corrected estimate is located within the common physical feasible set; Figure 3 (b) is a state in which the two physical feasible sets intersect and the corrected estimate is outside the common physical feasible set; Figure 3 (c) represents a state where the physical feasible set on the tractor side and the physical feasible set on the trailer side do not intersect. Detailed Implementation

[0085] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. These embodiments are used to illustrate the technical concept, data processing relationships, and collaborative processes of each module, and are not intended to limit the present invention to a specific vehicle model, a specific number of axles, a specific communication protocol, or a specific state observation algorithm. Without changing the bilateral dynamic redundancy estimation, common difference mode consistency evaluation, and bilateral physical feasible region self-verification relationships, each module can be centrally located in the tractor unit's overall controller, or it can be separately located in the tractor unit controller and the trailer controller, exchanging information through the onboard communication network.

[0086] This embodiment uses a main-trailer vehicle consisting of a three-axle electric tractor and a three-axle electric drive semi-trailer as the subject of this description. The tractor includes one steering axle and two rear axles, one or both of which are configured as electric drive axles; at least one of the three axles of the trailer is configured as an electric drive axle; each electric drive axle is equipped with a drive motor, and each wheel is equipped with a mechanical braking device; the rear axles and trailer axles use dual tires. For two-axle tractors, four-axle tractors, and vehicles with different numbers of trailer drive axles, it is only necessary to establish a corresponding axle index set according to the actual drive axle and brake axle configuration, without changing the subsequent estimation relationship.

[0087] In this embodiment, the equivalent unloaded mass of the trailer is taken as 7810 kg, the loaded mass as 34400 kg, and the total mass of the trailer when fully loaded is 42210 kg. The equivalent mass of the tractor unit is the weight of the entire vehicle, including the power battery, cab, chassis, and onboard equipment, or the effective value after calibration by the whole vehicle model. The mass of the vehicle body model, which does not include the mass of the power battery, is not directly used. The rated shaft end torque of the drive motor is 1600 N·m, and the peak shaft end torque is 2400 N·m. The rated voltage of the power battery is 673 V, and the rated capacity is 660 A·h. The voltage and capacity of the power battery are used to determine the usable state and power limit of the electric drive system and are not directly substituted into the longitudinal force dynamics equation of the articulation point.

[0088] The effective rolling radius of the tractor tires is set at 0.538m, and the effective rolling radius of the trailer tires is set at 0.510m. The rolling resistance coefficient for normal roads is calibrated within the range of 0.004 to 0.009. Based on the current vehicle braking system parameters, the maximum mechanical braking force at the wheel end of a single brake wheel is set to 9.0 × 10⁻⁶ m. 3 N is used as the physical upper limit of the braking torque; the corresponding upper limit of the longitudinal braking force of the tire is determined by dividing the mechanical braking torque at the wheel end by the effective rolling radius of the wheel. The actual estimated cycle uses the actual mechanical braking torque obtained from the output of the brake controller or the pressure-braking torque calibration relationship, and does not replace the actual action with the maximum mechanical braking torque.

[0089] The tractor side can provide actual motor torque, motor speed or wheel speed, mechanical braking status, longitudinal speed, longitudinal acceleration, vehicle attitude, and source timestamp; the trailer side can provide the same type of trailer drive and motion information, and provide trailer identity, axle configuration, drive system configuration, and tractor-trailer connection status. The tractor controller and trailer controller communicate via CAN, vehicle Ethernet, or other vehicle communication networks with timestamp transmission capabilities; the road gradient reference can be provided by at least one of the following: inertial measurement unit, satellite positioning unit, electronic map, road preview device, or existing gradient estimator.

[0090] This embodiment establishes a unified longitudinal coordinate axis for the road along the tangential direction of the road at the hinge point, with the direction of vehicle travel along the road as the positive direction of the coordinate axis. The longitudinal force at the hinge point is denoted as... It represents the component of the force exerted by the main and trailer connecting components on the longitudinal coordinate axis of the unified road; when the tractor pulls the trailer... When the trailer pushes the tractor, According to the action-reaction relationship, the longitudinal force at the articulation point is equal in magnitude and opposite in direction in the dynamic equations for the tractor side and the trailer side.

[0091] Throughout this text, mass is expressed in kg, longitudinal force in N, torque in N·m, length and effective rolling radius in m, longitudinal velocity in m / s, and longitudinal acceleration in m / s². 2 Angular velocity is expressed in rad / s, and angular acceleration is expressed in rad / s². 2 Time is expressed in seconds (s), road slope is uniformly expressed as slope angle in rad, and force variance is expressed in N. 2 The temporal mismatch, slip index, identifiable stimulus index, common mode and differential mode statistics, health confidence and output confidence are all dimensionless; parameters of different units are normalized according to the corresponding reference scale before entering the information matrix or distance evaluation.

[0092] After the system powers on, it reads the number of axles, drive axle positions, transmission ratios, effective rolling radius, equivalent moment of inertia of the wheel system, prior values ​​of vehicle mass and their uncertainties from the vehicle configuration file. A timestamp cache covering at least 3 seconds of data is established, and the initial value of the longitudinal force at the articulation point is set to zero or determined based on the slope and resistance balance relationship under static conditions. The initial estimation variance uses a value no less than... The conservative value is used; before the timing synchronization, longitudinal force reconstruction and at least one side pseudo measurement have reached the effective state, the output effective state is kept as invalid to avoid outputting the initialization process as a normal estimation result.

[0093] like Figure 1As shown, the main information chain from left to right consists of the integrated information acquisition and timing synchronization module, the axle longitudinal force reconstruction module, the main-trailer dual-side dynamic redundancy estimation module, the common difference mode consistency adaptive observation module, the dual-side physical feasible region self-verification and error attribution module, and the fault confirmation and graded degradation output module. The integrated information acquisition and timing synchronization module can set up front-end interfaces on the tractor side and the trailer side respectively, and after completing local signal acquisition, it sends the data with source timestamps to the unified time base processing unit.

[0094] Within each estimation cycle, the axle longitudinal force reconstruction module receives synchronized motor torque, mechanical braking torque, wheel speed, and transmission parameters, and outputs the longitudinal resultant force on the tractor side and trailer side, reconstruction uncertainty, and effective state. The main-trailer dual-side dynamic redundancy estimation module constructs two pseudo-measurements of the longitudinal force at the same hinge point by combining the mass, acceleration, gradient, and driving resistance on both sides. The common-mode consistency adaptive observation module adjusts the process uncertainty, measurement uncertainty, or fusion weight based on the offset relationship between the two pseudo-measurements and the predicted value. The dual-side physical feasible region self-verification and error attribution module performs physical constraints or conflict attribution on the corrected estimate.

[0095] The slow time-varying parameter selective identification module receives the time-series reliable state, longitudinal force reconstruction results, and the vehicle's longitudinal motion state. It classifies the trailer's equivalent mass, road slope correction, and driving resistance parameters as candidate parameters and updates only the subset of active parameters that currently meet the identification criteria. The update frequency of this module can be lower than the longitudinal force estimation main cycle frequency of 0.010s, for example, performing parameter updates once every 5 to 10 main cycles; cycles without parameter updates continue to use the previous valid parameter value and its uncertainty as it grows over time.

[0096] The trailer replacement identification and selective re-initialization module continuously monitors the trailer's identity, master-trailer connection status, and structural or drive fingerprints. When a trailer replacement is confirmed, this module only re-initializes the trailer's relevant parameters, trailer-side recursive history, bilateral innovation statistical status, and physical conflict history, and then... Figure 1 The system includes a selective reinitialization control path notification module for slow time-varying parameters, a common-mode consistency adaptive observation module, and a two-sided physical feasible domain self-verification and error attribution module. The tractor parameters and the still valid road conditions are not cleared with the trailer swapping event.

[0097] like Figure 2As shown, the system executes information acquisition and timing synchronization, longitudinal force reconstruction of the main trailer, selective identification of slow time-varying parameters, construction of bilateral pseudo-measurements, adaptive observation of common differential modes, self-verification of bilateral physical feasible regions, trailer replacement identification and selective re-initialization, and health confirmation and graded degradation output in sequence from S1 to S8. If trailer replacement is not confirmed, the result of S6 directly proceeds to S8; if trailer replacement is confirmed, S7 is executed first and the system enters the trailer replacement recovery state, and then S8 determines the current output estimation mode.

[0098] In the During one estimation cycle, the system obtains the actual motor torque on the tractor side and trailer side. The parameters include motor speed or wheel speed, mechanical braking torque or braking pressure, longitudinal speed, longitudinal acceleration, vehicle attitude, articulation angle, road gradient, main-trailer connection status, trailer identification information, and corresponding source timestamps, frame numbers, reception timestamps, and communication status. For the same physical quantity with multiple sources, the independent timing status and uncertainty of each source are preserved, and the average value is not unconditionally taken at the input.

[0099] The tractor controller establishes a uniform estimation time in each estimation cycle using a local monotonic clock. For situations where the trailer-side clock and the tractor-side clock are out of sync, the clock deviation is estimated using the transmission time, reception time, and periodic clock correction information in the master-trailer communication message; the effective sampling time after deviation correction is recorded as... The data age relative to the unified estimation time is When messages are out of order but the source timestamp is valid, they are reordered according to the source timestamp. When messages are duplicated, only the data with the latest sequence number and valid verification is retained.

[0100] For continuous signals such as motor torque, wheel speed, longitudinal velocity, longitudinal acceleration, and hinge angle, a low-pass filter, a sliding polynomial fitter, or a state filter with equivalent function is first used to obtain filtered values ​​and rates of change. Then, first-order extrapolation, interpolation, or model prediction compensation is performed based on the data age. The compensation value is limited to the physical boundaries of the signal and within a preset maximum rate of change. When the extrapolation reaches the limit, the signal is marked as time-constrained, rather than continuing to increase the extrapolation amplitude.

[0101] For discrete information such as gear position, drive enable, energy recovery enable, mechanical braking engagement, tractor-trailer connection status, and trailer identity, timestamp matching or the most recent valid state is used. If the unified estimation time spans the discrete state switching time, the corresponding data before and after the switch are used separately, and the continuous change trend before the switch is not allowed to be extrapolated to the state after the switch. For example, when mechanical braking switches from non-engaged to engaged, the zero braking torque before the switch must not continue to be used as a valid input after the switch across the switching time.

[0102]

[0103] In the formula, For the first The latency reference scale for information channels The filtered rate of change of the corresponding signal. As a reference scale for amplitude, This is a time-delay-dynamic coupling mismatch that simultaneously characterizes the degree of signal change during both the data age and the delay period. This implementation will use rapidly changing signals... Taking 0.010s, the information used for slow time-varying trend judgment will be... Take 0.60s.

[0104] This implementation method uses the time-normal threshold. Set to 1.0, timing failure threshold Take 6.0, recovery threshold Take 0.50. When Normally, this channel is used; when The time stamp is marked as time-constrained, and the corresponding input variance is increased or the weight is decreased according to the amount of mismatch; when The time stamp is considered invalid. Data with an age exceeding 0.05s, data verification failure, or unrecoverable out-of-order data is directly judged as invalid.

[0105] A time-series anomaly is confirmed as a persistent anomaly after four consecutive estimation periods, only when... Furthermore, the communication status remains normal for eight consecutive estimation cycles before returning to normal. Time-constrained information can be used for state prediction or expanding the physically feasible set, but is prohibited for parameter updates requiring high-precision correspondence at the same time. Time-invalid information does not participate in slow time-varying parameter updates or corresponding side pseudo-measurement corrections. This tiered usage method ensures that time anomalies are primarily manifested as increased uncertainty, preventing estimation mode switching from being directly caused by a single delayed message.

[0106] The system establishes an index set for the electric drive shaft on the tractor side according to the vehicle configuration file. Trailer-side electric drive shaft index set and the union of the two The axle index corresponds to the axle, not a single tire. When a drive axle is driven by a single motor for both wheels, the actual torque of the motor is first converted to the wheel end of the axle through the transmission system, and then combined with the mechanical braking action of the left and right wheels of the axle to form the longitudinal force of the axle. The normal load and braking capacity of dual tires are included in the parameters and physical boundaries of the axle, and the driving force term of the same motor is not repeatedly established.

[0107] The torque of the motor that generates the wheel-end force in the direction of vehicle travel is defined as positive, and the torque transmission coefficient under driving conditions is... Take the positive value corresponding to the positive power transmission. In the energy recovery state, the motor torque is input according to the actual sign of the controller, and the mechanical braking torque... The sign is negative during braking. If the torque sign given by the motor controller only indicates the direction of motor rotation, the power flow direction is also determined based on the motor speed, DC bus power, and drive or regeneration enable status. After completing the unified sign conversion, the longitudinal force at the wheel end is reconstructed.

[0108]

[0109] In the formula, and These are the transmission ratio of the transmission stage and the transmission ratio of the main reducer, respectively. To calculate the equivalent moment of inertia at the wheel end, This is the filtered wheel angular acceleration. Torque loss within the transmission system Effective rolling radius. Torque transmission coefficient. The parameters are selected based on the driving, gliding, and energy recovery directions, and include the transmission efficiency in the corresponding directions. When the transmission efficiency is represented by an independent parameter, the coefficient can be split accordingly, but the efficiency loss must not be counted twice.

[0110] For non-electrically driven brake axles, a motor torque term is not constructed. Instead, the actual mechanical braking effect of each brake wheel is converted into an additional longitudinal force based on its effective rolling radius and included in the resultant longitudinal force on the vehicle side. When braking pressure is available but wheel-end torque cannot be directly obtained, the actual value is calculated through the pressure-braking torque curve calibrated by test, and the curve error is included in the braking force uncertainty.

[0111] With input vector The input is represented by the reconstructed input along the i-th axis. The first-order sensitivity of each input is used to form a Jacobian vector based on the longitudinal force reconstruction relationship. and in accordance with Calculate the basic reconstruction variance. For errors originating from the same torque sensor, the same time synchronization channel, or the same effective rolling radius calibration, in The cross covariance is retained, and its repetition is not considered as mutually independent errors.

[0112]

[0113] In the formula, Let be the vehicle speed at position i along the longitudinal coordinate axis of the unified road. Denominator protection amount This serves as a reference scale for the longitudinal force of the axis. This embodiment will... Take 1.0 × 10 5 N、 Take 0.10, The threshold values ​​are set to 0.20; the slip warning threshold is set to 0.10, the slip invalid threshold is set to 0.20, the normal reconstruction threshold is set to 1.5, and the reconstruction failure threshold is set to 5.0. At low speeds, the longitudinal speed of the vehicle body or the inertial navigation speed is used instead of direct slip calculation based on the wheel speed validity to prevent false high slip caused by an excessively small denominator.

[0114] The longitudinal resultant force on the side of the tractor is caused by The effective drive axle reconfiguration force, the additional longitudinal force of the non-drive brake axle on the tractor side, and the action of other known longitudinal actuators are added together; the longitudinal resultant force on the trailer side is obtained by the same method. Information on the non-driving brake axles of the trailer is obtained. The combined variance is determined by the variance of each axle and the covariance of the common source error; when some axles are in a constrained state, constrained resultant forces are still allowed to form, but the corresponding variance increases and the resultant force is prohibited from being used for updating high-degree-of-freedom slow time-varying parameters.

[0115] when and When the axis is in normal condition, it outputs a normal reconstructed value; when any indicator is between the warning threshold and the invalid threshold, it outputs a restricted reconstructed value and increases the variance. If the restricted state fails to recover after 20 consecutive estimation periods, it becomes invalid. or If the actual mechanical braking force is missing under combined braking conditions, it should not be treated as zero. Instead, the corresponding force range should be increased and the corresponding side pseudo-measurement should be stopped from participating in normal observation and correction.

[0116] The driving force, longitudinal acceleration, and driving resistance on both the tractor side and the trailer side are converted to a unified road longitudinal coordinate axis. The driving resistance on both sides is defined as positive when it hinders the vehicle's movement in the positive longitudinal direction of the road, and negative when it provides a boost due to factors such as the gravity component during downhill driving. In one embodiment, the first... Side drag consists of rolling drag, air drag, slope drag, and unmodeled drag correction, with the weight of air drag parameters in identification reduced at low speeds.

[0117]

[0118] The above equation is obtained by adding the longitudinal dynamic equations of the tractor side and the trailer side. The longitudinal forces at the hinge point in the equations on both sides are eliminated because they are equal in magnitude and opposite in direction. and The equivalent masses on both sides are respectively. and These are the longitudinal accelerations on both sides, and These are the resultant forces in the longitudinal direction on both sides. and These are the equivalent driving resistances other than the longitudinal resultant force and the hinge point force. This vehicle relationship is used to identify slow time-varying parameters and is not directly used as the final hinge point longitudinal force output.

[0119] The road slope adopts the external slope benchmark. With bounded correction The sum is represented as the external slope reference. The external slope reference is primarily determined by fused information capable of distinguishing between road slope and vehicle longitudinal acceleration; when using only a single longitudinal acceleration sensor, the slope correction is not freely estimated simultaneously during periods of strong vehicle acceleration or forced movement. The absolute value of the slope correction is limited to the range of 0.0035–0.0087 rad to prevent the slope estimation from absorbing trailer mass errors or drive force reconstruction errors.

[0120] The candidate parameter vector includes trailer equivalent mass, slope correction, rolling resistance coefficient, air resistance coefficient, and unmodeled resistance correction. The system first eliminates parameters lacking sensitivity based on vehicle speed range, longitudinal acceleration, driving force variation, and road slope variation. Then, it forms a subset of active parameters based on the normalized correlation between the candidate parameters. For example, when driving at low speeds and with significant changes in driving force, the equivalent mass and rolling resistance parameters of the trailer should be prioritized. When driving at high speeds and with stable conditions, the air resistance parameter can be selected. The slope correction should only be included in the subset of active parameters when the slope reference is reliable and the slope changes slowly.

[0121]

[0122]

[0123] In the formula, This is the sensitivity vector of the vehicle's longitudinal dynamics to a subset of active parameters. The parameter reference scale is a diagonal matrix. As time window weight, The number of samples contained in the sliding time window. The input comprehensive reliability factor is determined based on driving force reconstruction, timing synchronization, and longitudinal motion state. This is a denominator protection quantity. After scale normalization, both the eigenvalue ratio of the information matrix and the identifiable excitation index are dimensionless quantities.

[0124] This implementation uses a 2-second sliding time window, setting the lower threshold for excitation to 0.03, the upper threshold for excitation to 0.10, the lower limit of the input comprehensive confidence factor to 0.70, and the upper limit of the parameter correlation to 0.95. Normal evaluation is allowed when the number of information matrix conditions is no higher than 30; restricted updates are allowed when the number of conditions is between 30 and 200; and the subset of active parameters is reduced or updates are frozen when the number exceeds 200. Freeze activity parameters, when When performing restricted updates, Normal updates will be performed if all other conditions are met.

[0125] Parameter updates can employ recursive least squares with a forgetting factor, projected gradient, moving window estimation, or other recursive algorithms with bounded constraints. Regardless of the algorithm used, the updated parameters are projected onto the corresponding physical boundaries, and the single-cycle change is limited. In this implementation, the single-cycle change of the trailer's equivalent mass is limited to 0.10% of the current value, and the single-cycle change of the slope correction is limited to 2 × 10⁻⁶. -5 rad limits the single-cycle variation of the rolling resistance coefficient to 2 × 10⁻⁶. -5 If the correlation between candidate parameters exceeds 0.95, parameters with lower sensitivity energy or higher prior confidence are removed first, and then the identifiable excitation index is recalculated.

[0126] The permissible range for the equivalent mass of the trailer is 7810–42210 kg, and the permissible range for the rolling resistance coefficient is 0.004–0.009. The slow time-varying parameter selective identification module outputs the current valid parameter value, parameter covariance, active parameter subset, identifiable excitation index, and parameter update valid state, and sends these results to the dual-sided pseudo-measurement construction and physical feasible region module. In the unidentifiable state, it continues to output the previous valid parameter value, but its uncertainty increases with the duration of the freeze, and the frozen value is not marked as a new identification result.

[0127] The dual-side dynamic redundancy estimation module again transforms the driving forces, accelerations, and driving resistances on both sides to a unified road longitudinal coordinate axis at the articulation point to eliminate the projection difference between the vehicle's longitudinal axis and the road tangential direction. The positive direction of the longitudinal force at the articulation point is still defined as the direction in which the tractor pulls the trailer; therefore, in the tractor-side dynamics, the positive articulation point longitudinal force manifests as a rearward load effect of the tractor, while in the trailer-side dynamics, it manifests as a forward traction effect of the trailer. After reorganizing the dynamic equations on both sides, the following equations are obtained for the same... Two pseudo-measurements.

[0128]

[0129] In the formula, and The longitudinal force pseudo-measurements at the articulation points on the tractor side and trailer side are respectively measured in N; and For the current effective equivalent quality, add the superscript. The force and acceleration are both components on the longitudinal coordinate axis of the unified road. If the tractor mass uses a fixed calibration value, its configuration uncertainty is still retained as a non-zero variance; the trailer mass uses the current effective value and covariance output by S3.

[0130]

[0131] In the formula, For the first The sensitivity of the spurious measurement relationship to the input quantity. The joint covariance of the driving force, mass, acceleration, gradient, driving resistance, and timing compensation on this side. This represents the cross-covariance of the pseudo-measurements on both sides. In this implementation, the upper limit of the allowable standard deviation of the pseudo-measurements on the tractor side is set to 2.0 × 10⁻⁶. 4 N, the upper limit of the permissible standard deviation of the pseudo-measurement on the trailer side is taken as 3.0 × 10. 4 N; A pseudo-measurement is marked as valid only when the key input is valid and the pseudo-measurement variance does not exceed the corresponding upper limit. Related terms arising from road gradient, air density, common vehicle speed or acceleration, and uniform time base error are retained in the cross-covariance.

[0132]

[0133] In the formula, The fundamental reliability factor on the q-th side is determined jointly by temporal sequence, longitudinal force reconstruction, motion state, and parameter state. This is the denominator protection quantity with the dimension of force variance. When both sides are valid, When only one side is valid, the base weight of the valid side is reset to 1, and the other side is set to 0; when both sides are invalid, no observation correction is generated. Decreased reliability is reflected by increasing the measurement variance or decreasing the fusion weights, not by directly reducing them. or The force value.

[0134]

[0135] In the formula, and These are the output value and estimated variance of the previous effective estimation period, respectively. This indicates the changes in lateral driving force, longitudinal acceleration, and driving resistance relative to the previous cycle. This is the mapping coefficient from the corresponding dynamic change to the longitudinal force change at the hinge point. This represents process uncertainty. If reliable dynamic change information is lacking, the predicted increment is set to zero and increased. Instead of using ineffective dynamic changes to extrapolate the longitudinal force at the hinge point.

[0136]

[0137] Common mode residual The differential residual characterizes the degree of common offset between the two spurious measurements and the predicted values. It characterizes the degree of inconsistency between the two pseudo-measurements. Since the common-mode residual can also be generated by rapid changes in the longitudinal force at the real hinge point, it is only permissible to use the common-mode anomaly primarily to increase process uncertainty when the common input is healthy and the changes in driving force or longitudinal acceleration can corroborate this. The differential-mode residual only reflects the inconsistency between the two sides and is not equivalent to a fault in a sensor or actuator on one side alone.

[0138]

[0139] In the formula, The measurement covariance matrix includes the pseudo-measurement variances on both sides and the cross covariance. , This is the protection quantity in the denominator of the force variance. If only one side's pseudo-measurement is valid, the innovation on that side is retained for single-sided observation correction, but the single-sided information is not used to calculate the differential mode statistic; if the common input is invalid, the common mode statistic can still be used as a consistency state record, but it cannot be used to determine the true dynamic change on its own.

[0140]

[0141] In this implementation, the abnormal entry thresholds for both common mode and differential mode statistics are set to 6.63, and the recovery thresholds are set to 3.84. The upper limit of the absolute change in the single-cycle fusion weight is set to 0.05, and the upper limit of the single-cycle amplification factor for process uncertainty and measurement uncertainty is set to 1.3. For the first cycle that has not undergone adaptive adjustment Basic innovation variance It is used to indicate the relative innovative contributions of each side in differential mode inconsistency, but only after combining independent timing, longitudinal force reconstruction and motion state health information can it be used to determine which side to increase the measurement uncertainty.

[0142] when and When both do not exceed the entry threshold, maintain the basic process uncertainty and measurement uncertainty; when only When the entry threshold is exceeded, under the condition that the shared input is healthy and the dynamic change is supported, proceed as follows: Increase process uncertainty; otherwise, prioritize increasing the uncertainty related to common inputs or suspend updates to related parameters. When the entry threshold is exceeded, according to Increase the measurement uncertainty of the side that contributes more to independent health information or reduce its fusion weight; when both exceed the entry threshold, suspend the update of relevant slow time-varying parameters and send the results of this period into the physical feasible region for verification, and do not directly determine the fault side in the statistical layer.

[0143]

[0144]

[0145] In the formula, When a pseudo-measurement on one side is invalid, its corresponding observation row and covariance row and column are deleted. When both sides are invalid, only state prediction is performed and the estimated variance is increased. Common mode anomalies or differential mode anomalies are confirmed as persistent anomalies after exceeding the entry threshold for four consecutive estimation periods, and are deactivated after not exceeding the recovery threshold for 12 consecutive estimation periods. Restrictions are applied during both confirmation and recovery periods. , The correction estimate is obtained by considering the single-cycle change of the fusion weights. and its variance .

[0146] Establish joint allowable sets for both the tractor side and the trailer side. The set consists of the physical upper and lower bounds and confidence intervals of longitudinal resultant force, equivalent mass, longitudinal acceleration, road slope, driving resistance, articulation state and timing compensation; inputs with common source errors are represented by relevant constraints or joint covariance; inputs that have been determined to be invalid are not used as deterministic point values, but are expanded into restricted intervals or the corresponding side of the physical feasible set is temporarily invalidated according to the degree of recoverability.

[0147]

[0148] In the formula, Indicates the first Side-joint allows for a set of dynamic inputs and vehicle parameters in the set. This is the pseudo-measurement mapping of the corresponding side dynamics in S4. and The sets of physically feasible longitudinal forces at the articulation points on the tractor side and trailer side are respectively. This represents the set of physically feasible common points on both sides. In one embodiment, interval operations, vertex enumeration, or constrained sampling are used. Perform mapping; if the mapping result is multiple discontinuous intervals, retain the interval union and do not replace it with its smallest envelope interval, so as to avoid artificially introducing feasible force values ​​that do not exist.

[0149]

[0150] In the formula, This means identifying the feasible value closest to the input value within the common physically feasible set. This is a physically self-verified estimate. This represents the minimum distance between the physically feasible sets on both sides. In this implementation, the upper limit of the longitudinal acceleration under the driving condition is set to 0.81 m / s². 2 The lower limit for normal braking conditions is taken as -1.04 m / s. 2 The lower limit for emergency braking is set at -4.5 m / s.2 And set the longitudinal force reference scale at the hinge point. Take 2.5 × 10 5 N.

[0151] Figure 3 In the diagram, the lateral position indicates the relative magnitude of the longitudinal force at the hinge point. , and These represent the tractor-side, trailer-side, and common physical feasible sets, respectively; the hollow circle represents the corrected estimate obtained from S5. The solid circle represents the self-verified estimate after physical constraints. . Figure 3 The continuous interval form is used to illustrate three typical relationships, but the physical feasible set in actual operations is not limited to continuous intervals.

[0152] like Figure 3 As shown in (a), when and When the corrected estimate satisfies the dynamic allowable conditions on both sides, the projection result is the same as the input value, that is... At this point, no additional reduction in the estimated variance is made; only the physically consistent state is recorded, and the common set boundary is used to monitor whether subsequent estimates approach the physical limits.

[0153] like Figure 3 As shown in (b), when but At that time, the feasible value closest to the corrected estimate is selected from the common physical feasible set. When multiple candidate points with the same projection distance exist in multiple discontinuous intervals, priority is given to selecting candidate points that are continuous with the previous valid estimate and consistent with the current driving force change direction; the projection distance is also used as a record of the physical constraint intensity in this cycle, but a single projection does not directly trigger fault determination.

[0154] like Figure 3 As shown in (c), when At this time, the dynamic conditions on both sides conflict with each other. Therefore, we should not... Forced projection to or Instead of discarding either side of the set, it retains it as a transient estimate, increasing the estimation variance and decreasing the output confidence; simultaneously, it calculates the set distance. This is to avoid masking common input anomalies or multi-source anomalies by forcibly selecting one side before the source of the anomaly has been reliably determined.

[0155] In this implementation, the normalized conflict warning threshold is set to 0.08, corresponding to an absolute set distance of 2.0 × 10⁻⁶. 4 N; The severe conflict threshold is set to 0.20, corresponding to an absolute set distance of 5.0 × 10⁻⁶.4 N. When Maintain transient estimates and continue observation; when The time is marked as a restricted conflict and the adjustment range of the fusion weight is limited; when Furthermore, candidate anomaly attribution and downgrade processing are initiated during four consecutive estimation cycles.

[0156]

[0157] In the formula, The errors include six categories: longitudinal force reconstruction, vehicle parameters, longitudinal motion state, road and driving resistance, timing synchronization, and articulation state. Indicates only for the first The class allows subsets to be implemented with the following scope: The first obtained after bounded relaxation Side-physical feasible set, The candidate category with the smallest relaxation required to restore the common intersection. This implementation takes the maximum bounded relaxation amount. That is, 6.25 × 10 4 If the intersection cannot be restored even after exceeding the upper limit N, the single-sided range will not be expanded further, and the process will switch to handling common input anomalies or multi-source anomalies.

[0158] Candidate anomaly sources are jointly determined by the minimum relaxed category, relaxed range, temporal health status of the corresponding channel, longitudinal force reconstruction health status, motion state health status, parameter identifiable status, and conflict duration period. Only when a candidate source continuously meets the confirmation conditions will the corresponding channel undergo bounded weight reduction, parameter update freezing, resynchronization, restricted reconstruction, or estimation mode switching. Upper limits are set for adjustments to weights, variance, and allowable set boundaries within a single period to prevent positive feedback between error attribution results and subsequent weight adjustments. The restricted conflict state is lifted after the common intersection is restored and the relevant conflict amount does not exceed half of the warning threshold for eight consecutive estimation periods.

[0159] S6 outputs the physical self-verified estimate or transient estimate, the corresponding estimate variance, the two sides of the physical feasible set, the common intersection state, the set distance, the physical consistency state, and the candidate anomaly sources and their contribution order. The candidate anomaly sources only represent the possible error categories obtained based on model consistency and are not equivalent to a deterministic hardware fault diagnosis of a specific sensor, actuator, or connecting component.

[0160] Trailer replacement identification is performed in the following order: trailer unique identification information, the event of reconnection after the main trailer disconnects, structural or drive fingerprints, and auxiliary features such as mass or drag. When the trailer's unique identification information is available, the change in identification and the connection status of the main trailer together form a candidate for trailer replacement. When the unique identification information is unavailable, a reconnection after the main trailer disconnects must be detected, along with at least one change in structural or drive fingerprint among the following: number of axles, drive axle position, number of motors, transmission parameters, effective rolling radius, or braking configuration. Changes in trailer mass or driving drag alone are insufficient to confirm trailer replacement, to avoid misidentifying the reloading of the same trailer as a trailer replacement.

[0161]

[0162] In the formula, For the normalized j-th trailer structure or drive fingerprint, For the corresponding feature weights, The value ranges from 0 to 1. In this implementation, the fingerprint difference threshold is set to 0.15; when unique identification information is missing, only when the reconnection event is established and... Only then is a replacement candidate formed. The replacement is confirmed after the replacement candidate meets the confirmation conditions for 30 consecutive estimation cycles; if the identity is restored to the original trailer, the connection status is broken again, or the fingerprint difference no longer meets the threshold before confirmation, the candidate status is cancelled.

[0163]

[0164] In the formula, and The parameters and covariance of the tractor unit that remain valid before the swapping are as follows: and For the prior values ​​of the new trailer parameters and the initial covariance, and The current road parameters and covariance are still valid. The prior parameters of the new trailer can be obtained from the trailer identity database, trailer controller self-description messages, or vehicle model default parameters; when the source is uncertain, a wider initial covariance is used.

[0165] After confirming the trailer swap, retain the tractor-side transmission parameters, tractor-side mass parameters, tractor-side longitudinal force reconstruction calibration, and the still valid road gradient benchmark. Clear or reinitialize the original trailer's longitudinal force reconstruction history, parameter recursion history, bilateral innovation statistics, differential mode confirmation count, trailer-side measurement uncertainty adaptive state, bilateral physical feasible sets, physical conflict count, and error attribution history. The tractor-side cache, which is related to the unified time base but does not depend on the specific trailer identity, is retained, and the trailer-side cache is re-established from the stable moment.

[0166] When both sides of the new trailer's pseudo-measurements are valid, the initial value of the longitudinal force observation at the articulation point is set based on their basic fusion results. When only the pseudo-measurement on the tractor side is valid, the initial value is set jointly by the pseudo-measurement on the tractor side and the current power change prediction. When neither side's pseudo-measurements are valid, the short-term prediction initial value is formed by the state of the previous valid tractor and the current driving force change. The initial observation variance is not lower than the variance after mapping the prior mass and resistance uncertainties of the new trailer to the longitudinal force at the articulation point, to avoid outputting excessively high confidence immediately after the trailer swap.

[0167] After the trailer swap, the system enters recovery mode. The single-cycle update range of trailer-related parameters is set to half of the normal upper limit, and the basic variance of the trailer-side pseudo-measurement is initialized at 1.3 times. This implementation requires an identifiable excitation index of not less than 0.10, a differential modulus statistic of not more than 3.84, and a non-empty common physical feasible set. The condition is maintained continuously for 5 seconds before exiting the trailer swapping recovery mode; if any condition is interrupted, the duration is re-accumulated. During the recovery period, the process of the new trailer re-converging is paused as the basis for confirming the continuous fault, but obvious timing invalidity, drive force reconstruction invalidity, or connection interruption status are not masked.

[0168] Establish channel health credibility for both the tractor side and the trailer side. and In one embodiment, a weighted geometric average is performed on the temporal reliability factor, longitudinal force reconstruction reliability factor, motion state reliability factor, parameter validity factor, pseudo-measurement innovation reliability factor, and physical consistency factor, with an upper limit set on the result by any key input invalid state. The weights are determined based on the sensitivity of each type of input to the corresponding pseudo-measurement; the health reliability is only allowed to vary within the range of 0 to 1, and the variation range of a single estimation period is limited.

[0169] When road gradient, common longitudinal velocity or acceleration, unified time base, or other information participating in pseudo-measurements on both sides are abnormal, a separate common input abnormality state is formed, and it is not directly counted as an independent fault on one side. The two sides' physical feasible sets do not intersect, nor do they automatically indicate that both sides are failed; only after combining the S6 candidate anomaly source, the independent health information on both sides, and the conflict persistence state can it be confirmed that the anomaly can be isolated to the tractor side or trailer side.

[0170] In this implementation, the channel fault entry threshold is set to 0.30, the channel recovery threshold to 0.70, the lower limit of output confidence to 0.60, and the upper limit of estimation variance failure to 4.0 × 10⁻⁶. 8 N 2 The maximum prediction hold time is set to 0.30s. An abnormal state is confirmed after four consecutive estimation cycles meeting the fault entry condition; a recovery threshold higher than the fault entry threshold creates a hysteresis interval. During trailer swapping recovery, anomalies related to the convergence of new trailer parameters are handled by extending the confirmation time or reducing the confirmation gain.

[0171] when , , with a total of healthy input and At that time, a two-sided fusion estimation mode is adopted, and the output of S6 is used. The longitudinal force output at the current hinge point is used as the output. If the corrected estimate is already within the common physical feasible set, its value is retained; if the corrected estimate is outside the common set, the common set projection value is output, while the projection distance is retained as the physical constraint state.

[0172] When the health confidence level on one side is no higher than 0.30 and the other side is no lower than 0.70, or when it is continuously confirmed that the anomaly can be isolated to one side, the corresponding one-sided health estimation mode is adopted. This mode only uses the pseudo-measurement on the health side to correct the state prediction value and restricts the correction result to the physically feasible set on the health side; if the physically feasible set on the health side is empty or the key input on the health side subsequently fails, the one-sided mode is not maintained, and the mode is switched to short-term prediction maintenance or failure mode.

[0173] When both sides are temporarily invalid, the common input is temporarily abnormal, the physical conflict cannot be reliably attributed, or there is no valid spurious measurement, but there was a valid output in the previous cycle and the driving force change and motion information required for state prediction are still valid, the short-time prediction hold mode is adopted, according to... Make predictions and make the estimated variance according to As the retention time increases, the reliability of the output decreases monotonically.

[0174] When the continuous holding time exceeds 0.30 s and the estimated variance exceeds 4.0 × 10⁻⁶, 8 N 2 The system enters failure mode when either of the following conditions is met: the output confidence level is below 0.60 or the information required for state prediction is invalid. In failure mode, the longitudinal force values ​​of the hinge points marked as valid will no longer be output; for the purpose of fault recording and recovery initialization, the last predicted value can be retained internally, but the valid external output states must be invalidated, and the boundary conditions that caused the failure must be output at the same time.

[0175] When recovering from the failure mode or short-term prediction hold mode, the system first requires that the health confidence of at least one channel reaches 0.70, the corresponding one-sided physical feasible set is valid, and the output variance and confidence meet the boundary conditions, and this condition is maintained continuously for 20 estimation cycles. Then, it recovers to the healthy one-sided estimation mode. Only when the conditions for both channels, common input, and common physical feasible set continue to be met continuously for 30 estimation cycles does it recover to the two-sided fusion estimation mode. During the recovery process, the single-cycle changes of fusion weights and uncertainties are further restricted to prevent repeated switching near the threshold.

[0176] The current mode is not a failure mode, the output confidence level is not lower than 0.60, and the estimation variance is not higher than 4.0 × 10⁻⁶.8 N 2 When the output is valid, the valid state is set to valid; when in short-term prediction hold mode, the continuous hold time must not exceed 0.30s. The output confidence is determined by the base upper limit of the current mode, the health confidence of the valid side channel, the state of the physical feasible set, the estimated variance, and the hold time. The base upper limit of the two-sided fusion mode is higher than that of the healthy single-sided mode, and the base upper limit of the healthy single-sided mode is higher than that of the short-term prediction hold mode.

[0177] The system ultimately outputs the estimated or transient longitudinal force at the articulation point, the estimation variance or standard deviation, output confidence level, output validity status, current estimation mode, health status of the tractor-side and trailer-side passages, common input anomaly status, confirmed anomaly passages, degree of physical conflict, and candidate anomaly source categories. The candidate anomaly source categories are used by the upper-level controller to determine whether to restrict cooperative drive, cooperative braking, or parameter updates; they are not considered as deterministic diagnostic conclusions that allow for hardware replacement without further testing.

[0178] The above implementation uses a scalar state prediction-innovative correction structure to illustrate the longitudinal force observation process at the hinge point. While maintaining the bilateral pseudo-measurement, common-mode statistical splitting, and uncertainty adaptive relationships unchanged, Kalman filtering, extended Kalman filtering, unscented Kalman filtering, moving window estimation, constrained least squares, or other state observation methods capable of outputting estimated uncertainties can also be used; the replacement of the observer form does not change the technical essence of this invention—utilizing bilateral dynamic redundancy information for self-verification.

[0179] The physically feasible sets on both sides can be represented by a single interval, the union of multiple intervals, a polyhedron, an ellipsoidal set, a combination of interval envelopes and related constraints, or a discrete set formed by constrained sampling. Regardless of the set representation used, the independent physical allowable conditions on both sides, the determination of common intersection, the constraint handling when the intersection is not empty, and the relationship of not forcibly selecting either side and performing conflict attribution when the intersection is empty should be retained.

[0180] Motor torque, mechanical braking, vehicle speed, acceleration, road gradient, trailer identity, and master-trailer connection information can be obtained from different types of sensors, controller internal status, vehicle communication messages, or external road information. Master-trailer communication can be wired or wireless, as long as it can obtain time information and valid status that meet the timing evaluation requirements. For centralized electric drive axles, wheel-side motors, or wheel hub motors, only the axle index, transmission parameters, and wheel system inertia need to be re-determined according to the actual power transmission path, without changing the bilateral dynamic estimation process.

[0181] Based on the disclosure of this invention, any equivalent substitutions made by those skilled in the art regarding module deployment location, signal source, state observation form, parameter identification algorithm, set representation method, and threshold calibration value, as long as the two-sided dynamic redundancy estimation, common difference mode consistency evaluation, two-sided physical feasible domain self-verification, and corresponding selective reinitialization and graded degradation processing under a unified time base are all within the scope of implementation of the technical concept of this invention.

Claims

1. A self-verifying estimation system for longitudinal force at the main hanger hinge point, characterized in that, It includes a comprehensive information acquisition and time-series synchronization module, an axle longitudinal force reconstruction module, a slow time-varying parameter selective identification module, a main and trailer dual-side dynamic redundancy estimation module, a common difference mode consistency adaptive observation module, and a dual-side physical feasible domain self-verification and error attribution module; The integrated information acquisition and timing synchronization module is used to acquire the main trailer operation, vehicle parameters and communication time information, complete unified time base synchronization and evaluate timing effectiveness; The axle longitudinal force reconstruction module is used to reconstruct the longitudinal force of the main trailer drive axle based on drive information, transmission parameters and wheel system dynamic parameters, and to evaluate reconstruction uncertainty. The slow time-varying parameter selective identification module is used to add the longitudinal dynamic equations of the main trailer to eliminate the longitudinal force at the articulation point, and to select a subset of active parameters that meet the identification conditions from the equivalent mass of the trailer, the road slope correction amount and the driving resistance parameters for updating. If the conditions are not met, the update is frozen or restricted. The main trailer dual-side dynamic redundancy estimation module is used to define the longitudinal force at the hinge point on a unified road longitudinal coordinate axis, and construct pseudo-measurements and their uncertainties on both sides of the longitudinal force at the same hinge point based on the driving force, mass, acceleration and resistance on both sides. The common-mode consistency adaptive observation module is used to construct common-mode residuals and differential-mode residuals from the pseudo-measurements on both sides relative to the predicted values. It combines common inputs, evidence of dynamic changes, and health information from both sides to adjust process uncertainty, measurement uncertainty, or fusion weights to obtain the corrected estimate. The bilateral physical feasible region self-verification and error attribution module is used to map the joint allowable set of both sides to the physical feasible set; when the intersection is not empty, the constraint correction estimate is used; when the intersection is empty, the projection is not forced; and the candidate anomaly source is determined by combining the degree of conflict and health information, and the basic estimation result and physical verification status are output.

2. The self-verifying estimation system for longitudinal force at the main hanger hinge point according to claim 1, characterized in that, The integrated information acquisition and timing synchronization module includes an integrated information acquisition unit, a unified time base establishment unit, a time delay compensation unit, and a timing reliability evaluation unit. The integrated information acquisition unit is used to acquire the actual motor torque, motor speed or wheel speed, actual mechanical braking torque or mechanical braking execution status of each electric drive shaft on the tractor side and trailer side, longitudinal speed and longitudinal acceleration of the tractor and trailer, articulation angle, road slope, main-trailer connection status, trailer identification information, transmission and vehicle structure parameters, as well as the source timestamp, frame sequence number, receiving timestamp and communication status corresponding to each piece of information; The unified time base establishment unit is used in the first... Each estimation period determines the unified estimation time. Based on the source timestamp and receiving timestamp of each piece of information, transmission delay, sampling misalignment, frame loss, out-of-order delivery, and clock deviation are identified, and the usable information is sorted, matched, and buffered according to the unified estimated time. The time delay compensation unit is used to first determine the filtering rate of change for continuously changing signals in motor torque, wheel speed, longitudinal speed, longitudinal acceleration and hinge angle, and then extrapolate or interpolate the corresponding signals to the unified estimation time according to their effective time delay; for discrete state signals in gear position, connection status, braking enable status and trailer identity information, timestamp matching or the most recent effective state retention method is used for alignment, and continuous extrapolation across switching times is prohibited in the state switching interval. The time-series reliability evaluation unit is used for any information channel participating in the estimation. Based on this information, the effective delay relative to the unified estimation time and the coupling amount between the effective delay and the signal rate of change, the delay-dynamic coupling mismatch is determined: In the formula, For information channels The effective sampling time As a time delay reference scale for corresponding information, This represents the rate of change of the filtered signal. As a reference scale for the amplitude of the corresponding signal, This is a dimensionless mismatch quantity that simultaneously characterizes the degree of signal change during both the data age and the time delay period; where, , , and All data are time units, and seconds (s) are used throughout the text; estimated period Set within the range of 0.008–0.015 s; filter change rate The unit is the corresponding signal unit per second. It has the same unit as the corresponding signal. It is a dimensionless quantity; The timing reliability evaluation unit is also used to determine the continuous timing reliability factor through monotonically decreasing mapping based on the delay-dynamic coupling mismatch, delay fluctuation, data integrity, and communication status, and to determine the timing validity flag through a preset hysteresis threshold and continuous period confirmation; when information is missing, timeout, out-of-order cannot be recovered, cross-discrete state switching extrapolation, or compensation exceeds a preset range, the corresponding information channel is marked as timing invalid; the integrated information acquisition and timing synchronization module sends the synchronization signal, timing reliability factor, and timing validity flag to the axle longitudinal force reconstruction module, the slow time-varying parameter selective identification module, the main and trailer dual-side dynamic redundancy estimation module, and the common difference mode consistency adaptive observation module; For rapidly changing information such as motor torque, wheel speed, longitudinal velocity, and longitudinal acceleration, the time delay reference scale... The time delay reference scale is set between 0.008 and 0.015 s; for information used to determine slow trends, the time delay reference scale is set between 0.5 and 0.75 s; a dimensionless time series normal threshold is set. Timing failure threshold and recovery threshold ,in Set at 0.8 to 1.5, Set at 4 to 9, Set the value to be between 0.36 and 0.64, and satisfy the following conditions: ;when When the information channel is deemed to be in a valid time sequence, it is allowed to be used normally; when When this occurs, it is judged as time-constrained, increasing the relevant uncertainty or reducing the weight used; when If the time series is invalid, it is considered invalid and stops participating in parameter updates or spurious measurement corrections; if the time series anomaly is confirmed after 3 to 5 consecutive estimation cycles, it is only considered invalid after 3 to 5 consecutive estimation cycles. The channel is restored after the recovery threshold is met for 5 to 10 consecutive estimation cycles; data age exceeding 0.05s for rapidly changing information is directly judged as timeout.

3. The self-verifying estimation system for longitudinal force at the main hanger hinge point according to claim 2, characterized in that, The axle longitudinal force reconstruction module includes a wheel-end torque determination unit, a wheel system dynamic compensation unit, a reconstruction uncertainty and reliability evaluation unit, and a longitudinal driving force synthesis unit on both sides of the main trailer. The wheel-end torque determination unit is used to establish an index set for the electric drive shafts on the tractor side and trailer side. and The union of the two is denoted as the electric drive shaft index set. For any electric drive shaft The electric drive wheel end torque is determined based on the actual motor torque with directional symbols, transmission ratio, main reducer speed ratio, and torque transmission coefficient corresponding to the direction of drive or energy recovery power flow, and the actual mechanical braking torque with directional symbols is superimposed. The gear train dynamic compensation unit is used to subtract the gear train inertial torque corresponding to the equivalent moment of inertia referred to the wheel end and the internal loss torque of the transmission system from the torque applied at the wheel end, and to determine the first [missing formula]. Longitudinal force base reconstruction value of the electric drive shaft: In the formula, This represents the actual motor torque with a direction sign. and These are the transmission ratio of the transmission stage and the transmission ratio of the main reducer, respectively. The torque transmission coefficient is determined based on the power flow direction of driving, coasting, or energy recovery. The actual mechanical braking torque with a direction sign. To calculate the equivalent moment of inertia at the wheel end, This is the filtered wheel angular acceleration. For the internal torque loss of the transmission system with directional signs, Effective rolling radius; Among them, the actual motor torque Actual mechanical braking torque and internal loss torque The unit is Equivalent moment of inertia The unit is Wheel angular acceleration The unit is Effective rolling radius The unit is Longitudinal force The unit is Transmission ratio and torque transmission coefficient All are dimensionless quantities; the upper limit of the mechanical braking torque of a single brake wheel or wheel end is determined according to the vehicle braking system configuration, the number of tires and the effective rolling radius. The actual mechanical braking torque of the i-th electric drive shaft is determined by summing the signed wheel end mechanical braking torques of each brake wheel on that shaft. The reconstruction uncertainty and reliability evaluation unit is used to determine the first uncertainty based on the estimated uncertainty of motor torque, mechanical braking torque, transmission efficiency, effective rolling radius, wheel speed change rate, and internal loss torque. Longitudinal force reconstruction uncertainty of electric drive shaft The non-negative wheel slip index is determined based on the absolute value of the difference between the wheel's circumferential speed and the vehicle's longitudinal speed. The reconstruction uncertainty, wheel slippage degree, and time delay-dynamic coupling mismatch are combined to form the driving force reconstruction mismatch: In the formula, For the first Longitudinal force reference scale for electric drive shaft As a sliding reference scale, The effect coefficient of temporal mismatch on the reconstructed reliable state; longitudinal force reconstruction variance. The unit is Longitudinal force reference scale The unit is Determined based on the maximum longitudinal force on the corresponding axis; slip index Slip reference scale Temporal action coefficient , and All are dimensionless quantities, among which Set at 0.08 to 0.12, Set at 0.1 to 0.3; The reconstruction uncertainty and reliability evaluation unit is also used to determine the continuous reconstruction reliability factor based on the driving force reconstruction mismatch through a monotonically decreasing mapping, and to determine the reconstruction validity flag by combining the integrity of mechanical braking information, power flow identification status, wheel speed and vehicle speed validity status, wheel slip status, and continuous period confirmation; under compound braking conditions, when the actual mechanical braking effect is unavailable or cannot be reliably estimated, the corresponding reconstruction uncertainty is increased, or the corresponding drive shaft is marked as reconstruction limited or reconstruction invalid; when the wheel is in a high slip state, the reconstruction result of this shaft is prohibited from being used for slow time-varying parameter updates, and pseudo measurements containing information of this shaft are reduced or stopped from participating in observation correction; Set slip warning threshold and slip invalid threshold ,in Set at 0.08 to 0.12, Set the value to be between 0.18 and 0.25, and satisfy the following conditions: Set the normal reconstruction threshold. and reconstruction failure threshold ,in Set at 1.2 to 2.0, Set the value to 4.0–6.0, and satisfy the following conditions: ;when and When any index is between the corresponding warning threshold and the invalid threshold, a restricted reconstruction value is output and the reconstruction uncertainty is increased, prohibiting or restricting its use for slow time-varying parameter identification. If the restricted state has not recovered for 10 to 30 consecutive estimation periods, it is converted to invalid reconstruction. or When this happens, the drive shaft is marked as invalid for reconstruction, and pseudo-measurements containing information about that shaft are stopped from participating in observation correction; when the actual mechanical braking force is missing under compound braking conditions, the mechanical braking torque must not be treated as zero directly; The longitudinal driving force synthesis units on both sides of the main hangar are used to respectively... and The effective electric drive axle longitudinal force is transformed and synthesized to obtain the longitudinal resultant force at the wheel end of the tractor side and the longitudinal resultant force at the wheel end of the trailer side. Based on the reconstruction uncertainty of each axle and the correlation of common source errors, the uncertainty, credible state and effective state of the forces on both sides are determined and sent to the slow time-varying parameter selective identification module, the main and trailer dual-side dynamic redundancy estimation module and the dual-side physical feasible domain self-verification and error attribution module. For the non-electric drive braking axle, the mechanical braking torque at the wheel end of each brake wheel of the axle is converted into an additional longitudinal force according to the effective rolling radius and included in the corresponding longitudinal resultant force and its uncertainty on the tractor side or trailer side.

4. The self-verifying estimation system for longitudinal force at the main hanger hinge point according to claim 3, characterized in that, The slow time-varying parameter selective identification module includes a vehicle longitudinal dynamics construction unit, an active parameter subset selection unit, a parameter identifiability evaluation unit, an excitation gating parameter update unit, and a parameter constraint and uncertainty output unit. The vehicle longitudinal dynamics structural unit is used to convert the longitudinal driving force, longitudinal acceleration, and driving resistance on the tractor side and trailer side to a unified road longitudinal coordinate system, and to add the longitudinal dynamic equations on both sides to eliminate the longitudinal forces at the hinge points with opposite directions in the equations, thus constructing the following vehicle longitudinal dynamics relationship: In the formula, and These are the equivalent mass of the tractor and the equivalent mass of the trailer, respectively. and These represent the longitudinal accelerations of the tractor side and the trailer side in a unified road longitudinal coordinate system, respectively. and These are the longitudinal driving forces on the tractor side and the trailer side, respectively. and These are the equivalent driving resistances formed by external forces other than the longitudinal driving force and the hinge point force on the longitudinal coordinate axis of the unified road, with resistance to the vehicle's movement along the positive longitudinal direction of the road being taken as positive and assistance to the vehicle's movement along the positive direction being taken as negative; where, equivalent mass , The unit is longitudinal acceleration , The unit is driving force , and driving resistance , The unit is The road slope and slope correction amount are consistently expressed in slope angle throughout the text. This indicates that it should not be used in conjunction with the percentage of slope. The vehicle longitudinal dynamics structural unit is also used to determine the road slope benchmark using at least one of on-board slope sensing information, inertial measurement information, satellite positioning information, electronic map information or existing slope observation results, and to use the slope correction amount as the slow time-varying parameter to be identified only within a preset bounded range, so as to avoid estimating the road slope and trailer mass as completely free unknowns at the same time. The active parameter subset selection unit is used to select the active parameter subset for the current estimation period from trailer equivalent mass, road slope correction, rolling resistance parameters, air resistance parameters, and unmodeled resistance correction based on vehicle speed range, longitudinal acceleration excitation, road slope change, driving force change, normalized sensitivity energy of each candidate parameter, and the correlation between parameters. And keep the parameters outside the subset of active parameters as the previous valid estimate or the preset prior value; The parameter identifiability evaluation unit is used to construct an activity subset information matrix within a preset sliding time window using the normalized sensitivity of the corresponding parameters for the activity parameter subset. The input comprehensive credibility factor is determined based on the reconstructed credibility state, temporal credibility state, and longitudinal motion state credibility state according to the driving force. The identifiable incentive index for the subset of activity parameters is calculated using the following formula: In the formula, and These are the minimum and maximum eigenvalues ​​of the matrix, respectively. This is the denominator protection value; since the sensitivity of the active parameters has been scaled, , eigenvalue ratio, and identifiable incentive index All are dimensionless quantities; set the threshold under excitation. and incentive threshold ,in Set at 0.02 to 0.05, Set the value to be between 0.08 and 0.15, and satisfy the following conditions: The lower limit of the input comprehensive confidence factor is set at 0.60 to 0.80; the activity information matrix is ​​constructed using a sliding time window of 1 to 3 seconds, with the upper limit of the condition number set at 20 to 50 for normal operation and 100 to 500 for restricted updates. The excitation gating parameter update unit is used to update only the parameters in the active parameter subset that satisfy the preset information content, matrix rank, conditionality, and input confidence threshold. When the active parameter subset does not meet the identifiable conditions, the active parameter subset is reduced and re-evaluated, or the update of the corresponding parameters is frozen or restricted. Within the same estimation period, the equivalent mass of the trailer, the road slope correction, and all driving resistance parameters are not required to be updated freely at the same time. The parameter update adopts a recursive estimation method with physical boundaries, an upper limit on the update rate, and anomaly feedback suppression. when When, freeze a subset of activity parameters; when When, limit the single-cycle update amplitude of parameters; when Normal updates are allowed; an upper limit for the correlation of dimensionless parameters is set. The correlation is between 0.90 and 0.97; when the correlation between any candidate parameter pair exceeds... If the activity information matrix does not meet the rank-sum conditional requirement, the subset of activity parameters is reduced and re-evaluated; the trailer equivalent mass, road slope correction, and rolling resistance parameters are constrained to the physical allowable range determined based on the current vehicle configuration, load conditions, and road type; the upper limit of the single-cycle update range of the trailer equivalent mass, slope correction, and rolling resistance coefficient is set at 0.05% to 0.2% of the current value of the trailer equivalent mass. and The slope continues to use an external slope benchmark plus a bounded correction method. Under unidentifiable conditions, the slope correction amount, trailer equivalent mass and all resistance parameters must not be updated freely at the same time. The parameter constraint and uncertainty output unit is used to output the updated trailer equivalent mass, road slope correction, driving resistance parameters, active parameter subset, identifiable excitation index, parameter update effective state and corresponding estimated uncertainty, and send them to the main trailer dual-side dynamic redundancy estimation module and the dual-side physical feasible region self-verification and error attribution module.

5. The self-verifying estimation system for longitudinal force at the main hanger hinge point according to claim 4, characterized in that, The main-mount dual-side dynamic redundancy estimation module includes a unified longitudinal coordinate transformation unit, a dual-side longitudinal dynamic construction unit, a dual-side pseudo-measurement generation unit, a pseudo-measurement uncertainty and credibility evaluation unit, and a basic fusion output unit. The unified longitudinal coordinate transformation unit is used to establish a unified road longitudinal coordinate axis with the road tangential direction at the articulation point location. Based on the tractor attitude, trailer attitude, and main-trailer articulation state, it transforms the longitudinal driving force, longitudinal velocity, longitudinal acceleration, and driving resistance of the tractor side and trailer side to the unified road longitudinal coordinate axis. The longitudinal force at the articulation point is defined as the component of the force exerted by the main-trailer articulation point on the unified road longitudinal coordinate axis, with the direction of the tractor pulling on the trailer as positive and the direction of the trailer pushing the tractor as negative. The dual-sided longitudinal dynamics construction unit is used to establish the longitudinal dynamic relationship between the tractor side and the trailer side on the unified road longitudinal coordinate axis, respectively. The longitudinal force at the articulation point in the tractor side equation is the same in magnitude and opposite in direction to the longitudinal force at the articulation point in the trailer side equation. The driving resistance is a signed quantity that is positive for hindering the positive longitudinal movement of the road, including gradient resistance, rolling resistance, air resistance, and unmodeled external resistance correction. The driving resistance is distinguished from the internal loss determined based on the internal loss torque of the transmission system during the reconstruction of the longitudinal force of the drive shaft, and the internal loss is not repeatedly included in the driving resistance. The dual-sided pseudo-measurement generation unit is used to construct pseudo-measurements of the longitudinal force at the same articulation point on the tractor side and trailer side respectively, based on the dual-sided longitudinal dynamic relationship: In the formula, and Redundant pseudo-measurements of the longitudinal force structure at the same articulation point are provided on the tractor side and trailer side, respectively. and These are the current valid values ​​of the equivalent mass of the tractor and the equivalent mass of the trailer, respectively. Other values ​​are marked with superscript. Both force and acceleration are components on the longitudinal coordinate axis of the unified road; among them, , , , , and The unit is N, equivalent mass , The unit is longitudinal acceleration , The unit is pseudo-measurement variance The unit is ; The pseudo-measurement uncertainty and reliability evaluation unit is used to determine the measurement uncertainty, basic reliability state, and effective state of the pseudo-measurements on both sides based on the joint uncertainty of the driving force, equivalent mass, longitudinal acceleration, road slope, driving resistance, and time synchronization information used by each pseudo-measurement on both sides. For the common errors on both sides caused by road slope, air density, articulation state, or common motion state information, the correlation between the pseudo-measurements on both sides is retained, and the pseudo-measurement errors on both sides are not assumed to be independent of each other. When the driving force reconstruction, motion state, parameter estimation, or time sequence state on one side is invalid, the pseudo-measurement on that side is marked as invalid, or its measurement uncertainty is increased. Set the upper limit of the allowable standard deviation for tractor side pseudo-measurements. for Set the upper limit of the permissible standard deviation for trailer-side pseudo-measurements. for Only when the reconstructed driving force, mass and drag parameters, longitudinal acceleration, timing information, and attitude and articulation information required for coordinate transformation are all valid, and When valid pseudo-measurements are generated, the corresponding pseudo-measurement uncertainty is increased when there are restricted inputs. When there are key invalid inputs or the pseudo-measurement variance exceeds the allowable upper limit, the pseudo-measurement on that side is marked as invalid. The decrease in credibility only affects the measurement uncertainty or fusion weight and cannot be represented by directly reducing the pseudo-measurement force value. The basic fusion output unit is used to determine the basic fusion weight based on the measurement uncertainty, cross-correlation, basic credible state, and effective state of the pseudo-measurements on both sides. When at least one pseudo-measurement is effective, it forms the basic estimate of the longitudinal force at the hinge point and its estimation uncertainty. The pseudo-measurements on both sides, the basic fusion weight, the pseudo-measurement uncertainty, the effective state, the basic estimate and its estimation uncertainty are sent to the common difference mode consistency adaptive observation module and the two-sided physical feasible domain self-verification and error attribution module.

6. The self-verifying estimation system for longitudinal force at the main hanger hinge point according to claim 5, characterized in that, The common mode consistency adaptive observation module includes a state prediction unit, a common mode statistic construction unit, an uncertainty diversion adjustment unit, and a prediction-innovation correction unit. The state prediction unit is used to obtain the predicted value of the longitudinal force at the hinge point for the current estimation period based on the estimated value of the longitudinal force at the hinge point in the previous estimation period, the current change in vehicle driving force, the longitudinal motion state, and the preset state transition relationship of the longitudinal force at the hinge point. and prediction uncertainty; The common mode statistics construction unit is used to construct tractor-side innovation, trailer-side innovation, common mode residual, and differential mode residual based on the pseudo-measurements on both sides and the predicted values: In the formula, and The weights are non-negative weights determined based on the basic credible states on both sides, the pseudo-measurement uncertainty, and the effective state, and the sum of the two is 1; Used to characterize the degree of common offset between the two spurious measurements and the predicted value. Used to characterize the degree of inconsistency between spurious measurements on both sides; The common mode statistic construction unit is also used to determine the common mode residual variance while taking into account the cross-correlation of pseudo-measures on both sides. Sum and difference model residual variance The dimensionless normalized common modulus statistic and the normalized difference modulus statistic are formed according to the following formula: In the formula, This is the denominator protection quantity that has the same dimensions as the corresponding residual variance; The uncertainty shunting adjustment unit is used to adjust the basic innovation variances of both sides. Determining the inconsistency of the differential mode in the first The innovation contribution coefficient on both sides is used to form the process uncertainty adjustment driver and the measurement uncertainty adjustment driver on both sides: In the formula, For the denominator protection quantity, Used for adjusting the uncertainty of the longitudinal force state process at the drive hinge point. Used to drive the Adjustment of spurious measurement uncertainty or fusion weights; among which, , , and The unit is , , , and The unit is ; , , , , , and All are dimensionless quantities; common mode entry thresholds are set separately. Common-mode recovery threshold Differential mode entry threshold Sum and difference mode recovery threshold ,in and All are set between 6.63 and 10.

83. and All are set between 2.71 and 3.84, and satisfy the following conditions: and ; when and If neither exceeds its respective entry threshold, it is determined that both sides are consistent and the prediction is normal, maintaining the basic process uncertainty and measurement uncertainty; when Exceeding the entry threshold When the threshold is not exceeded, only when there is a shared healthy input and evidence of a change in driving force or acceleration is the longitudinal force dynamic change at the possible articulation point determined and the process uncertainty adjusted; when Exceeding the entry threshold When the threshold is exceeded, it is determined that the two sides are inconsistent with each other due to pseudo-measurements. Based on the innovation contribution and independent health information, the measurement uncertainty of the side with the larger contribution is increased or its fusion weight is decreased. and When all exceed the entry threshold, it is determined that it is currently impossible to distinguish the real dynamic changes, common input anomalies, or multi-source mismatches by relying solely on statistical residuals. The update of the relevant slow time-varying parameters is then suspended and further verification is required. Common-mode anomalies or differential-mode anomalies are only confirmed after the corresponding statistic has exceeded the entry threshold for 3 to 5 consecutive estimation periods; anomaly recovery is only completed after the corresponding statistic has not exceeded the recovery threshold for 8 to 20 consecutive estimation periods; the absolute change in the fusion weight within a single estimation period shall not exceed 0.05 to 0.10, and the upper limit of the single-period amplification factor for process uncertainty and measurement uncertainty shall be set at 1.2 to 1.5; differential-mode anomalies shall not be used alone to determine the specific cause of the fault. The uncertainty diversion adjustment unit only increases or adjusts the process uncertainty based on the common mode statistics when the road slope, common speed or acceleration, unified time base, and other common inputs are in a healthy state, and the common offset is consistent with the dynamic change evidence information formed by changes in driving force, acceleration, or vehicle operating state. When the common input is abnormal or the dynamic change evidence is not valid, the uncertainty related to the pseudo-measurements on both sides caused by the abnormal common input is increased first, or the update of the corresponding slow time-varying parameters is suspended. The uncertainty shunting adjustment unit increases the measurement uncertainty of at least one pseudo-measurement or decreases the corresponding fusion weight based on the differential mode statistics, the innovation contribution coefficients of both sides, and the independent health information of both sides. The differential mode statistics are only used to measure the degree of inconsistency between the two channels and are not used as sufficient evidence to determine sensor failure, driving force reconstruction failure, or parameter failure. The single-cycle adjustment range of process uncertainty, measurement uncertainty, and fusion weight is constrained by a preset upper limit and continuous confirmation and hysteresis recovery are used to suppress repeated adjustments. The prediction-innovation correction unit is used to perform a correction based on two-sided effective innovation on the predicted value of the longitudinal force at the hinge point using adjusted process uncertainty, measurement uncertainty, or fusion weights. When only one side of the pseudo-measurement is effective, a correction based on the innovation on that side is performed. When there is no effective pseudo-measurement, the prediction is maintained and the estimation uncertainty is increased, and the corrected estimate is output. The estimation uncertainty, common mode statistic, and two-sided adjusted measurement uncertainty are included; the prediction-innovation correction can be implemented by a Kalman-type observer, an information filter, or other observers with state prediction and innovation correction structures.

7. A self-verifying estimation system for longitudinal force at the main hanger hinge point according to claim 6, characterized in that, The dual-sided physical feasible domain self-verification and error attribution module includes a joint permissible set construction unit, a physical feasible set mapping unit, a common feasible set determination and estimation constraint unit, and a physical conflict attribution and feedback unit. The joint allowable set construction unit is used to combine the physical boundaries, estimation uncertainties, interrelationships, and effective states of driving force, equivalent mass, longitudinal acceleration, road gradient, driving resistance, articulation state, and timing compensation into a joint allowable set for both the tractor side and the trailer side. ; The physically feasible set mapping unit is used to map the first... Side joint allow set via the first Side dynamics mapping Form the physically feasible set of longitudinal forces at the hinge point on this side, and determine the intersection of the physically feasible sets on both sides: In the formula, For the first The dynamic inputs and vehicle parameters used for side pseudo-measurements; the physically feasible set can be a single interval, multiple discontinuous intervals, or other sets formed by constraint mappings; The common feasible set determination and estimation constraint unit is used to determine the physical self-verification estimate with the smallest deviation from the corrected estimate within the common physical feasible set when the common physical feasible set is not empty; when the common physical feasible set is empty, forced projection is not performed, and the degree of physical conflict is characterized by the set distance between the two physical feasible sets. In the formula, This means determining the feasible value with the smallest deviation from the input estimate in the common physical feasible set; when there are multiple feasible values ​​with the same deviation, the physical self-verification estimate is determined based on the continuity with the previous valid estimate and the current direction of dynamic change. The minimum distance between the two physically feasible sets; where the longitudinal force at the hinge point and the set distance in the physically feasible set are... All units are The physical boundaries of vehicle longitudinal acceleration include the upper limit of drive acceleration, the lower limit of conventional braking deceleration, and the lower limit of emergency braking deceleration, all determined based on the current vehicle configuration. As a dimensionless normalized conflict level; setting a conflict warning threshold. The value is 0.05 to 0.

10. Severe conflict threshold for and satisfy ; The physical conflict attribution and feedback unit is used to maintain the corrected estimate as a transient estimate and increase its uncertainty or decrease its confidence state when the common physical feasible set is empty; within the preset maximum relaxation range, it separately or in a constrained combination relaxes the joint allowable subsets corresponding to driving force reconstruction, vehicle parameters, longitudinal motion state, road and driving resistance, timing synchronization and articulation state, and determines the minimum relaxation category and relaxation degree required to restore the intersection of the two physical feasible sets; The physical conflict attribution and feedback unit is also used to combine the minimum relaxation result with the independent temporal health status, driving force reconstruction health status, motion state health status, parameter identifiable status, and physical conflict duration of the corresponding channel to determine the candidate anomaly source and anomaly contribution order. The candidate anomaly source is not the sole basis for determining the cause of the fault in a single estimation period. It only triggers the bounded reduction of the corresponding channel, the freezing of the corresponding parameter update, resynchronization, reconstruction restriction, or observation mode switching after continuous confirmation. It also sets an upper limit and hysteresis recovery condition for the feedback adjustment amount in each estimation period to suppress the positive feedback between error attribution and weight adjustment. Physical projection is performed when the common physical feasible set is not empty; physical projection is performed when the common physical feasible set is empty and... When, maintain the transient estimate, increase the estimation uncertainty, and continue to observe; when And when there are 3 to 5 consecutive estimation periods, candidate anomaly attribution and downgrade processing is initiated; when When this occurs, it is classified as a restricted conflict, and the adjustment range of the fusion weight is limited; the maximum bounded relaxation of the joint allowable set is set at... If the intersection cannot be restored by relaxing the bounded set, then proceed to the common input anomaly or multi-source anomaly handling; neither side should be forcibly selected; the common physically feasible set should be restored to be non-empty and For 5 to 10 consecutive estimated periods, the value should not exceed [the specified value]. Afterwards, the restricted conflict state is lifted.

8. The self-verifying estimation system for longitudinal force at the main hanger hinge point according to claim 7, characterized in that, It also includes a trailer replacement identification and selective reinitialization module, which includes a trailer replacement feature extraction unit, a trailer replacement event determination unit, a parameter and observation status reinitialization unit, and a trailer replacement recovery status management unit. The trailer replacement feature extraction unit is used to obtain trailer identity information, main trailer connection status, trailer structural parameters, axle configuration, drive system configuration, wheel system parameters, and trailer mass or driving resistance characteristics. The trailer replacement event determination unit prioritizes the formation of trailer replacement candidate states based on changes in the trailer's unique identity information or events such as the disconnection and reconnection of the main trailer after disconnection. In the absence of unique identification information, trailer replacement is confirmed by combining the reconnection event with at least one of the following structural or drive fingerprints: trailer structural parameters, axle configuration, drive system configuration, or wheel system parameters; changes in trailer mass or driving resistance are only used as auxiliary confirmation information, and changes in mass themselves are not sufficient conditions for trailer replacement. The trailer replacement event determination unit is also used to continuously and periodically confirm the candidate status of trailer replacement, and set hysteresis conditions for trailer replacement entry and cancellation in combination with identity information recovery, connection status stability and structural or driving fingerprint consistency, so as to distinguish trailer replacement, reloading of the same trailer and temporary communication interruption. The order of comparison for swapping features is as follows: trailer unique identification information, reconnection event after main trailer disconnection, structural or drive fingerprint, and mass or drag auxiliary features. When there is a change in identity, a replacement candidate can be formed by combining the connection status; when identity information is missing, a reconnection event must be detected simultaneously, and the value must be a dimensionless structure or driving fingerprint difference quantity with a value of 0 to 1. Exceeding the fingerprint difference threshold set between 0.10 and 0.25 ; a trailer swap should not be confirmed if there is only a change in trailer mass or driving resistance; a re-initialization can only be triggered after the candidate trailer swap meets the confirmation conditions for 20 to 50 consecutive estimation cycles. The parameter and observation state reinitialization unit is used to divide the slow time-varying parameters into tractor parameter subvectors, trailer parameter subvectors, and current road parameter subvectors after confirming a trailer replacement. It replaces only the trailer parameter subvector and its covariance with the new trailer's prior values ​​and their prior uncertainties, while retaining the still valid tractor parameters and current road parameters. In the formula, and These are the initial values ​​of the relevant parameters and the initial covariance of the new trailer, respectively. The parameters and covariances marked with "-" are the corresponding estimates that are still valid before the trailer replacement. Each component of the parameter vector is represented by SI units corresponding to its physical meaning. The dimension of any element in the parameter covariance matrix is ​​the product of the units of the two corresponding parameters. Structural or driving fingerprints are scale-normalized before comparison. and All are dimensionless quantities; The parameter and observation state reinitialization unit is also used to clear or reinitialize the driving force reconstruction history, parameter recursion history, two-sided innovation statistics, common mode residuals, differential mode residuals, two-sided measurement uncertainty adaptive state, physical feasible set, physical conflict state and error attribution history related to the original trailer; when the two-sided pseudo-measurements of the new trailer are valid, the initial value of the longitudinal force observation at the articulation point is set with the basic fusion result corresponding to the new trailer; when the two-sided pseudo-measurements are not yet valid, the initial value of the observation is set according to the current valid vehicle state and longitudinal dynamics prediction result, and the observation uncertainty is initialized according to the prior uncertainty of the new trailer. The trailer replacement recovery status management unit is used to limit the single-cycle update amplitude of trailer-related parameters after trailer replacement, increase the initial uncertainty of pseudo-measurement and observation status of the new trailer, and determine the recovery status based on the identifiable status of the new trailer parameters, bilateral differential mode statistics, bilateral physical consistency, observation estimation uncertainty, and the continuous satisfaction period of the above conditions. When the recovery conditions are not met, the trailer replacement recovery mode is maintained and the misjudgment of the normal re-convergence process as a fault is suppressed. After the recovery conditions are continuously satisfied for a preset period, the trailer replacement recovery mode is exited, and the trailer replacement flag, recovery status, re-initialization parameters, and observation status are output for use in the process of selective identification of slow time-varying parameters, adaptive observation of common differential mode consistency, self-verification of bilateral physical feasible region, and optional fault confirmation and graded degradation output. The replacement recovery must simultaneously meet the following conditions: the identifiable incentive index of the activity parameter is not lower than the incentive upper threshold, the differential mode statistic is not higher than the differential mode recovery threshold, the common physical feasible set is not empty, and the estimated variance is not higher than the preset recovery upper limit, and the preset recovery time is maintained continuously; if the above conditions are not met, the replacement recovery mode is maintained.

9. A self-verifying estimation system for longitudinal force at the main hanger hinge point according to claim 7, characterized in that, It also includes a fault confirmation and graded degradation output module, which is used to receive the physical self-verification estimate or transient estimate, estimation uncertainty, physical consistency status, physical conflict status, candidate anomaly source, tractor-side and trailer-side channel health information and common input health information, and determine the final estimation mode, output credibility and output validity flag based on the information. The fault confirmation and graded degradation output module includes a dual-channel health status evaluation unit, an abnormality persistence confirmation unit, a graded degradation mode selection and estimation unit, and an output reliable status and effective status evaluation unit. The dual-side channel health status evaluation unit is used to evaluate the health status of the estimated channels on both sides based on the temporal reliability status, driving force reconstruction reliability status, motion state validity, parameter estimation validity, pseudo-measurement innovation and its uncertainty, physical feasible set status, and candidate anomaly sources of the tractor side and trailer side, respectively. Anomalies caused by road gradient, common speed or acceleration, unified time base, and other shared information are marked as shared input anomalies. The non-intersection of the dual-side physical feasible sets does not automatically indicate that both channels are simultaneously failed. Only by combining candidate anomaly sources, independent health information, and persistent status can the isolable anomaly side be determined. The abnormality continuous confirmation unit is used to determine the confirmed abnormal status of the tractor side, trailer side and common input respectively by adopting continuous multi-cycle confirmation and a hysteresis recovery method that separates the abnormality entry threshold and the abnormality recovery threshold. Within a single estimation period, only bounded adjustments are allowed to process uncertainty, measurement uncertainty, fusion weights, parameter update magnitude, and channel health status. When a trailer replacement event is detected and the system is in the trailer replacement recovery phase, related anomaly confirmation is paused, decayed, or handled using recovery phase thresholds to avoid identifying the normal reconvergence process of the new trailer parameters and the observed state as a persistent fault. Among them, the Side channel health reliability and output credibility The values ​​range from 0 to 1, and the variance of the longitudinal force estimation at the hinge point is... The unit is Duration The unit is Set the channel fault entry threshold. The channel recovery threshold is between 0.20 and 0.

40. It is between 0.65 and 0.80, and satisfies... Set the lower limit of output confidence. The range is 0.50–0.70; estimate the upper limit of variance. The maximum holding time is determined based on the longitudinal force reference scale at the hinge point and the preset relative error limit. It ranges from 0.2 to 0.

4. The abnormal state is confirmed after 3 to 5 consecutive estimation cycles. The graded degradation mode selection and estimation unit is used to select between the two-sided fusion estimation mode, the tractor-side single-sided estimation mode, the trailer-side single-sided estimation mode, the short-term prediction and maintenance mode, and the failure mode based on the health status of both sides, the status of the common physical feasible set, the common input abnormal status, and the previous effective estimation status. When the health confidence of both channels is not lower than the recovery threshold, the common input is healthy, and the common physical feasible set is not empty, the system enters the dual-side fusion mode; when anomalies in one channel have been continuously confirmed, and the health confidence of the other channel is not lower than the recovery threshold, the key input on that side is valid, and the physical feasible set on that side is not empty, the system enters the corresponding single-side health estimation mode for the other side; when both sides are temporarily invalid or conflicts cannot be reliably attributed, but... , , When the information required for state prediction is valid, it enters short-term prediction hold mode; when any of the hold time, estimation variance or output confidence exceeds the corresponding failure boundary, or when the information required for state prediction is missing, it enters failure mode. When both estimation channels are valid and a common physical feasible set exists, a dual-sided fusion estimation mode is adopted, and the longitudinal force estimate of the hinge point constrained by the common physical feasible set is output; when only one estimation channel is valid, or when physical conflicts occur on both sides and it is continuously confirmed that the candidate anomaly can be isolated to one side, the corresponding healthy single-sided estimation mode is adopted, and the longitudinal force prediction value of the hinge point is corrected only by using the pseudo-measurement of that side, and the result is corrected by using the physical feasible set constraint of that side. When both channels are temporarily invalid, the shared input is temporarily abnormal, physical conflicts cannot be reliably isolated to one side, or there is no effective pseudo-measurement, but there is an effective articulation point longitudinal force output in the previous estimation cycle and the motion information required for state prediction is still available, a short-term prediction hold mode is adopted. The previous effective estimated state and the current vehicle dynamic changes are used for prediction, and the estimation uncertainty is increased and the output reliability is reduced as the continuous hold time increases. When the hold time, estimation uncertainty, or the duration of invalid shared input exceeds the preset allowable conditions, or there is no effective vehicle motion information available for state prediction, the failure mode is entered and the current output is marked as having no effective estimate. The graded degradation mode selection and estimation unit is also used to restore the estimation capability in the order of failure mode or short-term prediction maintenance mode, healthy one-sided estimation mode, and two-sided fusion estimation mode after abnormal recovery and continuous satisfaction of hysteresis recovery conditions; during the recovery process, the single-cycle change amplitude of fusion weight and uncertainty is further limited to prevent repeated mode switching. When recovering from failure mode or short-term prediction hold mode, the recovery threshold of at least one channel, the one-sided physical feasibility condition and the output boundary condition should be met for 10 to 30 consecutive estimation cycles to recover to healthy one-sided mode; only after the conditions of both channels, common input and common physical feasibility set continue to be met for 20 to 50 consecutive estimation cycles can the recovery to two-sided fusion mode be restored. The output credibility and validity evaluation unit is used to determine the output credibility and output validity flag based on the current estimation mode, the health status of the valid side channel, the status of the common physical feasible set or the single-sided physical feasible set, the estimation uncertainty, and the continuous holding time. The output validity flag is set to valid only when the current estimation mode is not a failure mode, the output credibility is not lower than the lower limit of output credibility, and the estimation variance is not higher than the upper limit of estimation variance. The fault confirmation and graded degradation output module finally outputs the longitudinal force estimate of the hinge point, the estimation uncertainty, the output credibility, the output validity flag, the current estimation mode, the confirmed abnormal channel, and the candidate abnormal source category. Only if the current mode is not a failed mode and When the output is valid, the output status is set to valid; the final threshold within the above range is calibrated based on the false alarm rate, missed alarm rate and recovery time obtained from simulation or experiment, and remains fixed in the same vehicle configuration.

10. A self-verifying estimation method for longitudinal force at the main hanger hinge point, characterized in that, Includes the following steps: S1. Acquire the actual motor torque, motor speed or wheel speed, mechanical braking information, longitudinal speed, longitudinal acceleration, vehicle structure and transmission parameters, road conditions, articulation status, main-trailer connection status, trailer identification information, and communication time information of the tractor and trailer sides; establish a unified estimated time, filter and estimate the rate of change of continuously changing signals, and perform time delay compensation, interpolation or constrained extrapolation according to the effective sampling time of the corresponding information; perform timestamp matching or most recent valid state preservation for discrete state information; determine the time-series reliable state and time-series valid state of each information channel according to the data age, the degree of signal change during the time delay, time delay fluctuation, data integrity and communication status, and classify each information channel into time-series valid, time-series restricted or time-series invalid states; S2. For the electric drive shafts on the tractor side and trailer side, reconstruct the longitudinal force of each electric drive shaft based on the actual motor torque with direction symbols, actual mechanical braking torque, drive or energy recovery power flow direction, transmission parameters, equivalent rotational inertia of the wheel system, wheel angular acceleration, internal loss torque of the transmission system, and effective rolling radius. The actual mechanical braking force of the non-electrically driven braking axle is converted into an additional longitudinal force, and the longitudinal resultant forces on the tractor side and trailer side are synthesized separately. Based on the reconstruction input uncertainty, the completeness of mechanical braking information, the power flow identification status, the wheel slip status, and the time sequence validity status, the longitudinal force reconstruction uncertainty, reconstruction credibility status, and reconstruction validity status are determined. When the actual mechanical braking force is missing, there is high slip, or the key input is invalid, the corresponding reconstruction uncertainty is increased, and the corresponding results are restricted or prohibited from being used for slow time-varying parameter updates, or pseudo-measurements containing the corresponding axle information are stopped from participating in observation correction. S3. Transform the longitudinal resultant force, longitudinal acceleration, and driving resistance of the tractor side and trailer side to a unified road longitudinal coordinate system. Add the longitudinal dynamic equations of both sides to eliminate the longitudinal forces at the hinge point that are equal in magnitude and opposite in direction in the dynamic equations of both sides, and construct the longitudinal dynamic relationship of the whole vehicle. Determine the road slope benchmark based on the external road slope information and estimate the slope correction within a bounded range. Based on the longitudinal excitation, input confidence state, normalization sensitivity of candidate parameters, rank and conditionality of the activity information matrix, and correlation between candidate parameters, select a subset of active parameters that meet the identifiable conditions from the equivalent mass of the trailer, the road slope correction, and the driving resistance parameters for updating. Freeze or restrict the update of parameters that do not meet the identifiable conditions, and limit the single-cycle change and physical value range of the active parameters. S4. Establish a unified road longitudinal coordinate axis with the road tangential direction at the articulation point location. Determine the component of the force at the articulation point on the unified road longitudinal coordinate axis as the longitudinal force at the articulation point, with the direction of the tractor pulling the trailer as positive. Based on the longitudinal dynamic relationship between the tractor side and the trailer side, construct pseudo-measurements for the longitudinal force at the same articulation point on the tractor side and the trailer side respectively. Based on the joint uncertainty of the longitudinal resultant force, equivalent mass, longitudinal acceleration, road slope, driving resistance, and time synchronization information used by both sides, determine the uncertainty, basic reliability state, and effective state of the pseudo-measurements on both sides, and retain the correlation of pseudo-measurement errors caused by common inputs on both sides. When at least one pseudo-measurement is effective, generate corresponding observation information; when both pseudo-measurements are invalid, stop generating observation correction quantities. S5. Based on the previous valid estimation state and the current vehicle dynamics change, obtain the predicted value of the longitudinal force at the articulation point and its prediction uncertainty. Calculate the innovation of the pseudo-measurements on both sides relative to the predicted value, and construct the common mode residual, differential mode residual, and corresponding normalized statistics based on the innovation on both sides. Determine the innovation contribution of each side in the differential mode inconsistency based on the basic innovation variance and independent health information on both sides. When the common mode is abnormal and the differential mode is normal, adjust the process uncertainty only if the common input is healthy and the common offset provides evidence of vehicle dynamics changes; otherwise, adjust the relevant measurement uncertainty caused by the common input or suspend the update of relevant parameters. When the differential mode is abnormal, adjust the measurement uncertainty or fusion weight of at least one pseudo-measurement based on the innovation contribution and independent health information on both sides. When both the common mode and the differential mode are abnormal, suspend the update of relevant slow time-varying parameters and switch to the two-sided physical feasibility domain verification. Perform prediction-innovation correction using the adjusted process uncertainty, measurement uncertainty, or fusion weight to obtain the corrected estimate of the longitudinal force at the articulation point and its estimation uncertainty. S6. Establish joint allowable sets for both the tractor side and trailer side, including longitudinal resultant force, equivalent mass, longitudinal acceleration, road gradient, driving resistance, articulation state, time-series compensation, input correlation, and effective state. Map these sets to the corresponding longitudinal dynamic relationships to form physical feasible sets for the tractor side and trailer side. When the common intersection of the two physical feasible sets is not empty, determine the physical self-verification estimate with the smallest deviation from the corrected estimate within the common intersection. When the common intersection is empty, do not forcibly project the corrected estimate to either physical feasible set; instead, retain the corrected estimate. The value is used as a transient estimate, increasing its estimation uncertainty or reducing its credibility. The degree of physical conflict is determined based on the set distance between the two physical feasible sets. Bounded relaxation is performed on the joint allowable subsets corresponding to longitudinal force reconstruction, vehicle parameters, longitudinal motion state, road and driving resistance, temporal synchronization and articulation state respectively. The candidate anomaly sources and contribution order are determined by combining the minimum relaxation category, relaxation degree, independent health information and conflict persistence state required to restore the common intersection. Bounded weight reduction, parameter update freezing, resynchronization, reconstruction restriction or estimation mode switching of the corresponding channel are only performed after continuous confirmation. S7. Based on changes in trailer identity information or events such as reconnection after disconnection of the main trailer, a trailer replacement candidate state is formed. When unique identification information is lacking, trailer replacement is confirmed by combining the reconnection event with the trailer structural parameters, axle configuration, drive system configuration, or wheel system parameter fingerprints. Changes in trailer mass or driving resistance are only used as auxiliary confirmation information. After confirming that a trailer replacement has occurred, only the trailer-related parameters and their uncertainties, the trailer-related longitudinal force reconstruction and parameter recursion history, the bilateral innovation statistical state, the measurement uncertainty adaptive state, the physically feasible set, the physical conflict state, and the error attribution history are reinitialized, while the still valid tractor parameters and the current road state are retained. The initial value of the longitudinal force observation at the articulation point is set according to the valid bilateral pseudo-measurement or vehicle dynamics prediction results corresponding to the new trailer, and the observation uncertainty is initialized according to the prior uncertainty of the new trailer. During the trailer replacement recovery phase, the update range of the trailer-related parameters is limited and the corresponding initial uncertainty is increased until the parameter identifiable state, bilateral consistency, common physically feasible set state, and estimated uncertainty continuously meet the recovery conditions. S8. Based on the temporal reliability state, longitudinal force reconstruction reliability state, motion state validity, parameter estimation validity, pseudo-measurement innovation and its uncertainty, physical feasible set state and candidate anomaly sources of the tractor side and trailer side respectively, determine the health state of the estimation channels on both sides respectively, and identify the anomalies caused by road slope, common motion state, unified time base and other common information as common input anomalies. The system employs continuous multi-cycle confirmation and hysteresis recovery to determine the confirmed abnormal state. Based on the health status of both sides, the shared input health status, the common physical feasible set status, the estimation uncertainty, and the previous valid estimation status, it selects between a two-sided fusion estimation mode, a tractor-side health single-sided estimation mode, a trailer-side health single-sided estimation mode, a short-term prediction hold mode, and a failure mode. In the two-sided fusion estimation mode, it outputs an estimated value constrained by the common physical feasible set. In the health single-sided estimation mode, it uses a pseudo-measurement on the healthy side and the physical feasible set on that side to form an estimated value. In the short-term prediction hold mode, it uses the previous valid estimation status and the current vehicle dynamic changes to make a prediction, increasing the estimation uncertainty with the hold time. When the allowable hold conditions are exceeded or the information required for state prediction is lacking, it enters the failure mode. Finally, it outputs the longitudinal force estimate or transient estimate at the articulation point, the estimation uncertainty, the output credible state, the output valid state, the current estimation mode, the confirmed abnormal channel, and the candidate abnormal source category.

Citation Information

Patent Citations

  • A method and device for measuring the articulation force of a mid-axle trailer train

    CN111458058B

  • Vehicle control systems and methods based on trailer hitch joint forces

    US20240010036A1