A pure electric large-tonnage loader walking system and a decoupling control method thereof

Abstract

The application discloses a pure electric large-tonnage loader walking system, which comprises a driving intention acquisition unit for acquiring the driving demand of a driver, a hinged assembly for connecting a front frame and a rear frame, the front frame and the rear frame, a distributed electric drive execution unit composed of a wheel motor and a power inverter, a working condition and load identification unit for identifying a working condition and a load, a vehicle motion state measurement unit for acquiring a vehicle posture and dynamic information, a multi-mode walking collaborative controller, a multi-mode steering collaborative controller, a vehicle domain controller, a vehicle-mounted communication network, a power battery connected with the foregoing structure and used for providing energy, and a working mechanism connected with the front end of the front frame. The application solves the walking stability problem of the pure electric large-tonnage loader walking system on a non-paved road and the collaborative control difficulty caused by the longitudinal, transverse and vertical coupling.

Description

Technical Field

[0001] This invention relates to the technical field of engineering machinery, and in particular to a pure electric large-tonnage loader walking system and its decoupling control method. Background Technology

[0002] With the development of electric drive technology, battery technology, and intelligent control technology, pure electric loaders are gradually becoming an important direction for the electrification of construction machinery. However, existing technologies often overlook the coupling relationship between longitudinal drive, lateral steering, and vertical vibration in large-tonnage vehicles. Loaders frequently experience complex conditions such as acceleration, braking, steering, and loading impacts during operation, resulting in significant dynamic coupling effects between the vehicle's longitudinal driving force, lateral forces, and vertical vibrations caused by uneven road surfaces. Therefore, how to achieve dynamic decoupling between the longitudinal drive, lateral steering, and vertical vibration of the walking system, thereby improving the overall machine's stability and handling, has become a key technical problem that needs to be solved in the field of pure electric large-tonnage loaders.

[0003] Many publicly available solutions follow the approach of using sensor observations to construct coupled state mapping relationships, while others employ decoupling methods through mechanical layout and actuator configuration. For example, Chinese invention patent application number CN202310654264, entitled "A Fully Dynamically Decoupled Servo Control System and Method for a Complex Underactuated Robot"; and Chinese invention patent application number CN202210698675, entitled "Longitudinal Dynamically Decoupled Tiltrotor Aircraft and Its Flight Control Method." These technologies primarily apply to robots and aircraft, making it difficult to directly transfer to pure electric large-tonnage loaders to achieve longitudinal, transverse, and vertical dynamic decoupling.

[0004] Current research faces two key challenges: First, most existing control strategies for pure electric heavy loaders adopt the method of modeling and controlling the longitudinal, transverse and vertical directions separately and independently, ignoring the coupling effect between the three under complex working conditions. This causes the vehicle to easily experience pitching and tilting on uneven roads and under heavy-load loading conditions, affecting the vehicle's handling stability.

[0005] Secondly, the existing control methods for pure electric heavy-duty loaders usually do not fully consider the dynamic coupling relationship between load fluctuations in the operating system and changes in vehicle posture, resulting in large deviations in bucket spatial posture and unstable operating trajectories. This not only affects the smoothness of loading operations and the accuracy of material control, but also reduces the loader's docking positioning accuracy and operating efficiency.

[0006] Therefore, it is necessary to propose a dynamic decoupling control method for the longitudinal, transverse, and vertical coupling of the walking system of a pure electric heavy-duty loader. By using a deep reinforcement learning algorithm based on particle filtering, real-time calculations can be completed under sudden load impacts, thereby achieving synchronous and stable control of the loader's body posture and bucket posture under uneven road surfaces and heavy load conditions. Summary of the Invention

[0007] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0008] In view of the problems existing in the current pure electric heavy-duty loader walking system and its decoupling control method, the present invention is proposed.

[0009] Therefore, the purpose of this invention is to provide a pure electric large-tonnage loader walking system and its decoupling control method, which solves the problem of walking stability of pure electric large-tonnage loader walking system on unpaved roads, as well as the difficulty of coordinated control caused by longitudinal, transverse and vertical coupling.

[0010] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a pure electric large-tonnage loader walking system, comprising: A driving intention acquisition unit for collecting driver's driving needs; a driving input mechanism for driver to control the vehicle; an articulated assembly, front frame, and rear frame for connecting the front and rear frames; a steering hydraulic cylinder for generating relative rotation between the front and rear frames; a distributed electric drive execution unit composed of wheel motors and power inverters; a working condition and load identification unit for identifying operating conditions and loads; a vehicle motion state measurement unit for acquiring vehicle posture and dynamic information; a multi-mode walking cooperative controller; a multi-mode steering cooperative controller; a vehicle domain controller; an on-board communication network; a power battery connected to the aforementioned structures and used to provide energy; and a working mechanism connected to the front end of the front frame.

[0011] As a preferred embodiment of the pure electric heavy-duty loader walking system of the present invention, the driving intention acquisition unit is used to acquire the driver's target driving speed and target turning angle, and input them as the original intention quantity to the multi-mode walking cooperative controller and the multi-mode steering cooperative controller.

[0012] As a preferred embodiment of the pure electric heavy-duty loader walking system of the present invention, the distributed electric drive execution unit consists of drive motors and inverters arranged near the four wheels; the execution unit distributes the torque of each motor according to the differential torque command issued by the multi-mode walking coordination controller and the multi-mode steering coordination controller, so as to realize differential torque closed-loop regulation and motor current limiting protection.

[0013] As a preferred embodiment of the pure electric large-tonnage loader walking system of the present invention, the working condition and load identification unit calculates the changes in the front and rear axle normal loads and tire equivalent stiffness in real time by integrating the hinge angle, lifting height, bucket load inference and vehicle acceleration response, and is used to correct the reference model and related control constraints.

[0014] As a preferred embodiment of the pure electric heavy-duty loader walking system of the present invention, the vehicle motion state measurement unit consists of an inertial measurement component, a vehicle speed measurement module and a yaw rate sensor, used to acquire vehicle posture and yaw dynamic information, and its output and load identification results are provided to the multi-mode walking cooperative controller and the multi-mode steering cooperative controller.

[0015] As a preferred embodiment of the pure electric heavy-duty loader walking system of the present invention, the multi-mode walking coordination controller, the multi-mode steering coordination controller, the vehicle domain controller, the distributed electric drive execution unit, and multiple units achieve data interaction through the vehicle communication network.

[0016] A decoupling control method for the walking system of a pure electric heavy-duty loader includes the following steps: 1) Under static conditions, a nominal longitudinal and transverse model is constructed based on the principles of longitudinal, transverse, and vertical dynamics and bucket dynamics. Machine learning is used to fit the coupling terms of the model, and a three-dimensional coupled dynamic model is established based on the fusion of analytical model and data-driven approach. 2) The system state online correction model during dynamic operation is constructed based on the vehicle sensor signal to build a generalized disturbance observer. Deep reinforcement learning based on particle filtering is used to estimate the time-varying road surface excitation and load impact disturbance. The disturbance estimate containing the disturbance attitude quantity is injected back into the model. 3) Based on the online correction of the longitudinal and transverse coupling model, combined with nonlinear model predictive control and attitude feedforward decoupling compensation, synchronous stability control of the loader's body attitude and bucket attitude is achieved under uneven road surface and heavy load conditions.

[0017] As a preferred embodiment of the decoupling control method for the walking system of a pure electric large-tonnage loader according to the present invention, the process of establishing the three-dimensional coupled dynamic model in step 1) includes: 11) Establish longitudinal, lateral, vertical and bucket dynamic models respectively, combine the four models together, and establish the nominal longitudinal, lateral and vertical models through input vectors and state vectors; 12) Using vehicle motion parameters as input, fit the coupling terms of the longitudinal, transverse, and vertical models through a feedforward neural network; 13) Using parameters such as vehicle structural dynamics as input, a three-dimensional coupled dynamic model is established based on the fusion of analytical model and data-driven approach.

[0018] As a preferred embodiment of the decoupling control method for the walking system of a pure electric large-tonnage loader described in this invention, wherein: the disturbance attitude quantity estimate value in step 2) is injected back into the model: 21) Based on the three-dimensional coupled dynamic model in step 1), the state observation input is formed by multi-source measurement signals such as acceleration and angular velocity from the vehicle-mounted sensor; 22) A generalized disturbance observer is constructed based on vehicle sensor signals. Sensor signals such as speed and wheel speed, which contain uncertain noise, are used as inputs. Through deep reinforcement learning with particle filtering, the disturbance estimates such as attitude disturbance are obtained. 23) The disturbance estimate obtained in step 22) is injected back into the three-dimensional coupled dynamic model to realize the online correction of the longitudinal and transverse coupled model with dynamic parameter updates.

[0019] As a preferred embodiment of the decoupling control method for the walking system of the pure electric large-tonnage loader described in this invention, the specific control process in step 3) is as follows: 31) Based on vehicle dynamics characteristics, structural strength limitations, and road excitation conditions, comprehensively determine the relevant constraint boundaries for vehicle roll and pitch, impact caused by road surface unevenness, and bucket attitude angle and angular velocity. First, based on the vehicle's roll limit and pitch stability boundaries, calculate the maximum allowable roll and pitch angle ranges under extreme conditions using parameters such as center of gravity height and wheelbase, obtaining roll safety thresholds and pitch safety thresholds; then, combining the suspension's natural frequency and damping ratio, determine the boundary value for structural impact caused by the vehicle's vertical velocity, forming the vertical impact safety boundary; finally, by constraining the bucket leading edge height and moment, derive the bucket attitude angle and angular velocity safety boundaries. 32) Nonlinear model predictive control uses an online modified longitudinal and transverse coupled model for rolling optimization, comprehensively considering vehicle attitude, road surface unevenness, load changes, hydraulic actuator constraints and bucket status, to generate the optimal value that satisfies the safety boundary and tracking performance. 33) The feedforward decoupling module compensates in real time for the three-dimensional attitude deviation caused by pitch, roll, structural elasticity and bucket load based on the optimal value generated in 31) and the attitude disturbance output by the generalized disturbance observer, thereby achieving synchronous and stable control of the bucket attitude and the vehicle body attitude.

[0020] The beneficial effects of this invention are: 1. This invention establishes a three-dimensional coupled dynamic model, considers the coupling effect in the longitudinal, transverse and vertical directions under complex working conditions, and proposes a dual-channel control of "feedforward decoupling + disturbance compensation", which can actively predict future coupling effects and compensate for attitude deviations caused by roll and pitch in real time, thereby improving the handling stability of the vehicle.

[0021] 2. This invention introduces a particle filter-based deep reinforcement learning generalized disturbance observer to achieve unified observation of various disturbances, including load fluctuations in the operating system. It is more suitable for large-tonnage electric loaders with strong nonlinearity and multi-source coupling, improving the smoothness of loading operations and the accuracy of material control, and enhancing the robustness and anti-interference of the system.

[0022] 3. This invention introduces safety boundaries in the nonlinear control model based on roll and pitch, impact caused by road surface unevenness, bucket attitude angle and angular velocity, to ensure that the model output values ​​are within a reasonable range, thereby improving the stability and reliability of the vehicle in complex operating environments. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a schematic diagram of the overall structure of the pure electric large-tonnage loader walking system and its decoupling control method of the present invention.

[0024] Figure 2 This is a schematic diagram illustrating the principle of the pure electric large-tonnage loader of the present invention.

[0025] Figure 3 This is a schematic diagram illustrating the principle of the control method of the pure electric large-tonnage loader walking system and its decoupling control method according to the present invention.

[0026] Figure 4 This is a schematic diagram illustrating the principle of solving the roll and pitch control of the pure electric heavy-duty loader's walking system and its decoupling control method, as well as the principle of solving the safety boundary for impacts caused by road surface unevenness. Detailed Implementation

[0027] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0028] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0029] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0030] Secondly, the present invention is described in detail with reference to the schematic diagrams. When detailing the embodiments of the present invention, for ease of explanation, the cross-sectional views illustrating the device structure may be partially enlarged, not according to the usual scale. Furthermore, the schematic diagrams are merely examples and should not limit the scope of protection of the present invention. In addition, actual fabrication should include three-dimensional spatial dimensions of length, width, and depth.

[0031] Reference Figures 1-4 A pure electric heavy-duty loader travel system is provided, including: A driving intent acquisition unit 11 for collecting driver's driving needs; a driving input mechanism 3 for driver to control the vehicle; an articulated assembly 4, a front frame 2, and a rear frame 9 for connecting the front and rear frames; a steering hydraulic cylinder 5 for generating relative rotation between the front and rear frames; a distributed electric drive execution unit 12 composed of wheel motors and power inverters; a working condition and load identification unit 1 for identifying working conditions and loads; a vehicle motion state measurement unit 7 for acquiring vehicle posture and dynamic information; a multi-mode walking coordination controller 14; a multi-mode steering coordination controller 15; a vehicle domain controller 6; an on-board communication network; a power battery 8 connected to the aforementioned structures and used to provide energy; and a working mechanism connected to the front end of the front frame 2.

[0032] The driving intention acquisition unit 11 is used to acquire the driver's target driving speed and target turning angle, and input them as the original intention quantity to the multi-mode driving cooperative controller 14 and the multi-mode steering cooperative controller 15. The driving input mechanism 3 and the driving intention acquisition unit 11 are existing technologies and will not be described in detail here. The distributed electric drive execution unit 12 consists of drive motors and inverters arranged near the four wheels 13. The execution unit performs torque distribution of each motor according to the differential torque command issued by the multi-mode driving cooperative controller 14 and the multi-mode steering cooperative controller 15, so as to realize differential torque closed-loop regulation and motor current limiting protection.

[0033] Among them, the working condition and load identification unit 1 calculates the changes in the normal load of the front and rear axles and the equivalent stiffness of the tires in real time by integrating the hinge angle, lifting height, bucket load inference and vehicle acceleration response, and uses it to correct the reference model and related control constraints. The vehicle motion state measurement unit 7 consists of an inertial measurement component, a vehicle speed measurement module and a yaw rate sensor, and is used to acquire vehicle attitude and yaw dynamic information. Its output and load identification results are provided to the multi-mode walking cooperative controller 14 and the multi-mode steering cooperative controller 15.

[0034] Among them, the multi-mode walking coordination controller 14, the multi-mode steering coordination controller 15, the vehicle domain controller 6, the distributed electric drive execution unit 12, and multiple other units achieve data interaction through the vehicle communication network.

[0035] A decoupling control method for the walking system of a pure electric heavy-duty loader includes the following steps: 1) Under static conditions, a nominal longitudinal and transverse model is constructed based on the principles of longitudinal, transverse, and vertical dynamics and bucket dynamics. Machine learning is used to fit the coupling terms of the model, and a three-dimensional coupled dynamic model is established based on the fusion of analytical model and data-driven approach. 2) The system state online correction model during dynamic operation is constructed based on the vehicle sensor signal to build a generalized disturbance observer. Deep reinforcement learning based on particle filtering is used to estimate the time-varying road surface excitation and load impact disturbance. The disturbance estimate containing the disturbance attitude quantity is injected back into the model. 3) Based on the online correction of the longitudinal and transverse coupling model, combined with nonlinear model predictive control and attitude feedforward decoupling compensation, synchronous stability control of the loader's body attitude and bucket attitude is achieved under uneven road surface and heavy load conditions.

[0036] Specifically, the process of establishing the three-dimensional coupled dynamic model in step 1) includes: 11) Establish longitudinal, lateral, vertical and bucket dynamic models respectively, combine the four models together, and establish the nominal longitudinal, lateral and vertical models through input vectors and state vectors; 12) Using vehicle motion parameters as input, fit the coupling terms of the longitudinal, transverse, and vertical models through a feedforward neural network; 13) Using parameters such as vehicle structural dynamics as input, a three-dimensional coupled dynamic model is established based on the fusion of analytical model and data-driven approach.

[0037] Furthermore, step 11) includes: The longitudinal dynamic model is as follows: (1) In the formula: For vehicle quality; Longitudinal velocity; and The longitudinal force is between the front and rear wheels; For air resistance, This is the drag coefficient. For windward area, air density; For rolling resistance, The rolling resistance coefficient, The slope angle.

[0038] The lateral dynamics model is as follows: (2) (3) In the formula: For lateral velocity; This refers to the yaw rate; Let Z be the moment of inertia of the vehicle about the z-axis; and This represents the distance from the center of mass to the front and rear axles. and The lateral forces are those of the front and rear wheels. and For the front and rear wheel lateral stiffness, Input the steering angle for the front wheels.

[0039] The vertical dynamics model uses a quarter-vehicle model to describe the vertical motion of the suspension and tires: (4) (5) In the formula: For the sprung mass; Unsprung mass; This represents the vertical displacement of the sprung mass. The vertical displacement of the unsprung mass; For suspension spring force, For suspension stiffness; For suspension damping force, The damping coefficient; For tire elasticity, For tire vertical stiffness Input for road surface unevenness.

[0040] The bucket dynamics are as follows: (6) In the formula: The bucket attitude angle; The angular velocity of the bucket; This refers to the bucket angle acceleration; The equivalent moment of inertia of the bucket and its connecting mechanism; The equivalent stiffness of the bucket hinge point; The equivalent damping at the bucket hinge point; The load in the bucket.

[0041] Combining formulas (1)-(6), the nominal longitudinal and transverse dynamic model can be obtained: (7) Where: state vector Input vector System matrix

[0042] Input matrix .

[0043] Further, step 12) includes: Based on the nominal model, coupling terms are introduced. Then the system can be represented as: (8) In the formula: For vehicle parameters.

[0044] By utilizing machine learning to fit the coupling terms, this invention establishes a feedforward neural network model. The model input consists of vehicle motion-related parameters, and these parameters... Learning by minimizing the loss function: (9).

[0045] Furthermore, step 22) includes: A system containing uncertainty noise is represented as: (10) (11) In the formula: It is a three-dimensional coupled dynamic model; The generalized perturbation vector to be estimated; The perturbation input matrix; and This includes process noise and observation noise. This is the observation model.

[0046] Constructing augmented state vectors Particle filtering is used to sample particle importance. (12) in: Let be the state of the i-th particle at time k; For the proposed distribution, this invention adopts a priori proposals, namely: (13) Perform weight updates: (14) in: Let be the likelihood function.

[0047] Assuming the observed noise follows a Gaussian distribution, then: (15)

[0048] In the formula: These are the observed values ​​predicted for the particles.

[0049] Perform weight normalization: (16) Calculate the effective number of particles: (17) when Resampling is triggered at certain times. This is the resampling threshold.

[0050] Based on the particle filter algorithm, a reinforcement learning agent is constructed and trained using the following reward function: (18) in: It represents the negative log-likelihood, which measures the degree of match between the particle weight distribution and the observation; This is the root mean square error; To predict update time.

[0051] Using the algorithm's current distribution characteristics as state input, the agent's output is fed back into the particle filter algorithm to maintain the reasonableness of the weight distribution and estimation accuracy under complex noise and nonlinear coupling conditions. The extracted perturbation estimate is: (19) Furthermore, the online correction of the longitudinal and transverse coupling model in step 23) includes: (20) The model consists of a nominal longitudinal and transverse dynamic model, longitudinal and transverse coupling terms, and perturbation estimates.

[0052] Specifically, step 31) includes: In the optimization solution, constraints related to vehicle roll and pitch, impact caused by road unevenness, and bucket attitude and angular velocity are added: (twenty one) The safety boundaries for each term in the above formula are solved as follows: lateral acceleration The critical value is the value when the overturning moment it generates is equal to the gravitational moment: (twenty two) In the formula: The height of the vehicle's center of gravity; This is the distance from one tire to the longitudinal axis; This refers to the lateral ramp angle.

[0053] Equation (20) can be used to derive: (twenty three) In the formula: For safety factors, a value of 0.5-0.8 is generally used.

[0054] The critical roll angle value is: (twenty four) Critical longitudinal acceleration as follows: (25) In the formula: This is the distance from the center of gravity to the front wheel.

[0055] Equation (23) can be used to derive: (26) In the formula: This is for the safety factor.

[0056] The critical value for pitch angle is: (27) If the maximum vertical acceleration is permissible under structural limits and for passenger comfort, then... The natural frequency of the suspension is The critical value for impact velocity due to road surface unevenness is: (28) The bucket leading edge height is as follows: (29) In the formula: The installation height of the bucket hinge point relative to the ground; This is the length from the bucket hinge point to the leading edge of the bucket.

[0057] To avoid collisions, it is required ,in Let be the minimum height that the leading edge of the bucket should maintain relative to the ground. Then: (30) Summarized as follows: (31) Similarly, to prevent the bucket from being raised too high and exceeding structural limits, it is required that... Then we can obtain: (32) The safe range for the bucket attitude angle is: (33) The equation for the rotational motion of the bucket is as follows: (34) In the formula: This is the torque exerted when the bucket rotates.

[0058] Then for At that time, the range of bucket angular velocity was obtained: (35) Further, step 32) includes: Based on the online-corrected longitudinal and transverse coupling model, the cost function of nonlinear model predictive control is determined: (36) In the formula: For prediction in the time domain; It is a weighted matrix; For the system in The state vector at any given time; For reference only; For the system in Input at any moment; To predict the terminal state vector.

[0059] Based on the above objective function and safety boundary, rolling optimization of nonlinear model predictive control is performed: (37) Obtain the feedforward decoupling control reference input .

[0060] Further, step 33) includes: The feedforward decoupling module obtains the disturbance estimate output by the generalized disturbance observer: (38) in: This is a vertical disturbance; Disturbance caused by the bucket.

[0061] The module is designed with feedforward compensation based on the disturbance amount: (39) in The disturbance compensation gain matrix includes bucket attitude compensation and vehicle attitude compensation.

[0062] The final input to the bucket truck body controller is: (40) The final controller compensates for pitch disturbances, roll disturbances, and vertical disturbances to achieve synchronous and stable control of the vehicle body and the bucket.

[0063] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A pure electric heavy-duty loader walking system, characterized in that, include: A driving intention acquisition unit (11) for collecting the driver's driving needs; a driving input mechanism (3) for the driver to control the vehicle; an articulated assembly (4), a front frame (2), and a rear frame (9) for connecting the front and rear frames; a steering hydraulic cylinder (5) for generating relative rotation between the front and rear frames; a distributed electric drive execution unit (12) consisting of wheel motors and power inverters; a working condition and load identification unit (1) for identifying working conditions and loads; a vehicle motion state measurement unit (7) for acquiring vehicle posture and dynamic information; a multi-mode walking coordination controller (14); a multi-mode steering coordination controller (15); a vehicle domain controller (6); an on-board communication network; a power battery (8) connected to the aforementioned structures and used to provide energy; and a working mechanism (10) connected to the front end of the front frame (2).

2. The pure electric heavy-duty loader walking system according to claim 1, characterized in that: The driving intention acquisition unit (11) is used to acquire the driver's target driving speed and target turning angle, and input them as raw intention quantities to the multi-mode walking cooperative controller (14) and the multi-mode steering cooperative controller (15).

3. The pure electric large-tonnage loader walking system according to claim 2, characterized in that: The distributed electric drive execution unit (12) consists of drive motors and inverters arranged near the four wheels (13); the execution unit distributes the torque of each motor according to the differential torque command issued by the multi-mode walking coordination controller (14) and the multi-mode steering coordination controller (15), so as to realize differential torque closed-loop regulation and motor current limiting protection.

4. The pure electric heavy-duty loader walking system according to claim 3, characterized in that: The working condition and load identification unit (1) calculates the changes in the front and rear axle normal loads and tire equivalent stiffness in real time by integrating the hinge angle, lifting height, bucket load inference and vehicle acceleration response, and uses it to correct the reference model and related control constraints.

5. The pure electric heavy-duty loader walking system according to claim 1, characterized in that: The vehicle motion state measurement (7) unit consists of an inertial measurement component, a vehicle speed measurement module and a yaw rate sensor, and is used to acquire vehicle attitude and yaw dynamic information. Its output and load identification results are provided to the multi-mode walking cooperative controller (14) and the multi-mode steering cooperative controller (15).

6. The pure electric heavy-duty loader walking system according to claim 1, characterized in that: The multi-mode walking coordination controller (14), multi-mode steering coordination controller (15), vehicle domain controller (6), distributed electric drive execution unit (12) and multiple other units achieve data interaction through the vehicle communication network.

7. A decoupling control method for the walking system of a pure electric large-tonnage loader, characterized in that, Includes the following steps: 1) Under static conditions, a nominal longitudinal and transverse model is constructed based on the principles of longitudinal, transverse, and vertical dynamics and bucket dynamics. Machine learning is used to fit the coupling terms of the model, and a three-dimensional coupled dynamic model is established based on the fusion of analytical model and data-driven approach. 2) The system state online correction model during dynamic operation is constructed based on the vehicle sensor signal to build a generalized disturbance observer. Deep reinforcement learning based on particle filtering is used to estimate the time-varying road surface excitation and load impact disturbance. The disturbance estimate containing the disturbance attitude quantity is injected back into the model. 3) Based on the online correction of the longitudinal and transverse coupling model, combined with nonlinear model predictive control and attitude feedforward decoupling compensation, synchronous stability control of the loader's body attitude and bucket attitude is achieved under uneven road surface and heavy load conditions.

8. The decoupling control method for the walking system of a pure electric large-tonnage loader according to claim 7, characterized in that: The process of establishing the three-dimensional coupled dynamic model in step 1) includes: 11) Establish longitudinal, lateral, vertical and bucket dynamic models respectively, combine the four models together, and establish the nominal longitudinal, lateral and vertical models through input vectors and state vectors; 12) Using vehicle motion parameters as input, fit the coupling terms of the longitudinal, transverse, and vertical models through a feedforward neural network; 13) Using parameters such as vehicle structural dynamics as input, a three-dimensional coupled dynamic model is established based on the fusion of analytical model and data-driven approach.

9. The decoupling control method for the walking system of a pure electric large-tonnage loader according to claim 7, characterized in that: In step 2), the perturbation estimate of the perturbation attitude quantity is injected back into the model: 21) Based on the three-dimensional coupled dynamic model in step 1), the state observation input is formed by multi-source measurement signals such as acceleration and angular velocity from the vehicle-mounted sensor; 22) A generalized disturbance observer is constructed based on vehicle sensor signals. Sensor signals such as speed and wheel speed, which contain uncertain noise, are used as inputs. Through deep reinforcement learning with particle filtering, the disturbance estimates such as attitude disturbance are obtained. 23) The disturbance estimate obtained in step 22) is injected back into the three-dimensional coupled dynamic model to realize the online correction of the longitudinal and transverse coupled model with dynamic parameter updates.

10. The decoupling control method for the walking system of a pure electric large-tonnage loader according to claim 7, characterized in that: The specific control process in step 3) is as follows: 31) Based on vehicle dynamics characteristics, structural strength limitations, and road excitation conditions, comprehensively determine the relevant constraint boundaries for vehicle roll and pitch, impact caused by road surface unevenness, and bucket attitude angle and angular velocity. First, based on the vehicle's roll limit and pitch stability boundaries, calculate the maximum allowable roll and pitch angle ranges under extreme conditions using parameters such as center of gravity height and wheelbase, obtaining roll safety thresholds and pitch safety thresholds; then, combining the suspension's natural frequency and damping ratio, determine the boundary value for structural impact caused by the vehicle's vertical velocity, forming the vertical impact safety boundary; finally, by constraining the bucket leading edge height and moment, derive the bucket attitude angle and angular velocity safety boundaries. 32) Nonlinear model predictive control uses an online modified longitudinal and transverse coupled model for rolling optimization, comprehensively considering vehicle attitude, road surface unevenness, load changes, hydraulic actuator constraints and bucket status, to generate the optimal value that satisfies the safety boundary and tracking performance. 33) The feedforward decoupling module compensates in real time for the three-dimensional attitude deviation caused by pitch, roll, structural elasticity and bucket load based on the optimal value generated in 31) and the attitude disturbance output by the generalized disturbance observer, thereby achieving synchronous and stable control of the bucket attitude and the vehicle body attitude.

Images