A pump valve parallel steering system gain self-tuning adaptive robust control method

Abstract

This invention proposes a gain self-tuning adaptive robust control method for a pump-valve parallel steering system. Its mechanical part includes a trapezoidal steering mechanism, which is controlled by two steering assist cylinders in the hydraulic part of the pump-valve parallel steering system to drive the wheels. The hydraulic part includes a pump-valve parallel control system composed of a pump control circuit and a valve control circuit. This system uses a gain self-tuning adaptive robust strategy tailored to the vehicle's steering system to control both the pump control circuit and the valve control circuit. When the pump control circuit supplies oil as needed by adjusting the motor speed using a servo pump, it controls the direction of the directional valve to control the steering direction. When the valve control circuit supplies oil using a fixed-displacement pump, it controls the flow and direction using a servo valve. The pressurized oil from the pump control circuit and the valve control circuit converges through a parallel valve block to the two steering assist cylinders of the trapezoidal steering mechanism and is automatically coupled and distributed to jointly control the trapezoidal steering mechanism. Simultaneously, this invention achieves online self-tuning of the controller gain, ensuring good control performance over a long period.

Description

Technical Field

[0001] This invention relates to the field of automotive technology, and in particular to a gain self-tuning adaptive robust control method for a pump-valve parallel steering system, which is used for the joint control of vehicle subsystems of different types or functions. It can meet the vehicle control requirements that require simultaneous mobilization of the steering system and the power unit (motor, internal combustion engine) for joint control. Background Technology

[0002] As core equipment for long-distance transport of overweight cargo, heavy-duty multi-axle vehicles are the cornerstone of transportation for the smooth progress of many major national projects. Their electro-hydraulic steering system provides precise and flexible steering performance, which is crucial for ensuring efficient and safe transport operations.

[0003] Currently, heavy-duty vehicle steering primarily utilizes valve-controlled or pump-controlled electro-hydraulic steering systems, which are simple in structure, have a high power-to-weight ratio, and are highly reliable. Among these, valve-controlled systems are more widely used due to their high closed-loop control accuracy and engineering reliability, particularly in steering systems where safety is paramount. However, because they rely on throttling speed regulation and overflow unloading, these systems suffer from significant energy losses. In contrast, pump-controlled electro-hydraulic steering systems can theoretically achieve on-demand energy supply, offering significant energy-saving potential. However, limited by the dynamic response characteristics of the pump's flow rate and the pressure build-up rate on the low-pressure side of the steering cylinder, the system response exhibits significant lag and a large dynamic tracking error. This error increases significantly under high-frequency or step-steering demands, thus posing a considerable risk to handling safety.

[0004] Driven by the increasing demands for green industry, heavy-duty vehicles are transitioning from traditional internal combustion engine drives to hybrid powertrains, and may eventually evolve into pure electric vehicles. This leads to a significant decrease in overall vehicle power consumption, making the energy loss rate of valve-controlled steering extremely significant. Therefore, heavy-duty vehicles urgently need a steering system that balances high precision and high energy efficiency.

[0005] Electric power steering systems can achieve both high precision and high energy efficiency in vehicle steering, and are well-suited for hybrid and electric vehicles. However, the electric power steering structure struggles to meet the extremely high steering force requirements of heavy-duty vehicles. Therefore, it is still necessary to find an electro-hydraulic steering system that balances both high precision and high energy efficiency.

[0006] To address the shortcomings of low energy efficiency in valve-controlled systems and lag in dynamic response in pump-controlled systems, a pump-valve parallel system offers a viable solution that balances both. This system significantly improves energy utilization efficiency while achieving fast steering response and control safety under high-frequency and complex operating conditions, and has been validated in fields such as aviation EHA, hydraulic excavators, and cranes. Furthermore, the pump-controlled circuit of this parallel system is driven by an electric motor, while the valve-controlled circuit can be driven by either an electric motor or an internal combustion engine. This balances the high energy efficiency of pure electric drive with the long range of fuel-powered drive when electricity is insufficient, making it well-suited for hybrid vehicles. However, as a novel hydraulic system, the controller design for pump-valve parallel systems is still immature, and there is a lack of relevant research in the field of vehicle steering. Appropriate controller design is needed before its application in vehicle steering.

[0007] In the traditional design of hydraulic system controllers, PID controllers are widely used due to their simple structure and good control performance. However, hydraulic systems themselves have high model orders and suffer from strong nonlinearities such as flow and pressure nonlinearity, mechanical friction nonlinearity, and uncertainties in model parameters caused by pressure, temperature, and steering speed. Furthermore, vehicle steering systems are severely affected by external load disturbances, and the parallel connection of pumps and valves introduces flow coupling uncertainties into the system. PID control, being an error-driven model-free control strategy, completely ignores the nonlinear dynamics, uncertainties, and disturbances of the system model, making it difficult to meet the high-precision and fast-response control requirements of steering systems.

[0008] To achieve better nonlinear control performance and robustness, nonlinear robust controllers such as sliding diaphragm controllers can be used. However, in actual vehicle steering processes, the system model parameters will change with the operating conditions. The uncertainty of the model parameters leads to a reduction in the accuracy of feedforward compensation based on fixed parameter models. Furthermore, the robust control itself has low accuracy, which limits further improvement in control performance.

[0009] For time-varying model parameters, adaptive control has a natural advantage. Adaptive robust control, which combines adaptive and robust control laws, can effectively improve the control accuracy of steering systems. However, engineering practice shows that the performance of the controller gain in heavy vehicle steering systems is highly dependent on real-time driving conditions (vehicle speed, steering angle rate, load distribution, and road surface excitation, etc.). The sensitivity of performance indicators to gain varies greatly under different conditions, and a single fixed gain cannot cover all complex conditions, leading to severe performance degradation of the system under certain driving states. Furthermore, changes in flow distribution in parallel pump-valve systems also affect the controller's parameter requirements. Therefore, there is an urgent need to design an adaptive robust controller that incorporates online gain self-tuning to ensure good control performance over long periods. Summary of the Invention

[0010] This invention proposes an adaptive robust control method for a pump-valve parallel steering system with gain self-tuning. By using an adaptive robust controller to control the pump-valve parallel steering system, the method simultaneously achieves online self-tuning of the controller gain, ensuring good control performance over a long period of time. It can be used for the joint control of vehicle subsystems of different types or functions, and meets the vehicle control requirements that require simultaneous mobilization of the steering system and power unit (motor, internal combustion engine) for joint control.

[0011] The present invention adopts the following technical solution.

[0012] A gain-self-tuning adaptive robust control method for a pump-valve parallel steering system is disclosed. The mechanical part of the pump-valve parallel steering system includes a trapezoidal steering mechanism connected to the wheels. The control end of the trapezoidal steering mechanism is controlled by two steering assist cylinders of the hydraulic part of the pump-valve parallel steering system, driving the wheels to steer through an approximate Ackermann steering motion. The hydraulic control method of the hydraulic part includes a pump-valve parallel control system composed of a pump control circuit and a valve control circuit. The pump-valve parallel control system controls the pump control circuit and the valve control circuit with a gain-self-tuning adaptive robust strategy that matches the vehicle steering system. When the pump control circuit supplies oil as needed by adjusting the motor speed of the servo pump, it controls the direction of the reversing valve to control the steering direction. When the valve control circuit supplies oil by a fixed-displacement pump, it controls the flow and direction with a servo valve. The pressure oil from the pump control circuit and the valve control circuit flows through a parallel valve block to the two steering assist cylinders of the trapezoidal steering mechanism and is automatically coupled and distributed to jointly control the trapezoidal steering mechanism.

[0013] The trapezoidal steering mechanism is a single-degree-of-freedom hinged four-bar mechanism. When used in hybrid vehicles, the pump control circuit of its parallel control system is driven by a servo motor to drive the hydraulic pump, while the valve control circuit hydraulic pump is driven by an electric motor or an internal combustion engine.

[0014] The trapezoidal steering mechanism consists of a fixed axle frame, left and right steering angle members with one degree of rotational freedom mounted on the axle frame, steering wheels mounted on them, and a long shaft connecting the left and right angles.

[0015] The trapezoidal steering mechanism is a single-degree-of-freedom hinged four-bar linkage. The pump-valve parallel control system includes a parallel cylinder assembly formed by a left steering assist cylinder and a right steering assist cylinder (the two cylinders are connected in parallel, and the hydraulic oil will be automatically coupled and distributed, with only one driving degree of freedom counted). The parallel cylinder assembly serves as the actuator of the hydraulic system, and its two ends are connected to the axle frame and the corner piece by hinges respectively. The extension rod movement of the parallel cylinder assembly drives the trapezoidal steering mechanism to perform an approximate Ackermann steering motion to drive the wheels to turn.

[0016] The wheel is mounted on the steering angle component and rotates with it. The wheel rotation angle is measured by an angular displacement sensor installed at the pivot of the steering angle component.

[0017] The gain-self-tuning adaptive robust strategy is a particle swarm optimization (PSO) algorithm that includes adaptive inertia weights and a perturbation strategy, possessing stronger global search capabilities. The mathematical model of the controlled pump-valve parallel steering system is expressed by the following formula:

[0018] Formula 1;

[0019] Where u1 and u2 are the servo pump motor speed control voltage and the servo valve core displacement control voltage of the valve control circuit, respectively; x1 is the right wheel rotation angle of the controlled vehicle; θ is an uncertain parameter; and d is the disturbance.

[0020] The controller of the pump-valve parallel control system uses the steering system operating condition values ​​fed back in real time by sensors, and outputs control signals u1 and u2 according to the controller algorithm to control the steering system.

[0021] The controller operates according to an adaptive robust control law; the adaptive robust control law of the steering system is expressed by the following formula:

[0022] Formula 2;

[0023] Among them, u m For feedback control law, u a For adaptive control law, u s The robust control law is expressed in the following formula:

[0024] Formula 3;

[0025] in, The adaptive value of the uncertain parameter θ is calculated using the following formula:

[0026] Formula 4;

[0027] e1, e2, and e3 are the deviations of state values ​​x1, x2, and x3, respectively, calculated using the following formula:

[0028] Formula 5;

[0029] x 1d x 2d x 3d Let x1, x2, and x3 be the expected values ​​of the states, respectively. The calculation formula is as follows:

[0030] Formula 6;

[0031] k1, k2, k3, and ρ are controller gains, all of which are greater than zero. Given the strong dependence of the gain on the operating conditions, the proposed method achieves condition-driven gain self-tuning by designing an improved particle swarm optimization algorithm that integrates adaptive inertial weights and perturbation strategies.

[0032] The controller gain is self-tuned by using the particle swarm optimization algorithm to drive the steering system's operating conditions. The method is as follows: by traversing a typical set of operating conditions, the global optimization of the controller gain is completed efficiently, and the optimized set of the best gain is embedded into the controller.

[0033] The improved particle swarm optimization algorithm includes the following steps:

[0034] Step 1: Obtain the gain parameters k1, k2, k3, and ρ when the controller executes the adaptive robust control law, perform gain self-tuning, and set appropriate value ranges for each parameter;

[0035] Step 2: Set the number of particles and randomly select the initial position of the particles within the range of values.

[0036] Step 3: Perform simulation based on the gain parameters represented by the particle positions;

[0037] Step 4: Sort the control performance of each particle according to the fitness function value, and update the historical best position of the individual particles and the historical best position of the group.

[0038] Step 5: Determine whether most particles have converged to the vicinity of the group's historical optimal position or have not converged after too long an iteration; if they have converged to the optimal position or have failed to converge, output the result.

[0039] Step 6: If convergence is not yet complete, calculate the particle's velocity in each dimension representing different controller gains using the velocity calculation formula, based on parameters such as the individual particle's historical best value and the particle swarm's historical best value. Since the inertia weight and learning factor used for velocity calculation are fixed in the standard PSO algorithm, the algorithm's convergence speed is slow in the early stages and prone to getting trapped in local optima later. An improved PSO algorithm dynamically updates the inertia weight and learning factor, allowing the inertia weight and individual learning factor to gradually decrease as the particle swarm iterates and converges, while the swarm learning factor gradually increases. This optimization aims to achieve stronger early-stage global search capabilities and better late-stage local convergence capabilities.

[0040] Step 7: Update the particle position based on the individual particle position and velocity.

[0041] In step 3, the fitness function value is calculated based on the RMSE value and response time from the simulation results. This value characterizes the control accuracy and response time performance of the gain at that point; the smaller the value, the better the control performance. The specific formula is as follows:

[0042] Formula 7;

[0043] Where T is the total time for steering control; e1 is the control error; k s and t s The response time proportional coefficient and steering system response time are calculated using different methods and values ​​depending on the type of control signal.

[0044] Step 4: Sort the control performance of each particle according to the fitness function value, and update the individual particle's historical best position and the group's historical best position.

[0045] In step 6, the speed calculation formula used for optimization is:

[0046] Formula 8;

[0047] Where v is the particle velocity; i represents the dimension; n represents the number of iterations; gb and pb are the best historical position of the particle swarm and the best historical position of the individual particle, respectively; x is the particle position; r1 and r2 are random numbers from 0 to 1, representing the randomness of particle motion; ω, a1, and a2 are the inertia weight, the swarm learning factor, and the individual learning factor, respectively, and their variation with the number of iterations is as follows:

[0048] Formula 9;

[0049] Where ω0, a 10 a 20 Let ω, a1, and a2 be the initial values; k ω k a1 k a2 This represents the corresponding coefficient of variation.

[0050] In step 7, the formulas for updating the particle position based on the individual particle position and velocity are as follows:

[0051] Formula 10;

[0052] Where, x in v represents the particle's current position in dimension i; in Let x be the particle's current velocity in dimension i; i (n+1） Update the position of the particle in the i-th dimension.

[0053] In step 7, since the particle swarm is prone to getting trapped in local optima in the standard PSO algorithm, the improved PSO introduces a perturbation strategy. When updating the particle position, there is a small probability that the particle will be perturbed and its position will be randomly changed. This strategy is used to effectively avoid getting trapped in local optima. After the particle position is updated, the fitness function value is recalculated.

[0054] The advantages of this invention are:

[0055] (1) An adaptive robust controller is used to control the pump-valve parallel steering system, and the control effect is good. At present, there is little research on controllers for pump-valve parallel steering systems. There is a lack of controller designs with high precision, fast response and strong robustness. However, the pump-valve parallel system has high requirements for controllers. It is necessary to design a suitable controller for practical applications. This invention is the first to apply an adaptive robust controller to the pump-valve parallel steering system and achieve good control effect.

[0056] (2) By combining particle swarm optimization (PSO) and an adaptive robust controller, online self-tuning of the controller gain is achieved, ensuring good control performance over a long period. The performance of the controller gain in the steering system of heavy vehicles is highly dependent on real-time driving conditions (vehicle speed, steering angle rate, load distribution, and road excitation, etc.), and the sensitivity of performance indicators to gain varies greatly under different conditions. Heavy-duty transport vehicles typically travel long distances, with varying terrain, road conditions, and weather, resulting in complex operating conditions. Conventional controllers have relatively simple gain, and their performance degrades significantly under certain driving conditions, making it difficult to achieve high-precision control over long periods under complex conditions. At the same time, the flow distribution and coupling characteristics of the pump-valve parallel system also exacerbate the need for adaptive controller parameters. Therefore, this invention combines an adaptive robust controller with a particle swarm optimization (PSO) algorithm to ensure good control performance over a long period.

[0057] (3) Based on the particle swarm optimization algorithm, an improved particle swarm optimization algorithm was designed by combining inertial weight adaptation and perturbation strategy, which improves its global search capability and can effectively avoid getting trapped in local optima. The gain of the steering system controller is directly related to the vehicle driving safety. Using conventional random search poses certain safety hazards. It is necessary to select a gain self-tuning adaptive robust strategy that fits the vehicle steering system. Given that the model of the pump-valve parallel electro-hydraulic steering system is relatively complex, the parameters vary greatly, and the global search requirement is large, the particle swarm optimization algorithm, which has the advantages of not requiring gradient information and having strong global search capability, is a more suitable choice. However, this algorithm has the defect of too fast convergence speed and easy to get trapped in local optima. Therefore, this invention introduces inertial weight adaptation and perturbation strategy into the particle swarm optimization algorithm and designs an improved particle swarm optimization algorithm with stronger global search capability. Attached Figure Description

[0058] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0059] Appendix Figure 1 This is a schematic diagram of the pump-valve parallel electro-hydraulic steering system in an embodiment of the present invention;

[0060] Appendix Figure 2 This is a control flowchart of the adaptive robust control law used by the adaptive robust controller in an embodiment of the present invention;

[0061] Appendix Figure 3 This is a flowchart illustrating the improved particle swarm optimization algorithm in an embodiment of the present invention.

[0062] Appendix Figure 4 This is a schematic diagram of the simulation test results in an embodiment of the present invention (a in the figure is a comparison of the trajectory tracking effects of the two controllers, and b in the figure is a comparison of the trajectory deviation of the two controllers). Detailed Implementation

[0063] This example is used for control. Figure 1 The diagram shows a heavy-duty vehicle pump-valve parallel electro-hydraulic steering system. The mechanical structure of this system consists of a trapezoidal steering mechanism with one degree of freedom, controlled by two power steering cylinders in the hydraulic section, performing an approximate Ackermann steering motion. The hydraulic control section is a pump-valve parallel control system, composed of a pump control circuit and a valve control circuit. The pump control circuit is supplied with oil by a servo pump and controls the flow rate, while a directional valve controls the direction. The valve control circuit is supplied with oil by a fixed displacement pump, and a servo valve controls the flow rate and direction. The pressure oil from both circuits converges through parallel valve blocks to jointly control the steering mechanism. Because the controlled mechanism is a single-degree-of-freedom mechanism, the hydraulic oil flowing into the two power steering cylinders will automatically couple and distribute.

[0064] like Figure 1 As shown, a gain self-tuning adaptive robust control method for a pump-valve parallel steering system is disclosed. The mechanical part of the pump-valve parallel steering system includes a trapezoidal steering mechanism connected to the wheels. The control end of the trapezoidal steering mechanism is controlled by two steering assist cylinders of the hydraulic part of the pump-valve parallel steering system, driving the wheels to steer through an approximate Ackermann steering motion. The hydraulic control method of the hydraulic part includes a pump-valve parallel control system composed of a pump control circuit and a valve control circuit. The pump-valve parallel control system controls the pump control circuit and the valve control circuit with a gain self-tuning adaptive robust strategy that matches the vehicle steering system. When the pump control circuit supplies oil as needed by adjusting the motor speed of the servo pump, it controls the direction of the reversing valve to control the steering direction. When the valve control circuit supplies oil by the fixed displacement pump, it controls the flow and direction with a servo valve. The pressure oil of the pump control circuit and the valve control circuit flows through the parallel valve block to the two steering assist cylinders of the trapezoidal steering mechanism and is automatically coupled and distributed to jointly control the trapezoidal steering mechanism.

[0065] The trapezoidal steering mechanism is a single-degree-of-freedom hinged four-bar mechanism. When used in hybrid vehicles, the pump control circuit of its parallel control system is driven by a servo motor to drive the hydraulic pump, while the valve control circuit hydraulic pump is driven by an electric motor or an internal combustion engine.

[0066] The trapezoidal steering mechanism consists of a fixed axle frame, left and right steering angle members with one degree of rotational freedom mounted on the axle frame, steering wheels mounted on them, and a long shaft connecting the left and right angles.

[0067] The trapezoidal steering mechanism is a single-degree-of-freedom hinged four-bar linkage. The pump-valve parallel control system includes a parallel cylinder assembly formed by a left steering assist cylinder and a right steering assist cylinder (the two cylinders are connected in parallel, and the hydraulic oil will be automatically coupled and distributed, with only one driving degree of freedom counted). The parallel cylinder assembly serves as the actuator of the hydraulic system, and its two ends are connected to the axle frame and the corner piece by hinges respectively. The extension rod movement of the parallel cylinder assembly drives the trapezoidal steering mechanism to perform an approximate Ackermann steering motion to drive the wheels to turn.

[0068] The wheel is mounted on the steering angle component and rotates with it. The wheel rotation angle is measured by an angular displacement sensor installed at the pivot of the steering angle component.

[0069] The gain-self-tuning adaptive robust strategy is a particle swarm optimization (PSO) algorithm that includes adaptive inertia weights and a perturbation strategy, possessing stronger global search capabilities. The mathematical model of the controlled pump-valve parallel steering system is expressed by the following formula:

[0070] Formula 1;

[0071] Where u1 and u2 are the servo pump motor speed control voltage and the servo valve core displacement control voltage of the valve control circuit, respectively; x1 is the right wheel rotation angle of the controlled vehicle; θ is an uncertain parameter; and d is the disturbance.

[0072] The controller of the pump-valve parallel control system uses the steering system operating condition values ​​fed back in real time by sensors, and outputs control signals u1 and u2 according to the controller algorithm to control the steering system.

[0073] The controller operates according to an adaptive robust control law, and its control flow is as follows: Figure 2 As shown;

[0074] The adaptive robust control law of the steering system is expressed by the following formula:

[0075] Formula 2;

[0076] Among them, u m For feedback control law, u a For adaptive control law, u s The robust control law is expressed in the following formula:

[0077] Formula 3;

[0078] in, The adaptive value of the uncertain parameter θ is calculated using the following formula:

[0079] Formula 4;

[0080] e1, e2, and e3 are the deviations of state values ​​x1, x2, and x3, respectively, calculated using the following formula:

[0081] Formula 5;

[0082] x 1d x 2d x 3d Let x1, x2, and x3 be the expected values ​​of the states, respectively. The calculation formula is as follows:

[0083] Formula 6;

[0084] k1, k2, k3, and ρ are controller gains, all of which are greater than zero. Given the strong dependence of the gain on the operating conditions, the proposed method achieves condition-driven gain self-tuning by designing an improved particle swarm optimization algorithm that integrates adaptive inertial weights and perturbation strategies.

[0085] The controller gain is self-tuned by using the particle swarm optimization algorithm to drive the steering system's operating conditions. The method is as follows: by traversing a typical set of operating conditions, the global optimization of the controller gain is completed efficiently, and the optimized set of the best gain is embedded into the controller.

[0086] The improved particle swarm optimization algorithm process is as follows: Figure 3 As shown. The specific process includes the following steps:

[0087] Step 1: Obtain the gain parameters k1, k2, k3, and ρ when the controller executes the adaptive robust control law, perform gain self-tuning, and set appropriate value ranges for each parameter;

[0088] Step 2: Set the number of particles and randomly select the initial position of the particles within the range of values.

[0089] Step 3: Perform simulation based on the gain parameters represented by the particle positions;

[0090] Step 4: Sort the control performance of each particle according to the fitness function value, and update the historical best position of the individual particles and the historical best position of the group.

[0091] Step 5: Determine whether most particles have converged to the vicinity of the group's historical optimal position or have not converged after too long an iteration; if they have converged to the optimal position or have failed to converge, output the result.

[0092] Step 6: If convergence is not yet complete, calculate the particle's velocity in each dimension representing different controller gains using the velocity calculation formula, based on parameters such as the individual particle's historical best value and the particle swarm's historical best value. Since the inertia weight and learning factor used for velocity calculation are fixed in the standard PSO algorithm, the algorithm's convergence speed is slow in the early stages and prone to getting trapped in local optima later. An improved PSO algorithm dynamically updates the inertia weight and learning factor, allowing the inertia weight and individual learning factor to gradually decrease as the particle swarm iterates and converges, while the swarm learning factor gradually increases. This optimization aims to achieve stronger early-stage global search capabilities and better late-stage local convergence capabilities.

[0093] Step 7: Update the particle position based on the individual particle position and velocity.

[0094] In step 3, the fitness function value is calculated based on the RMSE value and response time from the simulation results. This value characterizes the control accuracy and response time performance of the gain at that point; the smaller the value, the better the control performance. The specific formula is as follows:

[0095] Formula 7;

[0096] Where T is the total time for steering control; e1 is the control error; k s and t s The response time proportional coefficient and steering system response time are calculated using different methods and values ​​depending on the control signal.

[0097] Step 4: Sort the control performance of each particle according to the fitness function value, and update the individual particle's historical best position and the group's historical best position.

[0098] In step 6, the speed calculation formula used for optimization is:

[0099] Formula 8;

[0100] Where v is the particle velocity; i represents the dimension; n represents the number of iterations; gb and pb are the best historical position of the particle swarm and the best historical position of the individual particle, respectively; x is the particle position; r1 and r2 are random numbers from 0 to 1, representing the randomness of particle motion; ω, a1, and a2 are the inertia weight, the swarm learning factor, and the individual learning factor, respectively, and their variation with the number of iterations is as follows:

[0101] Formula 9;

[0102] Where ω0, a 10 a 20 Let ω, a1, and a2 be the initial values; k ω k a1 k a2This represents the corresponding coefficient of variation.

[0103] In step 7, the formulas for updating the particle position based on the individual particle position and velocity are as follows:

[0104] Formula 10;

[0105] Where, x in v represents the particle's current position in dimension i; in Let x be the particle's current velocity in dimension i; i (n+1） Update the position of the particle in the i-th dimension.

[0106] In step 7, since the particle swarm is prone to getting trapped in local optima in the standard PSO algorithm, the improved PSO introduces a perturbation strategy. When updating the particle position, there is a small probability that the particle will be perturbed and its position will be randomly changed. This strategy is used to effectively avoid getting trapped in local optima. After the particle position is updated, the fitness function value is recalculated.

[0107] Figure 4 The simulation results shown demonstrate the control performance of the gain self-tuning adaptive robust controller based on improved particle swarm optimization and the traditional PID controller under disturbance conditions, indicating that the controller designed in this example has good performance in both robustness and control accuracy.

Claims

1. A gain-self-tuning adaptive robust control method for a pump-valve parallel steering system, characterized in that: The mechanical part of the pump-valve parallel steering system includes a trapezoidal steering mechanism connected to the wheels. The control end of the trapezoidal steering mechanism is controlled by two steering assist cylinders of the hydraulic part of the pump-valve parallel steering system to drive the wheels to steer. The hydraulic control method of the hydraulic part includes a pump-valve parallel control system composed of a pump control circuit and a valve control circuit. The pump-valve parallel control system controls the pump control circuit and the valve control circuit with a gain self-tuning adaptive robust strategy that matches the vehicle steering system. When the pump control circuit supplies oil on demand by adjusting the motor speed of the servo pump, it controls the direction of the reversing valve to control the steering direction. When the valve control circuit supplies oil by the fixed displacement pump, it controls the flow and direction with a servo valve. The pressure oil of the pump control circuit and the valve control circuit flows through the parallel valve block to the two steering assist cylinders of the trapezoidal steering mechanism and is automatically coupled and distributed to jointly control the trapezoidal steering mechanism. When used in hybrid vehicles, the pump control circuit of the parallel control system is driven by a servo motor to drive the hydraulic pump, while the hydraulic pump of the valve control circuit is driven by an electric motor or an internal combustion engine.

2. The gain self-tuning adaptive robust control method for a pump-valve parallel steering system according to claim 1, characterized in that: The trapezoidal steering mechanism is a single-degree-of-freedom hinged four-bar mechanism. The trapezoidal steering mechanism consists of a fixed axle frame, left and right steering angle members with one degree of rotational freedom mounted on the axle frame, steering wheels mounted on them, and a long shaft connecting the left and right angles. The trapezoidal steering mechanism is a single-degree-of-freedom hinged four-bar linkage. The pump-valve parallel control system includes a parallel cylinder assembly formed by a left steering assist cylinder and a right steering assist cylinder. The parallel cylinder assembly serves as the actuator of the hydraulic system. Its two ends are connected to the axle frame and the corner piece by hinges respectively. The extension rod movement of the parallel cylinder assembly drives the trapezoidal steering mechanism to perform an approximate Ackerman steering motion to drive the wheels to turn. The wheel is mounted on the steering angle component and rotates with it. The wheel rotation angle is measured by an angular displacement sensor installed at the pivot of the steering angle component.

3. The gain self-tuning adaptive robust control method for a pump-valve parallel steering system according to claim 1, characterized in that: The gain self-tuning adaptive robust strategy is a particle swarm optimization (PSO) algorithm that includes adaptive inertia weights and a perturbation strategy. When used to optimize the controller gain, the mathematical model of the parallel pump-valve steering system is expressed by the following formula: Official 1; Where u1 and u2 are the servo pump motor speed control voltage and servo valve core displacement control voltage of the pump control loop and the valve core displacement control voltage of the valve control loop, respectively; x1 is the wheel angle of the controlled vehicle; θ is an uncertain parameter; and d is the disturbance. The controller of the pump-valve parallel control system uses the steering system operating condition values ​​fed back in real time by sensors, and outputs control signals u1 and u2 according to the controller algorithm to control the steering system.

4. The gain self-tuning adaptive robust control method for a pump-valve parallel steering system according to claim 3, characterized in that: The controller operates according to an adaptive robust control law; The adaptive robust control law of the steering system is expressed by the following formula: Official 2; Among them, u m For feedback control law, u a For adaptive control law, u s The robust control law is expressed in the following formula: Official 3; in, The adaptive value of the uncertain parameter θ is calculated using the following formula: Official 4; e1, e2, and e3 are the deviations of state values ​​x1, x2, and x3, respectively, calculated using the following formula: Official 5; x 1d x 2d x 3d Let x1, x2, and x3 be the expected values ​​of the states, respectively. The calculation formula is as follows: Official 6; k1, k2, k3, and ρ are the controller gains and are all greater than zero.

5. The gain self-tuning adaptive robust control method for a pump-valve parallel steering system according to claim 4, characterized in that: The controller gain is self-tuned by using the particle swarm optimization algorithm to drive the steering system's operating conditions. The method is as follows: the controller gain is globally optimized by traversing a typical set of operating conditions, and the optimized set of the best gain is embedded into the controller.

6. The gain self-tuning adaptive robust control method for a pump-valve parallel steering system according to claim 5, characterized in that: The specific process of the particle swarm optimization algorithm includes the following steps: Step 1: Obtain the gain parameters k1, k2, k3, and ρ when the controller executes the adaptive robust control law, perform gain self-tuning, and set the value range for each parameter; Step 2: Set the number of particles, and randomly select the initial position of the particles within the range of values; Step 3: Perform simulation based on the gain parameters represented by the particle positions; Step 4: Sort the control performance of each particle according to the fitness function value, and update the historical best position of the individual particles and the historical best position of the group. Step 5: Determine whether most particles have converged to the vicinity of the group's historical optimal position or have not converged after too long an iteration; if they have converged to the optimal position or have failed to converge, output the result. Step 6: If the convergence is not yet complete, calculate the particle's velocity in each dimension representing different controller gains using the velocity calculation formula based on parameters such as the individual particle's historical best value and the particle swarm's historical best value. Dynamically update the inertia weight and learning factor, so that the inertia weight and individual learning factor gradually decrease as the particle swarm iterates and converges, while the swarm learning factor gradually increases, in order to obtain strong early global search capability and late local convergence capability through optimization. Step 7: Update the particle position based on the individual particle position and velocity.

7. The gain self-tuning adaptive robust control method for a pump-valve parallel steering system according to claim 6, characterized in that: In step 3, the fitness function value is calculated based on the RMSE value and response time from the simulation results. This value is used to characterize the control accuracy and response time performance of the gain. The specific formula is as follows: Official 7; Where T is the total time for steering control; e1 is the control error; k s and t s The response time proportional coefficient and steering system response time are calculated using different methods and values ​​depending on the type of control signal. Step 4: Sort the control performance of each particle according to the fitness function value, and update the individual particle's historical best position and the group's historical best position.

8. The gain self-tuning adaptive robust control method for a pump-valve parallel steering system according to claim 7, characterized in that: In step 6, the speed calculation formula used for optimization is: Official 8; Where v is the particle velocity; i represents the dimension; n represents the number of iterations; gb and pb are the best historical position of the particle swarm and the best historical position of the individual particle, respectively; x is the particle position; r1 and r2 are random numbers from 0 to 1, representing the randomness of particle motion; ω, a1, and a2 are the inertia weight, the swarm learning factor, and the individual learning factor, respectively, and their variation with the number of iterations is as follows: Official 9; Where ω0, a 10 a 20 Let ω, a1, and a2 be the initial values; k ω k a1 k a2 This represents the corresponding coefficient of variation.

9. The gain self-tuning adaptive robust control method for a pump-valve parallel steering system according to claim 8, characterized in that: In step 7, the formulas for updating the particle position based on the individual particle position and velocity are as follows: Official 10; Where, x in v represents the particle's current position in dimension i; in Let x be the particle's current velocity in dimension i; i (n+1） Update the position of the particle in the i-th dimension.

10. The gain self-tuning adaptive robust control method for a pump-valve parallel steering system according to claim 8, characterized in that: In step 7, a perturbation strategy is introduced. When updating the particle position, the particle is perturbed with a small probability and its position is randomly changed to avoid getting trapped in a local optimum. After the particle position is changed and updated, the fitness function value is recalculated.

Images