A Multi-Objective Real-Time Lower-Level Torque Distribution Method for Overdrive Electric Vehicles

Abstract

This invention discloses a multi-objective real-time lower-level torque allocation method for overdrive electric vehicles, belonging to the field of intelligent electric vehicle motion control and chassis cooperative control technology. First, it acquires the generalized control commands output by the upper-level controller, the real-time vehicle state variables, and the wheel-end torques of the four wheels in the previous control cycle. Based on the vehicle's geometric parameters, a control efficiency matrix is ​​constructed, establishing a linear mapping relationship between the generalized control variables and the wheel-end torques. Using the four-wheel wheel-end torques as decision variables, a multi-objective quadratic optimization model is constructed, incorporating generalized command tracking, tire load rate balancing, drive energy consumption optimization, and torque smoothing. A convex quadratic programming problem is constructed by combining motor torque saturation, torque change rate, and linearized tire adhesion constraints. A hot-start time-limited solution strategy is used to solve the problem, triggering a multi-level fault-tolerant mechanism to output a safe and feasible torque when an anomaly occurs. This invention balances vehicle stability, economy, and system robustness, and is applicable to overdrive electric vehicles with four-wheel independent drive and other configurations.

Description

Technical Field

[0001] This invention relates to the field of motion control and chassis cooperative control technology for intelligent electric vehicles, and in particular to a multi-objective real-time lower-level torque distribution method for overdrive electric vehicles. Background Technology

[0002] Overdrive electric vehicles (typically four-wheel independent drive electric vehicles) have actuator redundancy characteristics, meaning that under the premise of satisfying the generalized control commands such as the total longitudinal force and yaw moment of the vehicle target, there are multiple wheel end torque combinations that can achieve the control target, providing sufficient degrees of freedom for vehicle dynamics optimization.

[0003] In existing technologies, torque distribution is often achieved using fixed-ratio rule allocation, pseudo-inverse methods, or simple single-objective optimization. However, in scenarios such as low-adhesion road surfaces, rapid switching of operating conditions, and limited actuator capabilities (e.g., torque saturation, over-temperature derating, insufficient power), existing methods have significant drawbacks: First, rule allocation cannot dynamically adapt to changes in operating conditions, making it difficult to simultaneously consider tracking accuracy, tire adhesion balance, energy economy, and command smoothness; second, the high complexity of solving nonlinear optimization models cannot meet the real-time requirements of millisecond-level control cycles for onboard controllers; third, the lack of explicit handling of tire adhesion limit constraints can easily lead to premature saturation and instability of a single wheel; fourth, the lack of a robust fault-tolerance and degradation mechanism makes it prone to control interruption or output instability when abnormal solutions occur, failing to meet automotive-grade functional safety requirements.

[0004] Therefore, there is an urgent need for a method for lower-level torque distribution in overdrive electric vehicles that can explicitly handle multi-dimensional physical constraints, take into account the collaborative optimization of multiple performance objectives, meet real-time requirements, and have good fault tolerance. Summary of the Invention

[0005] The purpose of this invention is to propose a multi-objective real-time lower-level torque distribution method for overdrive electric vehicles. This method transforms the torque distribution problem into a standard convex quadratic programming problem. Under the premise of meeting automotive-grade real-time control requirements, it achieves multi-objective coordination of generalized command tracking, tire load balancing, drive energy consumption optimization, and torque smoothing. At the same time, through a hot-start solution strategy and a multi-level fault-tolerant mechanism, it ensures solution efficiency and system robustness, while taking into account the stability and economy of vehicle driving.

[0006] To achieve the above objectives, the present invention provides a multi-objective real-time lower-level torque distribution method for overdrive electric vehicles, comprising the following steps: Step S1: Obtain the vehicle generalized control command, real-time vehicle operating status, and four-wheel wheel torque output values ​​from the upper-level motion controller. Step S2: Based on the vehicle's geometric parameters and the mapping relationship between wheel-end torque and tire longitudinal force, construct a control efficiency matrix and establish a linear mapping relationship between the vehicle's generalized control quantity and the wheel-end torque of the four wheels; Step S3: Using the torque vectors at the wheel ends of the four wheels as decision variables, construct a multi-objective quadratic objective function. The objective function includes a generalized command tracking error term, a tire load rate balancing term, a drive energy consumption optimization term, and a torque smoothing term. Step S4: Estimate the real-time normal load of the four wheels based on the real-time longitudinal and lateral acceleration of the vehicle, and dynamically update the weight coefficient of the tire load rate equilibrium term and the boundary parameters of the tire adhesion constraint based on the normal load. Step S5: Set constraints including motor wheel end torque saturation constraint, torque change rate constraint and tire adhesion constraint, and transform the multi-objective optimization problem into a standard convex quadratic programming problem; Step S6: Use a time-limited solution strategy with hot start to solve the convex quadratic programming problem. If the solution is successful, output the optimal wheel end torque command for all four wheels; if the solution is not feasible or the timeout occurs, trigger the fault tolerance degradation mechanism to output a safe and feasible wheel end torque command.

[0007] Preferably, in step S2, the control performance matrix The expression is: ; in, The effective rolling radius of the tire. The front axle track of the vehicle. This refers to the rear axle track of the vehicle. The linear mapping relationship between the generalized control variables of the vehicle and the wheel-end torques of the four wheels is as follows: ; in, This is the generalized control vector for the vehicle. This is the torque vector at the wheel ends of the four wheels.

[0008] Preferably, in step S3, the expression for the multi-objective quadratic objective function is: ; in, This is the generalized instruction tracking error term, used to minimize the deviation between the actual generalized control quantity and the upper-level generalized control instruction; This is a tire load factor balancing term, used to balance the adhesion utilization rate of the four tires; This is a drive energy consumption optimization item used to reduce the energy consumption of the vehicle's drive system; This is a torque smoothing term used to suppress torque fluctuations between adjacent control cycles.

[0009] Preferably, the generalized instruction tracking error term The calculation formula is as follows: ; in, This is the generalized control command vector output from the upper layer. To track the weight diagonal matrix; Tire load balance item The calculation formula is as follows: ; in, For the first The load factor of each wheel. For the first The wheel-end torque of each wheel; Drive energy consumption optimization items The calculation formula is as follows: ; in, For the first The wheel-end torque of each wheel, For the first The corresponding drive motor speed for each wheel The non-negative energy consumption weighting coefficient related to motor speed is obtained by fitting a motor efficiency map or looking up a table. Torque smoothing term The calculation formula is as follows: ; in, This represents the wheel-end torque vector of the four wheels in the current control cycle. This is the wheel-end torque vector of the four wheels from the previous control cycle. It is a non-negative smoothed weight diagonal matrix.

[0010] Preferably, in step S4, the real-time normal loads of the four wheels are estimated using a quasi-static load transfer model, and the calculation formula is as follows: ; in, For the first Real-time normal load of each wheel For the first Static normal load of each wheel Longitudinal acceleration The resulting longitudinal load transfer amount, lateral acceleration The resulting lateral load transfer; Load factor Real-time normal load of the corresponding wheel It is inversely proportional to the square of the number of digits, and the expression is: ; in, This represents the average value of the normal loads on the four wheels. Based on the weighting coefficient, To prevent the default small positive number with a denominator of zero.

[0011] Preferably, in step S5, the expression for the motor wheel-end torque saturation constraint is: ; in, and The first The maximum and minimum allowable output torque of each wheel drive motor is dynamically determined by the motor's external characteristics, battery state of charge, and motor winding temperature. The expression for the torque change rate constraint is: ; in, The maximum allowable torque change within a single control cycle is determined by the motor response bandwidth. The tire adhesion constraint is a linearized polyhedral approximation of the friction ellipse constraint, used to limit the resultant force of the longitudinal and lateral forces of the wheel from exceeding the tire's usable adhesion limit. ,in For the first The road adhesion coefficient of each wheel.

[0012] Preferably, in step S6, the time-limited solution strategy with hot start uses the optimal or feasible solution of the previous control cycle as the initial value for the quadratic programming solution of the current cycle, adopts the active set method or interior point method as the solver, and sets the maximum number of iterations and the maximum solution time threshold to ensure that the solution process is completed within the preset control cycle.

[0013] Preferably, in step S6, the fault tolerance and degradation mechanism includes two levels of fault tolerance: The first-level fault tolerance is a relaxed solution. When the optimization problem is not feasible, a relaxed variable is introduced to the generalized instruction tracking constraint. Priority is given to satisfying the safety constraints of motor torque saturation and tire adhesion, and a feasible solution is obtained by solving again. The second-level fault tolerance is a rule rollback. When the solution times out or the first-level fault tolerance solution fails, it switches to the preset rule allocation strategy, allocates longitudinal force according to the axle load ratio, allocates yaw moment through differential allocation of left and right wheels, and outputs safe and feasible wheel end torque commands.

[0014] Therefore, the present invention employs the above-described multi-objective real-time lower-level torque distribution method for overdrive electric vehicles, which has the following advantages: (1) Improve vehicle driving stability and safety: By explicitly introducing motor torque saturation, torque change rate and linearized tire adhesion constraints in the optimization model, the output torque command is always within the physical feasible domain of the actuator and the tire, effectively preventing motor overload and tire saturation instability, and greatly improving vehicle handling stability under extreme conditions. (2) Achieve multi-objective global collaborative optimization: Construct a multi-objective quadratic optimization model containing four sub-objectives, which breaks through the limitations of single-objective control. It can dynamically adjust the torque distribution of each wheel according to the real-time working conditions. Under the premise of ensuring the accuracy of generalized command tracking, it takes into account tire load balance, drive energy consumption optimization and torque smoothing, so as to achieve the coordinated improvement of vehicle power, safety and economy. (3) Meets automotive-grade real-time requirements: The complex nonlinear torque distribution problem is transformed into a standard convex quadratic programming problem. Combined with a hot-start solution strategy, the number of iterations and computation time are greatly reduced. By setting thresholds for solution time and number of iterations, the solution process is ensured to be completed within the millisecond-level control cycle of the on-board controller, thus meeting automotive-grade real-time control requirements. (4) High system robustness and fault tolerance: The designed multi-level fault tolerance and degradation mechanism can cope with numerical calculation problems such as infeasibility and timeout. Under extreme conditions such as sensor noise, sudden changes in road surface adhesion, and partial failure of actuators, it can continuously output safe and feasible torque commands, avoid control divergence or system crash, and improve the robustness of the system.

[0015] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0016] Figure 1 This is an overall flowchart of the multi-target real-time lower-level torque distribution method in an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the mapping relationship between the control performance matrix and the generalized instruction in an embodiment of the present invention; Figure 3 This is a schematic diagram of the objective and constraint terms of a multi-objective convex quadratic programming problem in an embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0018] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0019] Example like Figure 1-3 As shown, this embodiment proposes a multi-objective real-time lower-level torque distribution method for overdrive electric vehicles. This embodiment takes a four-wheel independent drive electric vehicle as the application object, and the specific implementation steps are as follows: Step S1: Data Acquisition and Initialization: At the beginning of each control cycle (e.g., 10ms) The lower-level controller performs the following data acquisition operations via the vehicle bus (such as CAN or FlexRay): (1) Obtaining generalized control commands: Receive the generalized control command vector calculated and output by the upper-level motion controller. .in, The total longitudinal driving force of the vehicle. The target yaw moment for the vehicle.

[0020] (2) Obtain vehicle state information: Collect or obtain the current motion state of the vehicle through an estimator, including longitudinal acceleration. lateral acceleration yaw rate and the wheel speed of each wheel .

[0021] (3) Obtain feedback information: Read the actual output torque of the previous control cycle. This is used to calculate the torque change rate and smoothing control term in subsequent steps.

[0022] Step S2: Control Performance Matrix Construction: Based on the vehicle's geometry, establish the mapping relationship between generalized control forces and tire longitudinal forces. Define the decision variable as the four-wheel wheel-end torque vector. subscript These represent the front left, front right, rear left, and rear right wheels, respectively.

[0023] Constructing a control effectiveness matrix This makes the mapping relationship In this embodiment, the performance matrix The specific form is as follows: ; In the formula, The effective rolling radius of the tire; The front axle track. This represents the rear axle track. This matrix linearly maps the drive / braking torque of the four wheels to the total longitudinal force and yaw moment of the entire vehicle.

[0024] Step S3: Construction of a multi-objective quadratic objective function: To balance tracking accuracy, stability, and economy, this embodiment constructs a multi-objective quadratic objective function, such as... Figure 3 As shown, the function contains the following four sub-objectives, expressed as follows: ; The specific definitions and physical meanings of each sub-item are as follows: (1) Generalized instruction tracking error term : The aim is to minimize the deviation between the actual generated generalized force and the upper-level instructions, expressed as follows: ; in, To track the weight matrix. In actual tuning, it is usually... Set to a larger value to prioritize the vehicle's yaw stability.

[0025] (2) Tire load factor balancing item : The aim is to distribute torque according to the vertical load capacity of each wheel to prevent premature saturation of a single wheel. The expression is as follows: ; Among them, weight Designed to be in harmony with tire normal load It is inversely proportional to the square of . This embodiment uses the following dynamic update formula: ; in, For average load, Based on the weighting coefficient, To prevent small positive numbers with a denominator of zero, this mechanism assigns a larger penalty weight to wheels with lighter loads (i.e., those prone to slippage), thereby actively reducing their torque output.

[0026] (3) Drive energy consumption optimization items : The aim is to reduce the total power loss of the system. Based on the motor efficiency map, under the convex optimization framework, it can be approximately represented as a quadratic term related to the speed, as shown in the following expression: ; in It is the energy consumption coefficient that varies with the motor speed.

[0027] (4) Torque smoothing term : The purpose is to suppress high-frequency jitter in control commands and protect the mechanical components of the actuator. The expression is as follows: ; in, This represents the four-wheel torque vector for the current control cycle. This is the four-wheel torque vector from the previous control cycle. This is the smoothing weight matrix.

[0028] Step S4: Real-time Normal Load Estimation and Dynamic Parameter Update: The quasi-static load transfer model is used to estimate the real-time normal loads of the four wheels. The calculation formula is as follows: ; in, For the first Real-time normal load of each wheel For the first Static normal load of each wheel Longitudinal acceleration The resulting longitudinal load transfer amount, lateral acceleration The resulting lateral load transfer; Load factor Real-time normal load of the corresponding wheel It is inversely proportional to the square of the number of digits, and the expression is: ; in, This represents the average value of the normal loads on the four wheels. Based on the weighting coefficient, To prevent the default small positive number with a denominator of zero.

[0029] Step S5: Setting Constraints and Constructing the Convex Quadratic Programming Problem: Set three types of core constraints, transform them into standard linear inequality constraint forms, and construct a standard convex quadratic programming problem in combination with the objective function.

[0030] (1) Motor wheel end torque saturation constraint: Considering the motor's external characteristics and the battery's discharge capacity, upper and lower limits are set as follows: ; Specifically, a dynamic derating factor is introduced when motor overheating or battery SOC is detected. compression and The range.

[0031] (2) Torque change rate constraint: Due to the limitations of the motor's response bandwidth, the torque increment between adjacent cycles must be restricted: ; (3) Tire adhesion constraint (linearization process): To maintain the convexity of the problem, this invention addresses the nonlinear friction circle constraint. Perform a linearized polyhedral approximation. Assume lateral forces. If the longitudinal force constraint is known or can be estimated, then it is transformed into: ; This ensures that the distribution result is always within the tire's physical adhesion limits.

[0032] Step S6: Quadratic Programming Solution and Fault Tolerance Mechanism: Combining the above objective function and constraints, the original problem is transformed into a standard convex quadratic programming (Convex QP) problem: ; This embodiment employs the following strategy for solving and fault tolerance: (1) Hot start: The optimal solution from the previous time step. This serves as the initial guess for the current solution algorithm (such as the effective set method) to significantly reduce the number of iterations.

[0033] (2) Solver configuration: Set the maximum number of iterations (e.g., 50 times) and the maximum solution time threshold (e.g., 5ms) to meet the real-time requirements.

[0034] (3) Multi-level fault tolerance mechanism: Level 1 tolerance (relaxed solution): If the solver returns "no solution", it indicates that the constraints are too strong. In this case, slack variables are introduced. Relax the generalized instruction tracking constraints and prioritize meeting hard physical safety constraints (such as motor saturation).

[0035] Level 2 fault tolerance (rule rollback): If the solution times out or a numerical error occurs, the rule rollback strategy is immediately triggered. At this time, the system switches to a static allocation rule based on axle load ratio, outputting a safe and usable torque command until the optimized solver returns to normal.

[0036] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. 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 still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A multi-objective real-time lower-level torque distribution method for overdrive electric vehicles, characterized in that, Includes the following steps: Step S1: Obtain the vehicle generalized control command, real-time vehicle operating status, and four-wheel end torque output values ​​from the upper-level motion controller. Step S2: Based on the vehicle's geometric parameters and the mapping relationship between wheel-end torque and tire longitudinal force, construct a control efficiency matrix and establish a linear mapping relationship between the vehicle's generalized control quantity and the wheel-end torque of the four wheels; Control effectiveness matrix The expression is: ； in, The effective rolling radius of the tire. The front axle track of the vehicle. This refers to the rear axle track of the vehicle. The linear mapping relationship between the generalized control variables of the vehicle and the wheel-end torques of the four wheels is as follows: ； in, This is the generalized control vector for the vehicle. This represents the torque vector at the wheel ends of all four wheels. Step S3: Using the torque vectors at the wheel ends of the four wheels as decision variables, construct a multi-objective quadratic objective function. The objective function includes a generalized command tracking error term, a tire load rate balancing term, a drive energy consumption optimization term, and a torque smoothing term. The expression for the multi-objective quadratic objective function is: ； in, This is the generalized instruction tracking error term, used to minimize the deviation between the actual generalized control quantity and the upper-level generalized control instruction; This is a tire load factor balancing term, used to balance the adhesion utilization rate of the four tires; This is a drive energy consumption optimization item used to reduce the energy consumption of the vehicle's drive system; This is a torque smoothing term used to suppress torque fluctuations between adjacent control cycles; Generalized instruction trace error term The calculation formula is as follows: ； in, This is the generalized control command vector output from the upper layer. To track the weight diagonal matrix; Tire load balance item The calculation formula is as follows: ； in, For the first The load rate weighting coefficient for each wheel. For the first The wheel-end torque of each wheel; Drive energy consumption optimization items The calculation formula is as follows: ； in, For the first The wheel-end torque of each wheel, For the first The corresponding drive motor speed for each wheel The non-negative energy consumption weighting coefficient related to motor speed is obtained by fitting a motor efficiency map or looking up a table. Torque smoothing term The calculation formula is as follows: ； in, This represents the wheel-end torque vector of the four wheels in the current control cycle. This is the wheel-end torque vector of the four wheels from the previous control cycle. It is a non-negative smoothed weight diagonal matrix; Step S4: Estimate the real-time normal load of the four wheels based on the real-time longitudinal and lateral acceleration of the vehicle, and dynamically update the weight coefficient of the tire load rate equilibrium term and the boundary parameters of the tire adhesion constraint based on the normal load. Step S5: Set constraints including motor wheel end torque saturation constraint, torque change rate constraint and tire adhesion constraint, and transform the multi-objective optimization problem into a standard convex quadratic programming problem; Step S6: Use a time-limited solution strategy with hot start to solve the convex quadratic programming problem. If the solution is successful, output the optimal wheel end torque command for all four wheels; if the solution is not feasible or the timeout occurs, trigger the fault tolerance degradation mechanism to output a safe and feasible wheel end torque command.

2. The multi-objective real-time lower-level torque distribution method for overdrive electric vehicles according to claim 1, characterized in that: In step S4, the real-time normal loads on the four wheels are estimated using a quasi-static load transfer model. The calculation formula is as follows: ； in, For the first Real-time normal load of each wheel For the first Static normal load of each wheel Longitudinal acceleration The resulting longitudinal load transfer amount, Lateral acceleration The resulting lateral load transfer; Load factor Real-time normal load of the corresponding wheel It is inversely proportional to the square of the number of digits, and the expression is: ； in, This represents the average value of the normal loads on the four wheels. Based on the weighting coefficient, To prevent the default small positive number with a denominator of zero.

3. The multi-objective real-time lower-level torque distribution method for overdrive electric vehicles according to claim 1, characterized in that: In step S5, the expression for the motor wheel end torque saturation constraint is: ； in, and The first The maximum and minimum allowable output torque of each wheel drive motor is dynamically determined by the motor's external characteristics, battery state of charge, and motor winding temperature. The expression for the torque change rate constraint is: ； in, The maximum allowable torque change within a single control cycle is determined by the motor response bandwidth. The tire adhesion constraint is a linearized polyhedral approximation of the friction ellipse constraint, used to limit the resultant force of the longitudinal and lateral forces of the wheel from exceeding the tire's usable adhesion limit. ,in For the first The road adhesion coefficient of each wheel.

4. The multi-objective real-time lower-level torque distribution method for overdrive electric vehicles according to claim 1, characterized in that: In step S6, the time-limited solution strategy with hot start uses the optimal or feasible solution of the previous control cycle as the initial value for the quadratic programming solution of the current cycle, adopts the active set method or interior point method as the solver, and sets the maximum number of iterations and the maximum solution time threshold to ensure that the solution process is completed within the preset control cycle.

5. A multi-objective real-time lower-level torque distribution method for overdrive electric vehicles according to claim 1, characterized in that: In step S6, the fault tolerance and degradation mechanism includes two levels of fault tolerance: The first-level fault tolerance is a relaxed solution. When the optimization problem is not feasible, a relaxed variable is introduced to the generalized instruction tracking constraint. Priority is given to satisfying the safety constraints of motor torque saturation and tire adhesion, and a feasible solution is obtained by solving again. The second-level fault tolerance is a rule rollback. When the solution times out or the first-level fault tolerance solution fails, it switches to the preset rule allocation strategy, allocates longitudinal force according to the axle load ratio, allocates yaw moment through differential allocation of left and right wheels, and outputs safe and feasible wheel end torque commands.

Images