Five-axis transfer robot energy-saving torque control method based on dimension reduction display MPC

Abstract

The invention belongs to the technical field of torque control, and particularly relates to a five-axis transfer robot energy-saving torque control method based on dimension reduction display MPC. The method comprises the steps that current vehicle state information is collected in real time; on the basis of the current vehicle state information, an optimal torque distribution scheme is obtained by quickly looking up a table from a pre-generated explicit control strategy, and an optimal torque instruction is obtained; and the obtained optimal torque instruction is transmitted to a bottom driver, and each wheel motor is controlled to execute a driving control task under energy consumption optimization.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of torque control, and particularly relates to a five-axis transfer robot energy-saving torque control method based on dimension reduction display MPC. BACKGROUND

[0002] With the rapid development of intelligent ports and industrial logistics systems, transfer robots have been widely used in complex environments such as closed sites, high-density loading and unloading, and narrow passage passing, for completing high-precision and high-efficiency cargo carrying tasks. In order to meet the dynamic control requirements under complex paths and heavy load working conditions, some high-performance transfer robots adopt a multi-wheel independent drive structure to improve the mobility, load capacity and terrain adaptability of the whole machine. However, in actual operation, the robot often faces working conditions such as slope passing, frequent starting and stopping, and complex paths, which puts higher requirements on the energy consumption control, dynamic response and execution coordination of the drive system.

[0003] The existing drive control method still has obvious deficiencies in energy consumption optimization. On the one hand, the traditional control strategy mostly adopts a fixed proportion or average distribution torque scheduling method, which fails to fully consider the efficiency differences of each drive motor under different operating states, resulting in low overall energy utilization rate. On the other hand, although the control method based on online optimization has good adjustment performance in theory, it has heavy computational burden and slow response speed in high-dimensional multi-wheel systems, which makes it difficult to meet the real-time requirements under high-frequency control period, affecting the path tracking accuracy and system stability. SUMMARY

[0004] In view of the above technical problems, the application provides a transfer robot energy-saving torque vector control method based on explicit model predictive control, in order to solve the problems of low energy consumption control accuracy, poor real-time performance, insufficient response ability to slope and load changes, heavy controller calculation burden, and rigid and rigid energy efficiency scheduling strategy of the existing multi-wheel drive system.

[0005] The technical solution of the application to solve the above technical problems is as follows: The application provides a five-axis transfer robot energy-saving torque control method based on dimension reduction display MPC, comprising the following steps: Step 100: Real-time acquisition of current vehicle state information; Step 200: Based on the current vehicle state information, the optimal torque distribution scheme is obtained from the pre-generated explicit control strategy by fast table lookup, and the optimal torque instruction is obtained; Step 300: The obtained optimal torque instruction is transmitted to the bottom drive to control each wheel motor to execute the drive control task under energy consumption optimization.

[0006] Furthermore, based on the current vehicle state information, before quickly looking up the optimal torque distribution solution from the pre-generated explicit control strategy, the following steps are also included: The five axes of the five-axis transfer robot are simplified into two three-degree-of-freedom dynamic models and one two-degree-of-freedom dynamic model, and the vehicle dynamics state equation is established, with longitudinal velocity, lateral velocity and yaw angular velocity as the core vehicle state variables.

[0007] Furthermore, based on the current vehicle state information, the optimal torque distribution scheme is quickly obtained from the pre-generated explicit control strategy by looking up the table, including: Perform state dimensionality reduction processing on the original high-dimensional vehicle state vector to obtain an equivalent low-dimensional vehicle state vector; A control equation is introduced, in which the control variable is the control torque of each wheel. By applying different control torques to different wheels, the vehicle state information is controlled; The energy consumption per unit time of each wheel motor is constructed and the total energy consumption is derived from it. Based on the equivalent low-dimensional vehicle state vector, the reference vehicle state vector, the control torque and the total energy consumption, a cost function is constructed to achieve a multi-objective unified optimization of path tracking accuracy and energy minimization. Based on the current vehicle state information and control torque, a vehicle discrete dynamics model is constructed to describe the vehicle state information at the next moment; Based on the vehicle discrete dynamics model, a model predictive control problem is constructed, taking into account multiple constraints; An offline approach is used to partition and optimize the model predictive control problem into multiple parameter regions, generate a mapping between the vehicle state information space and the optimal torque control command, and deploy it to the controller in the form of a lookup table. The vehicle status information is obtained in real time and used as the table input to retrieve the optimal torque distribution result.

[0008] Furthermore, the energy consumption per unit time of each wheel motor is constructed and the total energy consumption is obtained based on it, including: Construct the energy consumption model of each wheel motor and its energy consumption per unit time for: ; in, Indicates time k Wheels wheel speed, Indicates time k Wheels The control torque, For the Efficiency function of a motor under different working conditions; According to the above formula, it is defined at time Total energy consumption for: ; Where n represents the number of motors.

[0009] Furthermore, based on the equivalent low-dimensional vehicle state vector, the reference vehicle state vector, the control torque, and the total energy consumption, a cost function is constructed to achieve a multi-objective unified optimization of path tracking accuracy and energy minimization, including: ; Among them, Q and R are the weight matrices of state error and control input respectively; For the moment k The reference vehicle state vector of For the moment k The actual state vector of N is the equivalent low-dimensional vehicle state vector obtained after state dimensionality reduction; p The length of the time range for calculation.

[0010] Furthermore, multiple constraints include upper and lower limits of control input, vehicle state boundaries, tire adhesion limitations, and motor efficiency operating range constraints.

[0011] Furthermore, the vehicle status information includes longitudinal speed, lateral speed, yaw rate, and wheel speed.

[0012] Compared with the prior art, the present invention has the following technical effects: This paper proposes an energy-efficient torque vectoring control method based on explicit model predictive control (eMPC). By constructing an offline dynamics and energy consumption model of the multi-wheel system of a transfer robot, the optimal control law is pre-solved and deployed in the online control system via a lookup table, achieving a highly efficient energy-efficient control strategy without the need for real-time iteration. Considering the high state dimensionality and table lookup complexity inherent in multi-wheel systems, a state dimensionality reduction mechanism is introduced to compress multiple wheel-end variables into a small number of equivalent representative states. While retaining key dynamic characteristics, this method effectively reduces the controller input dimensionality and the number of lookup tables, thereby improving the controller's computational efficiency and deployment adaptability, meeting the requirements for high-performance, energy-saving control under complex operating conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0014] Figure 1A structural schematic diagram of a transport robot of the present application; Figure 2 A general architecture diagram of an energy-saving control system; Figure 3 A three-degree-of-freedom dynamics model; Figure 4 A two-degree-of-freedom dynamics model; Figure 5 A control flowchart of an energy-saving torque vector control method based on eMPC; Figure 6 A schematic diagram of an eMPC lookup table control strategy execution process. DETAILED DESCRIPTION

[0015] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined object of the application, the specific embodiments, structures, features and effects of the technical solutions proposed according to the present application are described in detail below in combination with the drawings and preferred embodiments. The specific features, structures or characteristics in one or more embodiments can be combined in any suitable form. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as understood by those skilled in the art of the technology to which the present application belongs.

[0016] Figure 1 A structural schematic diagram of a transport robot of the present application, as shown in Figure 1 The transport robot adopts a five-axis ten-wheel independent drive structure, in which the ten wheels are independently controlled in torque output by a hub motor. In order to simplify the control modeling process, in combination with the actual structural characteristics of the vehicle: every two axes in the rear four axes are respectively fixedly installed on the front piece and the rear piece, and have similar motion characteristics, so the present application simplifies them into two three-degree-of-freedom dynamics subsystems; the first axis has independent driving and steering functions, and its two wheels cooperatively realize vehicle longitudinal and lateral control, so it can be simplified into a two-degree-of-freedom dynamics model. Through the above model division, the complexity of the overall modeling and control solution of the system is effectively reduced.

[0017] Figure 2 A general architecture diagram of an energy-saving control system, as shown in Figure 2 The five-axis transport robot energy-saving torque control system based on dimension reduction display MPC includes four core functional modules of path planning module, state acquisition module, controller and actuator.

[0018] On the basis of the path planning module generating the target running trajectory of the vehicle, the state acquisition module is used to collect the current vehicle state information in real time, and the vehicle state information includes longitudinal speed , lateral speed , yaw angular velocity , wheel speed , road slope etc. The state variables required by the vehicle are obtained by IMU, wheel speed sensor and other on-board devices, and constitute the current state vector of the vehicle, which is the basis for the subsequent eMPC controller optimization calculation, and finally forms a complete system state vector for the subsequent eMPC optimization calculation.

[0019] A controller is used to obtain the optimal torque distribution scheme from the pre-generated explicit control strategy based on the current vehicle state information, and output the calculated wheel control torque instruction to the underlying actuator.

[0020] The actuator is used to receive the wheel control torque instruction and drive the wheel motor to run cooperatively, so as to realize the energy saving of the vehicle while meeting the requirements of path tracking accuracy and stability. The process is executed in a high-frequency closed-loop manner in real time, ensuring the feasibility, real-time performance and energy efficiency of the control strategy in the multi-wheel drive transfer robot system.

[0021] In an embodiment of the present application, a five-axis transfer robot energy-saving torque control method based on dimension reduction explicit MPC is provided, comprising the following steps: Step 100: Real-time acquisition of current vehicle state information, the state variables required by the vehicle include longitudinal speed, lateral speed, yaw rate, wheel speed, road slope; Step 200: Based on the current vehicle state information, the optimal torque distribution scheme is obtained from the pre-generated explicit control strategy, and the optimal torque instruction is obtained; Step 300: The obtained optimal torque instruction is transmitted to the underlying drive to control the wheel motor to execute the driving control task under the energy consumption optimization.

[0022] The above steps are described in detail as follows: Step 100: Real-time acquisition of current vehicle state information, the vehicle state information includes longitudinal acceleration, lateral acceleration, vehicle yaw rate, and wheel speed.

[0023] The above state information is obtained by IMU, wheel speed sensor and other on-board devices, and constitutes the current system state vector of the vehicle, which is the basis for the subsequent eMPC controller optimization calculation, and finally forms a complete system state vector for the subsequent eMPC controller optimization calculation.

[0024] Step 200: Based on the current vehicle state information, the optimal torque distribution scheme is obtained from the pre-generated explicit control strategy, and the optimal torque instruction is obtained.

[0025] In step 200, the specific process is as follows: Step 210: Simplify the five axes of the five-axis transfer robot into two three-degree-of-freedom dynamic models and one two-degree-of-freedom dynamic model, establish the vehicle dynamics state equation, and use the longitudinal velocity, lateral velocity and yaw angular velocity as the core vehicle state variables to describe the dynamic behavior of the vehicle in complex scenarios.

[0026] Figure 3 is a three-degree-of-freedom dynamic model, such as Figure 3 The five-axis transfer robot model is simplified into a three-degree-of-freedom dynamic model, and the dynamic equation is established based on the three-degree-of-freedom dynamic model: ; in, is the gross vehicle mass, is the air density, is the aerodynamic drag coefficient, is the frontal area of ​​the vehicle, B is the wheelbase, is the longitudinal velocity of the vehicle's center of mass, is the first-order time derivative of the longitudinal velocity of the vehicle's center of mass, is the lateral velocity of the vehicle's center of mass, is the first-order derivative of the lateral velocity of the vehicle's center of mass with respect to time, is the vehicle's yaw rate, is the first-order derivative of the vehicle's yaw rate with respect to time, is the moment of inertia of the vehicle about the vertical axis, and is the distance from the vehicle's center of mass to the front and rear axles, and are the steering angles of the left and right front wheels, are the longitudinal forces of the left front wheel, right front wheel, left rear wheel and right rear wheel respectively, They are the lateral forces of the left front wheel, right front wheel, left rear wheel and right rear wheel respectively.

[0027] Figure 4 is a two-degree-of-freedom dynamic model, such as Figure 4 As shown. The front axis of the five-axis transfer robot is simplified into a two-degree-of-freedom dynamic model, and the dynamic equation is established based on the two-degree-of-freedom dynamic model: ; in, is the steering angle.

[0028] Step 220: Based on the current vehicle state information, quickly look up the optimal torque distribution scheme and the optimal torque command from the pre-generated explicit control strategy.

[0029] Figure 5 This is a schematic diagram of the execution process of the lookup table control strategy, as shown in Figure 5The proposed table-based control strategy uses the reduced-dimensional explicit model predictive control (eMPC) method to achieve dynamic optimal control of the multi-wheel drive transfer robot between path tracking accuracy and energy efficiency. The specific steps are as follows: Step 221: Perform state dimensionality reduction processing on the original high-dimensional vehicle state vector to obtain an equivalent low-dimensional vehicle state vector as the input state of the controller.

[0030] Based on the vehicle state information, the vehicle dynamics state equation is defined as: ; Among them, X(t) represents the original high-dimensional vehicle state vector, a x is the vehicle longitudinal acceleration, a y is the vehicle lateral acceleration, is the vehicle yaw angular velocity, For wheels i Speed, i=1~10.

[0031] By designing a linear mapping matrix T , the original high-dimensional state vector X ( t ) is mapped to an equivalent low-dimensional vehicle state vector X r Here, the average value of the 10 wheel speeds is used as part of the state variable after dimensionality reduction, which simplifies the information expression about the wheel speed. After dimensionality reduction, it becomes: ; ; In the above formula, Indicates wheels i The average value of the rotation speed; T represents the linear mapping matrix.

[0032] Among them, the mapping matrix T is: ; The above equivalent low-dimensional vehicle state vector It is used as the lookup table input of the display controller to perform the control law matching stage.

[0033] Step 222: Introduce a control equation, where the control quantity in the control equation is the control torque of each wheel, and a control torque vector composed of the control torque of each wheel of the vehicle.

[0034] After obtaining the low-dimensional state vector, the control variable needs to be determined to control the vehicle's motion state. The control equation clarifies that the control variable is the control torque of each wheel. By applying different torques to different wheels, the vehicle's steering, acceleration, deceleration and other control objectives can be achieved. This is the key operation variable for achieving vehicle dynamics control. The control equation is: ; In the above formula, T i For wheels i Control torque; It represents the control torque vector composed of the control torques of each wheel of the vehicle at the discrete time point k, which is the control torque for adjusting the vehicle motion state and achieving the control target.

[0035] Step 223: Construct the energy consumption per unit time of each wheel motor and derive the total energy consumption based on it. Based on the vehicle state vector, the reference vehicle state vector, the control torque and the total energy consumption, construct a cost function to achieve multi-objective unified optimization of path tracking accuracy and energy minimization.

[0036] In order to build a unified path tracking and energy consumption optimization goal, we first build an energy consumption model for each wheel motor, and its energy consumption per unit time is for: ; in, Indicates time k Wheels wheel speed, Indicates time k Wheels The control torque, For the Efficiency function of a motor under different working conditions.

[0037] According to the above formula, it is defined at time The total energy consumption is: ; in, n Indicates the number of motors.

[0038] In order to achieve multi-objective unified optimization of path tracking accuracy and energy minimization, the cost function is defined as follows: ; Among them, Q and R are the weight matrices of state error and control input respectively; For the moment k The reference vehicle state vector of For the moment kThe actual state vector of , that is, the equivalent low-dimensional vehicle state vector obtained after state dimensionality reduction; N p The length of the time range for calculation.

[0039] Path tracking accuracy (represented by the state error term) and energy consumption (represented by the control input term) are unified in a single optimization framework. During the optimization process, these two objectives need to be comprehensively considered to find an optimal control strategy that allows the system to track the reference path as accurately as possible while operating at low energy consumption.

[0040] Step 224: Based on the vehicle state and control torque at the current moment, a vehicle discrete dynamics model is constructed to describe the vehicle state at the next moment.

[0041] In order to predict the future state of the vehicle and design and optimize the control strategy, it is necessary to establish a vehicle dynamics model.

[0042] The discrete dynamics model is used to calculate the vehicle state at the next moment based on the current vehicle state and control torque. It describes the discrete changes in the vehicle state over time and serves as the basis for predicting and controlling the vehicle's future state. For example, based on the current longitudinal acceleration, lateral acceleration, yaw rate, average wheel speed, and control torque for each wheel, the discrete dynamics model can infer the values ​​of these state variables at the next moment, thereby enabling dynamic tracking and control of the vehicle's motion.

[0043] The vehicle discrete dynamics model is: ; exist k +1 moment status is k State of the moment and k Moment Jointly determined function f is the state transition function.

[0044] Step 225: Based on the vehicle discrete dynamics model, a model predictive control problem is constructed, taking into account multiple constraints such as energy minimization, path tracking accuracy, and motor operating range.

[0045] A series of physical and performance constraints are introduced, including upper and lower limits of control input, vehicle state boundaries, tire adhesion limitations, and motor efficiency operating range constraints, to ensure that the generated control instructions are executable and energy-efficient.

[0046] ; in, respectively, are the lower and upper limits of the control input, restricting the physical boundary of the motor torque, respectively, are the lower and upper limits of the state variable, constraining the vehicle dynamic state within a safe range; denotes the longitudinal tire force of the th driving wheel at time , corresponding to the driving force or braking force during acceleration or deceleration; is the adhesion limit region of the th tire under current working conditions, usually represented as a friction circle or friction ellipsoid, used to ensure that the tire force does not exceed the ground adhesion capacity; is the acceptable interval of motor operating efficiency, used to avoid entering the low-efficiency zone or overheating zone.

[0047] Step 226: In the control strategy design phase, the model predictive control problem is solved by offline multi-parameter region division and optimization, generating the mapping relationship between the state space and optimal torque control instructions, and deploying it in the controller in the form of a lookup table.

[0048] First, the state space is divided into regions , for each divided region, an explicit control law is derived through complex optimization algorithms. Assuming the state vector is x , the explicit control law in the jth region can be expressed as: ; In the above formula, denotes the optimal control torque vector when the state x belongs to the region ; denotes the state feedback gain matrix corresponding to the region , which determines the weight of the linear relationship between the control instruction and the state vector; is the bias term, used to further adjust the control instruction to better meet the optimal control requirements of the region; denotes the jth state space region.

[0049] Through such processing, an accurate mapping relationship between the state space and the optimal torque control instruction is generated. Subsequently, this mapping relationship is deployed in the controller in the form of a lookup table, forming an offline constructed explicit strategy library.

[0050] In the actual operation of the vehicle, online control is required. First, in the offline constructed explicit strategy library, the current vehicle state region needs to be quickly located. This relies on the real-time provision of vehicle current state information by the state acquisition module, including the vehicle's position, speed, acceleration, yaw angle, and other key state parameters.

[0051] After receiving the current state information, the controller quickly determines the region of the state space where it is located. Once the region of the state is determined, the controller can retrieve the corresponding control law from the explicit policy library.

[0052] At the same time, the controller also combines the reference speed and the desired heading angle information obtained. The reference speed is the ideal speed set in advance according to the driving task and road conditions of the vehicle, and the desired heading angle is the direction angle that the vehicle should travel.

[0053] Based on the current state, reference speed and desired heading angle, the controller directly calculates the optimal torque allocation result. This table lookup process strictly depends on the control strategy database generated by multi-parameter optimization in the offline stage.

[0054] During online operation, the state acquisition module provides the current state , and after the controller determines the region where it is located, the optimal control command is output according to the corresponding explicit control law : ; Through this combination of offline design and online application, the control system can ensure control accuracy while achieving fast response, effectively achieving vehicle dynamics control, so that the vehicle can accurately track the reference trajectory while meeting performance requirements such as energy consumption.

[0055] Step 227: During system operation, real-time acquisition of current vehicle state information is performed as a table lookup input to quickly retrieve the optimal torque allocation result, and the result is issued to each drive motor to achieve high-frequency, low-energy torque coordination control.

[0056] The control signal is sent to the bottom driver, and the corresponding torque is executed by the motor to achieve the unified goal of path accuracy control and energy minimization.

[0057] Step 300: The obtained optimal torque command is transmitted to the bottom driver to control each wheel motor to perform driving control tasks under energy optimization.

[0058] The above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A five-axis transfer robot energy-saving torque control method based on dimensionality reduction display MPC, characterized in that: The following steps are involved: Step 100: Collect current vehicle status information in real time; Step 200: Based on the current vehicle state information, quickly look up the optimal torque distribution scheme from the pre-generated explicit control strategy and obtain the optimal torque command; Step 300: The obtained optimal torque command is transmitted to the underlying drive to control each wheel motor to perform the drive control task under energy consumption optimization.

2. The energy-saving torque control method of a five-axis transfer robot based on dimensionality reduction display MPC according to claim 1 is characterized in that: Based on the current vehicle state information, the optimal torque distribution scheme is quickly obtained from the pre-generated explicit control strategy by looking up the table. The following steps are also included: The five axes of the five-axis transfer robot are simplified into two three-degree-of-freedom dynamic models and one two-degree-of-freedom dynamic model, and the vehicle dynamics state equation is established, with longitudinal velocity, lateral velocity and yaw angular velocity as the core vehicle state variables.

3. The energy-saving torque control method of a five-axis transfer robot based on dimensionality reduction display MPC according to claim 1 is characterized in that: Based on the current vehicle state information, the optimal torque distribution scheme is quickly obtained from the pre-generated explicit control strategy table, including: Perform state dimensionality reduction processing on the original high-dimensional vehicle state vector to obtain an equivalent low-dimensional vehicle state vector; A control equation is introduced, in which the control variable is the control torque of each wheel. By applying different control torques to different wheels, the vehicle state information is controlled; The energy consumption per unit time of each wheel motor is constructed and the total energy consumption is derived from it. Based on the equivalent low-dimensional vehicle state vector, the reference vehicle state vector, the control torque and the total energy consumption, a cost function is constructed to achieve a multi-objective unified optimization of path tracking accuracy and energy minimization. Based on the current vehicle state information and control torque, a vehicle discrete dynamics model is constructed to describe the vehicle state information at the next moment; Based on the vehicle discrete dynamics model, a model predictive control problem is constructed, taking into account multiple constraints; An offline approach is used to partition and optimize the model predictive control problem into multiple parameter regions, generate a mapping between the vehicle state information space and the optimal torque control command, and deploy it to the controller in the form of a lookup table. The vehicle status information is obtained in real time and used as the table input to retrieve the optimal torque distribution result.

4. The energy-saving torque control method of a five-axis transfer robot based on dimensionality reduction display MPC according to claim 3 is characterized in that: Construct the energy consumption per unit time of each wheel motor and use it to derive the total energy consumption, including: Construct the energy consumption model of each wheel motor and its energy consumption per unit time for: ； in, Indicates time k Wheels wheel speed, Indicates time k Wheels The control torque, For the Efficiency function of a motor under different working conditions; According to the above formula, it is defined at time Total energy consumption for: ； Where n represents the number of motors.

5. The energy-saving torque control method of a five-axis transfer robot based on dimensionality reduction display MPC according to claim 4 is characterized in that: Based on the equivalent low-dimensional vehicle state vector, the reference vehicle state vector, the control torque, and the total energy consumption, a cost function J is constructed to achieve a multi-objective unified optimization of path tracking accuracy and energy minimization, including: ； Among them, Q and R are the weight matrices of state error and control input respectively; For the moment k The reference vehicle state vector of For the moment k The actual state vector of N is the equivalent low-dimensional vehicle state vector obtained after state dimensionality reduction; p The length of the time range for calculation.

6. The energy-saving torque control method of a five-axis transfer robot based on dimensionality reduction display MPC according to claim 5 is characterized in that: Multiple constraints include upper and lower limits of control input, vehicle state boundaries, tire adhesion limitations, and motor efficiency operating range constraints.

7. The energy-saving torque control method of a five-axis transfer robot based on dimensionality reduction display MPC according to claim 1 is characterized in that: Vehicle status information includes longitudinal speed, lateral speed, yaw rate, and wheel speed.

Images