An unmanned tracked vehicle model predictive contour control method based on fusion modeling

Abstract

The application discloses a model predictive contour control method for unmanned tracked vehicle based on fusion modeling, and relates to the technical field of tracked vehicle control methods.The method comprises the following steps: connecting an ICR kinematic model and a neural network model in series, using the ICR model to provide physical constraints and basic prediction, and using the neural network model to fit nonlinear dynamic residual to construct a hybrid model; based on the hybrid model, constructing a model predictive contour control (MPCC) algorithm, optimizing path parameters, minimizing contour error and lag error, and realizing high-precision tracking of the trajectory of the off-road unmanned tracked vehicle.The method has the advantages of simple tracking control and high control precision.

Description

Technical Field

[0001] This invention relates to the field of tracked vehicle control methods, and in particular to a model prediction contour control method for unmanned tracked vehicles based on fusion modeling. Background Technology

[0002] Off-road unmanned tracked vehicles, due to their large ground contact area, possess superior mobility on soft and rugged terrain, and are widely used in agriculture, military, disaster relief, and other fields. Establishing an accurate and computationally efficient vehicle model is a core challenge in achieving autonomous driving or path-tracking control for tracked vehicles.

[0003] Currently, modeling methods for off-road unmanned tracked vehicles are mainly divided into two categories: The first type is the kinematic model, which commonly uses the instantaneous center of rotation (ICR) model. This model assumes a specific geometric relationship between the vehicle and the ground. Its advantages are simple structure, extremely fast computation speed, and suitability for real-time control; its disadvantages are that it ignores complex track-ground interactions (such as slippage and drift), and its prediction accuracy is poor at high speeds or in complex terrain, failing to accurately describe vehicle dynamics.

[0004] The second type is the dynamic model, based on the Newton-Euler equations and considering soil mechanical properties. Its advantages are that it can describe the force and motion state of the vehicle well and has high accuracy; however, its disadvantages are that the model parameters are extremely complex (involving soil shear modulus, subsidence coefficient, etc.), parameter identification is difficult, and the solution of differential equations is time-consuming, making it difficult to meet the millisecond-level real-time calculation requirements of the controller.

[0005] Existing single kinematic models lack sufficient accuracy, while dynamic models have excessive computational load, making it difficult for tracked vehicles to achieve both high accuracy and high real-time performance in path tracking control on complex road surfaces. Summary of the Invention

[0006] The technical problem to be solved by the present invention is how to provide a model prediction contour control method for unmanned tracked vehicles based on fusion modeling that is simple to track and control and has high control accuracy.

[0007] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a method for predicting the contour of an unmanned tracked vehicle model based on fusion modeling, the method comprising the following steps: S1, Constructing a hybrid model: The ICR kinematic model and the neural network model are connected in series. The ICR model provides physical constraints and basic predictions, and the neural network model is used to fit the nonlinear dynamic residuals to construct a hybrid model. S2, MPCC control application: Based on the hybrid model, a model prediction contour control (MPCC) algorithm is constructed. By optimizing the path parameters, the contour error and hysteresis error are minimized simultaneously to achieve the tracking of the trajectory of the off-road unmanned tracked vehicle.

[0008] The beneficial effects of adopting the above technical solution are as follows: 1) High computational speed: Compared with complex dynamic differential equation models, the ICR model is mainly composed of algebraic equations, which is extremely fast; the neural network is only used to fit the residuals, has a small scale, and has a fast inference speed. This combination allows the model prediction step to be completed in milliseconds, meeting the real-time requirements of MPCC.

[0009] 2) High prediction accuracy: Traditional ICR models cannot describe track slippage and ground deformation. This solution uses a neural network to learn the difference between real vehicle data and the ICR model, effectively capturing the complex nonlinear characteristics of vehicle-ground interaction and significantly improving the prediction accuracy of the model under different terrains.

[0010] 3) Excellent path tracking performance: Compared to traditional path tracking (which only focuses on lateral error), MPCC dynamically adjusts path parameters. This allows vehicles to flexibly adjust their speed along the path while meeting contour accuracy requirements, which not only solves the problem of "keeping up with the target" but also optimizes driving efficiency. Attached Figure Description

[0011] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0012] Figure 1 This is the main flowchart of the method described in Embodiment 1 of the present invention; Figure 2 This is a flowchart of the method for constructing a hybrid model as described in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of the ICR kinematic model in the method described in Embodiment 1 of the present invention; Figure 4 This is a schematic diagram of the structure of a computer device according to Embodiment 2 of the present invention. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the spirit of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0014] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0015] Example 1: Overall, such as Figure 1 As shown in the figure, this invention discloses a method for predicting the contour of an unmanned tracked vehicle model based on fusion modeling. The method includes the following steps: S1, Constructing a hybrid model: The ICR kinematic model and the neural network model are connected in series. The ICR model provides physical constraints and basic predictions, and the neural network model is used to fit the nonlinear dynamic residuals to construct a hybrid model. S2, MPCC control application: Based on the hybrid model, a model prediction contour control (MPCC) algorithm is constructed. By optimizing the path parameters and minimizing the contour error and hysteresis error, high-precision tracking of the trajectory of the off-road unmanned tracked vehicle is achieved.

[0016] The steps described above will be explained in detail below: Specifically, the construction of the hybrid model in the method is achieved through steps 1-6 (flowchart shown below). Figure 2 As shown), the MPCC control application part is implemented through steps 7-8; Step 1: The remote-controlled tracked vehicle travels under different working conditions (different speeds, turning radii), and the vehicle's state quantity sequence is measured and collected in real time through the onboard inertial navigation unit. Simultaneously, the corresponding left and right motor speed control sequence is recorded and acquired. Among them, state variables , respectively representing the abscissa, ordinate, and heading angle in the geodetic coordinate system, and the longitudinal velocity, lateral velocity, and yaw rate of the vehicle in the vehicle coordinate system; control variables , representing the rotational speeds of the left and right track drive motors, respectively, in rpm. The collected state variable sequences... and control quantity sequence Perform filtering and timestamp alignment to construct the original dataset. .in, and They represent the first state quantity at time 1 and control quantity Training dataset A total of The state quantity at time 1 and control quantity constitute.

[0017] Step 2: Construct a general ICR kinematic model The diagram is as follows Figure 3 As shown, where The parameters to be identified include the transmission ratio from motor speed to track linear speed. The geometric position parameters of the ICR in the vehicle coordinate system. The formula expression for the ICR kinematic model is: ; in, and These represent the linear velocity of the left and right tracks, and the speed control values ​​of the left and right motors, respectively. and The transmission ratios are all ; , and These are the geometric coordinates of the instantaneous turning centers of the vehicle body, left track, and right track in the vehicle body coordinate system, respectively, and are parameters of the ICR kinematic model. Only used in , and ; This represents the time interval between two adjacent moments.

[0018] Step 3: Apply the ICR kinematic model obtained in Step 2 Construct the optimization objective function Its formula expression is: ; For optimizing the objective function Solving parameters using a nonlinear least squares algorithm This makes the dataset obtained by the ICR model in step 1... The optimal ICR model parameters are obtained by minimizing the prediction error. ,Right now Optimal ICR model parameters Substituting these values ​​into the ICR model above, we obtain the optimal ICR model, whose formula is as follows: ; Step 4: Transfer the dataset obtained in Step 1 In and Substitute the optimal ICR model obtained in step 3 Calculate the theoretical state variables at the next time step. For theoretical state quantities Compared with the actual vehicle status quantity By taking the difference, we obtain the residual value. This residual represents the dynamic characteristics (ingress slip, rotation, terrain drag, etc.) not captured by the ICR model. The obtained residual... Compared with the dataset obtained in step 1 At the same time and Combine them to obtain the neural network training dataset. .

[0019] Step 5: Construct a fully connected neural network, with the current state variable as the input layer. and control quantity The output layer is the predicted residual value for the next time step. Set the network training loss function to mean squared error (MSE), based on the neural network training dataset obtained in step 4. The network weights are trained using the backpropagation algorithm. After training, a neural network residual prediction model is obtained. .

[0020] Step 6: Apply the optimal ICR model obtained in Step 3. Compared with the neural network residual prediction model obtained in step 5 By connecting the series of data, the final system prediction model is obtained. Its formula expression is: ; The model incorporates both the physical and geometric constraints of ICR and the correction of nonlinear dynamic errors by neural networks.

[0021] Step 7: Introduce path variables Parameterize the reference path as ,in, The length of the curved path from the starting point to the specified point. , and The reference paths are respectively in The x-coordinate, y-coordinate, and heading angle at the location. Define the contour error. This represents the normal distance from the vehicle's position to the reference path, and defines the hysteresis error. This represents the longitudinal distance between the vehicle and the reference point along the path. The current vehicle state is known. In the MPCC prediction time domain Within this framework, the MPCC optimization problem is constructed, requiring the minimization of the cost function. Its formula expression is: ; in, Path parameters The speed of travel along the path, , , and These are the weighting coefficients for contour error, hysteresis error, control variable change, and forward speed, respectively.

[0022] Solving the cost function When the minimum value is reached, the system prediction model obtained in step 6 must also be satisfied. Constraints, i.e. The above optimization problem is solved using a nonlinear programming solver to obtain the optimal control sequence. And the optimal control quantity of the left and right motor speeds at the current moment. .

[0023] Step 8: Apply the optimal control quantity obtained in Step 7 The data is sent to the left and right track motors. After the vehicle moves, the onboard sensors collect new state data again. Return to step 8 and enter the next control cycle.

[0024] Example 2 In one exemplary embodiment, the present invention also provides a computer device, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 4 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements the unmanned tracked vehicle model prediction contour control method based on fusion modeling described in Embodiment 1.

[0025] Those skilled in the art will understand that Figure 4The structures shown are merely block diagrams of some structures related to the present application and do not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0026] In one exemplary embodiment, the present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0027] In one exemplary embodiment, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0028] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0029] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0030] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units, etc., and are not limited to these.

[0031] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0032] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for predictive contour control of an unmanned tracked vehicle model based on fusion modeling, characterized in that, The method includes the following steps: S1, Constructing a hybrid model: The ICR kinematic model and the neural network model are connected in series. The ICR model provides physical constraints and basic predictions, and the neural network model is used to fit the nonlinear dynamic residuals to construct a hybrid model. S2, MPCC control application: Based on the hybrid model, a model prediction contour control (MPCC) algorithm is constructed. By optimizing the path parameters, the contour error and hysteresis error are minimized simultaneously to achieve the tracking of the trajectory of the off-road unmanned tracked vehicle. The construction of the hybrid model includes the following steps: 1) Construct the original dataset; 2) Construct the ICR kinematic model; 3) Construct an optimization objective function for the obtained ICR kinematic model; 4) Obtain the dataset Substitute the obtained optimal ICR model into the equation and calculate the theoretical state variables for the next time step. 5) Construct a fully connected neural network with the current state and control variables as inputs and the residual prediction value for the next time step as output. 6) Connect the obtained optimal ICR model with the obtained neural network residual prediction model to obtain the final system prediction hybrid model; The method for constructing an optimization objective function from the obtained ICR kinematic model includes the following steps: The obtained ICR kinematic model Construct the optimization objective function Its formula expression is: ； For optimizing the objective function Solving parameters using a nonlinear least squares algorithm This makes the dataset obtained by the ICR model... The optimal ICR model parameters are obtained by minimizing the prediction error. : ； Optimal ICR model parameters Substituting these values ​​into the ICR model above, we obtain the optimal ICR model, whose formula is as follows: ； The obtained dataset The method for calculating the theoretical next-time state variables by substituting the optimal ICR model obtained from the previous step includes the following steps: The obtained dataset In and The optimal ICR model obtained by substituting Calculate the theoretical state variables at the next time step. ; for theoretical state quantities Compared with the actual vehicle status quantity Taking the difference, we get the residual as follows: The residual represents the dynamic characteristics not captured by the ICR model; the obtained residual With the obtained dataset At the same time and Combine them to obtain the neural network training dataset. ; The method for constructing a fully connected neural network, with the current state and control variables as inputs and the residual prediction value for the next time step as output, includes the following steps: Construct a fully connected neural network, with the current state as the input layer. and control quantity The output layer is the predicted residual value for the next time step. Set the network training loss function to mean squared error (MSE), and then use the obtained neural network training dataset. The network weights are trained using the backpropagation algorithm. After training, a neural network residual prediction model is obtained. ; MPCC control applications: Introducing path variables Parameterize the reference path as ,in, The length of the curved path from the starting point to the specified point. , and The reference paths are respectively in The x-coordinate, y-coordinate, and heading angle of the location; Define contour error Hysteresis error is defined as the normal distance from the vehicle's position to the reference path. This indicates the longitudinal distance between the vehicle and the reference point along the path; Given the current vehicle state quantity In the MPCC prediction time domain Internally, we construct the MPCC optimization problem and minimize the cost function. Its formula expression is: ； in, Path parameters The speed of travel along the path, , , and These are the weighting coefficients for contour error, hysteresis error, control variable change, and forward speed, respectively. Solving the cost function When the minimum value is reached, the system prediction model is satisfied. Constraints, i.e. The above optimization problem is solved using a nonlinear programming solver to obtain the optimal control sequence. And the optimal control quantity of the left and right motor speeds at the current moment. .

2. The method for predictive contour control of unmanned tracked vehicle models based on fusion modeling as described in claim 1, characterized in that, The method for constructing the original dataset includes the following steps: The remote-controlled tracked vehicle operates under different working conditions, and the vehicle's state sequence is measured and collected in real time through the onboard inertial navigation unit. Simultaneously, the corresponding left and right motor speed control sequence is recorded and acquired. Among them, state variables , respectively representing the abscissa, ordinate, and heading angle in the geodetic coordinate system, and the longitudinal velocity, lateral velocity, and yaw rate of the vehicle in the vehicle coordinate system; control variables , representing the rotational speeds of the left and right track drive motors, respectively; for the collected state variable sequences and control quantity sequence Perform filtering and timestamp alignment to construct the original dataset. ;in, and They represent the first state quantity at time 1 and control quantity Training dataset A total of The state quantity at time 1 and control quantity constitute.

3. The unmanned tracked vehicle model prediction contour control method based on fusion modeling as described in claim 2, characterized in that, The method for constructing an ICR kinematic model includes the following steps: Constructing an ICR kinematic model ,in The parameters to be identified include the transmission ratio from motor speed to track linear speed. The geometric position parameters of the ICR in the vehicle coordinate system; the formula expression of the ICR kinematic model is: ； in, and These represent the linear velocity of the left and right tracks, and the speed control values ​​of the left and right motors, respectively. and The transmission ratios are all ; , and These are the geometric coordinates of the instantaneous turning centers of the vehicle body, left track, and right track in the vehicle body coordinate system, respectively. This represents the time interval between two adjacent moments.

4. The unmanned tracked vehicle model prediction contour control method based on fusion modeling as described in claim 1, characterized in that, The method of concatenating the obtained optimal ICR model with the obtained neural network residual prediction model to obtain the final system prediction hybrid model includes the following steps: The obtained optimal ICR model With the obtained neural network residual prediction model By connecting the series of data, the final system prediction model is obtained. Its formula expression is: 。 5. The unmanned tracked vehicle model prediction contour control method based on fusion modeling as described in claim 1, characterized in that, The optimal control quantity obtained by the step The data is sent to the left and right track motors. After the vehicle moves, the onboard sensors collect new state data again. Then, proceed to the next control cycle.

Images