Limited time domain unmanned vehicle obstacle collision prevention control method and device

Abstract

The invention provides a limited time domain unmanned vehicle obstacle collision prevention control method and device, and the method comprises the steps: describing the safety of a control system based on forward invariance, and constructing an obstacle function of the control system according to an obstacle based on the safety of the control system; integrating the obstacle function into a cost function of the control system, and converting a collision avoidance control problem of the control system into an optimal control problem based on the cost function; establishing a control system model of unknown parameters based on a neural network function approximation method, and designing a parameter adaptive law by using an integral parallel learning method to identify the unknown parameters in the model; and based on the control system model, a quadratic optimization framework is utilized, and an adaptive iterative learning method is adopted to solve the optimal control input meeting the requirements of the collision avoidance control task. According to the scheme, the method does not depend on an accurate parameter model of the system, the optimization performance and safety of the control system are both considered, and the collision avoidance control problem in a limited time domain can be effectively solved.

Description

Technical Field

[0001] This application relates to the fields of control theory and artificial intelligence technology, and particularly to a method and device for obstacle avoidance control of a finite-horizon driverless vehicle. Background Art

[0002] Intelligent driving technology is regarded as an effective means to reduce traffic accidents, improve traffic efficiency and relieve the driver's burden. It is generally divided into an active safety system, an intelligent driving assistance system and an autonomous driving system. The active safety system plays a crucial role in intelligent vehicle control by perceiving and monitoring the internal and external conditions of the vehicle, identifying potential dangers, and helping to avoid or mitigate collisions through various intervention means. The main methods for solving the obstacle avoidance control problem include model predictive control and safety reinforcement learning, etc. The key advantage of model predictive control is its ability to dynamically adjust control inputs to adapt to environmental changes and complex obstacle avoidance requirements. This method usually divides the obstacle avoidance task into two parts: path planning and trajectory tracking. However, it consumes a large amount of online computing resources, and there may be performance contradictions between hierarchical subtasks. Safety reinforcement learning optimizes the control strategy and numerically solves the Hamilton-Jacobi-Bellman equation to enable the vehicle to strictly comply with safety constraints while pursuing the minimum cost. However, this method is often challenging when faced with a finite-horizon optimal control problem. Summary of the Invention

[0003] This application aims to at least partly solve one of the technical problems in the related art.

[0004] To this end, the first objective of this application is to propose a method for obstacle avoidance control of a finite-horizon driverless vehicle, which does not rely on the accurate parameter model of the system, takes into account the optimization performance and safety of the control system, and can effectively solve the obstacle avoidance control problem within a finite time domain.

[0005] The second objective of this application is to propose a device for obstacle avoidance control of a finite-horizon driverless vehicle.

[0006] The third objective of this application is to propose a computer device.

[0007] To achieve the above object, an embodiment of the first aspect of the present application proposes a method for obstacle avoidance control of a finite-time-domain driverless vehicle, including: describing the safety of the control system based on forward invariance, and constructing an obstacle function of the control system according to the obstacles based on the safety of the control system; integrating the obstacle function into the cost function of the control system, and transforming the obstacle avoidance control problem of the control system into an optimal control problem based on the cost function; establishing a control system model with unknown parameters based on the neural network function approximation method, and designing a parameter adaptive law using the integral parallel learning method to identify the unknown parameters in the model; based on the control system model, using the quadratic optimization framework, and adopting the adaptive iterative learning method to solve the optimal control input that meets the requirements of the obstacle avoidance control task.

[0008] To achieve the above object, an embodiment of the second aspect of the present application proposes a device for obstacle avoidance control of a finite-time-domain driverless vehicle, including: a first system establishment module, configured to describe the safety of the control system based on forward invariance, and construct an obstacle function of the control system according to the obstacles based on the safety of the control system; a second system establishment module, configured to integrate the obstacle function into the cost function of the control system, and transform the obstacle avoidance control problem of the control system into an optimal control problem based on the cost function; a system identification module, configured to establish a control system model with unknown parameters based on the neural network function approximation method, and design a parameter adaptive law using the integral parallel learning method to identify the unknown parameters in the model; an iterative solution module, configured to based on the control system model, use the quadratic optimization framework, and adopt the adaptive iterative learning method to solve the optimal control input that meets the requirements of the obstacle avoidance control task.

[0009] To achieve the above object, an embodiment of the third aspect of the present application proposes a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the computer program, the above method for obstacle avoidance control of a finite-time-domain driverless vehicle is implemented.

[0010] The method for obstacle avoidance control of a finite-time-domain driverless vehicle according to the embodiments of the present application describes the safety of the control system based on forward invariance. On this basis, a corresponding obstacle function is constructed according to the obstacles; the constructed obstacle function is integrated into the cost function, and the obstacle avoidance control problem is transformed into an optimal control problem; a parameter identification model of the unknown system is established based on the neural network function approximation method, and an adaptive law is designed using the integral parallel learning method to online identify the unknown parameters in the model; based on the identification model and the identified parameters, a quadratic optimization framework is used to propose an adaptive iterative learning algorithm to solve the optimal control input that meets the requirements of the obstacle avoidance control task. The present application does not depend on the accurate parameter model of the system, takes into account the optimization performance and safety of the control system, and can effectively solve the obstacle avoidance control problem within a finite time domain.

[0011] Additional aspects and advantages of the present application will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The above-mentioned and / or additional aspects and advantages of the present application will become apparent and be readily understood from the following description of embodiments in conjunction with the accompanying drawings, in which: Figure 1 is a schematic flowchart of a method for obstacle avoidance control of a finite-time domain driverless vehicle provided in Embodiment 1 of the present application; Figure 2 is a schematic diagram of a "bicycle" model for vehicle steering control in Embodiment of the present application; Figure 3 is a design diagram of a driving scenario in Embodiment of the present application; Figure 4 is an iterative graph of a cost function in Embodiment of the present application; Figure 5 is an iterative graph of a vehicle trajectory in Embodiment of the present application; Figure 6 is a comparison graph of the true value and the estimated value of the vehicle state in Embodiment of the present application; Figure 7 is a path schematic diagram of a vehicle under an initial input in Embodiment of the present application; Figure 8 is a path schematic diagram of a vehicle in the 20th iteration in Embodiment of the present application; Figure 9 is a path schematic diagram of a vehicle when the algorithm converges in Embodiment of the present application; Figure 10 is a schematic structural diagram of a device for obstacle avoidance control of a finite-time domain driverless vehicle provided in Embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0013] Embodiments of the present application will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary and are intended to explain the present application, and should not be construed as limiting the present application.

[0014] In actual situations, it is often difficult to obtain an accurate model of a vehicle, and the driving environment is complex and changeable. Utilizing the trajectory data of the vehicle to avoid the dependence on an accurate system model in traditional obstacle avoidance control methods, while considering the unity of planning and control and effectively solving a control strategy that meets the requirements of the obstacle avoidance task, is an effective way to improve driving safety.

[0015] The method and device for obstacle avoidance control of a finite-time domain driverless vehicle according to an embodiment of the present application will be described below with reference to the accompanying drawings.

[0016] Figure 1 It is a schematic flow chart of a method for obstacle avoidance control of a finite-time domain driverless vehicle provided in the first embodiment of the present application.

[0017] As Figure 1 shown, the method for obstacle avoidance control of the finite-time domain driverless vehicle includes the following steps: Step 101, describe the safety of the control system based on forward invariance, and construct an obstacle function of the control system according to the obstacles based on the safety of the control system; In this embodiment, consider a control system model in the following form:

[0018] Among them, is the system state variable, and the initial value is , is the system control input, is the dynamic equation of the unknown system. Assume that and its partial derivatives with respect to the state and the input , satisfy the local Lipschitz continuity condition, which is a common assumption in control theory and guarantees the existence and uniqueness of the solution trajectory .

[0019] The set formed by the control input is defined as a convex set in the following form:

[0020] Assume that for the closed and bounded control input , at the initial value , the solution trajectory of the control system is bounded, which is also a common assumption in control theory and avoids the phenomenon of the system escaping within a finite time. It is described formulaically as: for , , such that ; The obstacle avoidance control problem requires that the state of the control system always remains within the given safety set. The safety set is defined as follows:

[0021]

[0022]

[0023] Among them, is a continuously differentiable function, and the set is the boundary of, and the set is the interior of, and there are and ; If for any initial state , the solution trajectory always remains within the set , then the set is said to have forward invariance. If the system state satisfies the forward invariance condition with respect to the set , then the control system is said to be safe with respect to the set ; Based on the properties of the safety set, the following barrier function can be defined:

[0024] The barrier function can be obtained to satisfy the following properties:

[0025]

[0026] Considering there are obstacles, in order to ensure that the system can avoid collisions, a circular region in the following form is defined as the unsafe region of the -th obstacle:

[0027] Among them, is the center of the -th obstacle, and is the radius of the corresponding unsafe region. For the circular region, the barrier function of the -th obstacle can be specifically expressed as:

[0028] Among them, ; In actual numerical simulations, in order to avoid the situation of infinity, a saturation hyperbolic function can be used to limit the barrier function, and the barrier function can be re-expressed as:

[0029] In this embodiment, taking a "bicycle" simplified model of vehicle steering control in a plane rectangular coordinate system as an example, its schematic diagram is asFigure 2 As shown, the vehicle dynamics equation described by the model is as follows:

[0030] Wherein, and respectively represent the horizontal and vertical coordinates of the vehicle in the earth coordinate system. and respectively represent the lateral speed and longitudinal speed in the vehicle coordinate system. In the "bicycle" model, it is assumed that is a fixed value. is the yaw angle, is the yaw rate, represents the steering wheel rotation angle, and are the front wheel steering stiffness and rear wheel steering stiffness respectively, is the vehicle mass, and are the distances from the vehicle center of mass to the front axle and rear axle respectively, is the moment of inertia, is the steering ratio; In this embodiment, taking the above "bicycle" model as an example, the model in CarSim is selected as the simulation model. The vehicle model is selected as a type C hatchback, and the tire model is selected as 215 / 55 R17. Some vehicle model parameters provided by CarSim are shown in Table 1.

[0031] Table 1

[0032] Select the vehicle state as , the control input as , and the allowable control input set is set as , and the formulation is:

[0033] In this embodiment, considering that the obstacle avoidance problem pays more attention to the relative position between the vehicle's horizontal and vertical coordinates and the obstacle, the obstacle function is related to the vehicle's horizontal and vertical coordinates. Define the vehicle trajectory , proposes a situation with 5 obstacles, and the specific positions of the obstacles are shown in Table 2 below. Table 2 is: Table 2

[0034] The dangerous area radius of the obstacle is set as , therefore, the obstacle function can be described as:

[0035] Among them, .

[0036] Step 102, integrate the obstacle function into the cost function of the control system, and transform the collision avoidance control problem of the control system into an optimal control problem based on the cost function; In this embodiment, in order to represent the perception degree of the driverless vehicle for obstacles, the following form of circular perception area is defined for each obstacle :

[0037] Among them, is the outer diameter of the circular area, representing the minimum distance at which the system can start to perceive the obstacle, is the inner diameter of the circular area, representing the maximum distance at which the system must perceive the obstacle. Within the circular area, the system's perception ability for the obstacle decreases as the distance from the center of the obstacle increases. The perception degree of the control system for the obstacle can be formulated as:

[0038] Among them, is the distance from the center of the obstacle; In order to transform the collision avoidance control problem into an optimal control problem, the following form of cost function is first defined:

[0039] Among them, is the instantaneous cost, including system performance and the obstacle function, and can be specifically described as:

[0040] Among them, the matrix and the matrix are real symmetric positive definite matrices; in actual numerical simulations, a saturated hyperbolic function is used to limit the obstacle function, and the cost function can be re-expressed as:

[0041]

[0042]

[0043] Based on the purpose of the collision avoidance problem, the control objective is to obtain an optimal control input , such that for , the following inequality holds:

[0044] According to the Pontryagin minimum principle, the optimal control input satisfies the following necessary conditions:

[0045] where, is the Hamiltonian function, is the partial derivative of the Hamiltonian function with respect to the state , is the co-state variable.

[0046] In this embodiment, taking the above "bicycle" model as an example, the inner diameter and outer diameter of the sensing area are set to , ; the driving scenario is designed as shown in Figure 3 . The initial position of the vehicle is at the origin of coordinates, and the control objective is to make the vehicle track a given trajectory , while avoiding hitting obstacles. The position information of the obstacles is the same as that in Table 2. The target trajectory tracked by the vehicle is a cosine function path in the coordinate axis plane, and the specific description is:

[0047] where, is a constant, and the value of the abscissa of the target trajectory at any time can be solved by the following formula:

[0048] where, is the second kind of elliptic integral, and can be calculated by the following boundary conditions:

[0049] Based on the proposed control objective, the obstacle function is integrated into the cost function through the sensing function, and the finite time domain is selected as . The cost function is formulated as:

[0050] where, the matrix and the matrix are real symmetric positive definite matrices. Select , ; Therefore, the optimal control problem is described as optimizing the control input based on the given vehicle control system in the given admissible control input set to minimize the proposed cost function of the vehicle in the given time domain.

[0051] ​Step 103: Establish a control system model of unknown parameters based on the neural network function approximation method, and design a parameter adaptation law using the integral parallel learning method to identify the unknown parameters in the model; In this embodiment, the control system of unknown parameters is expressed as:

[0052] where, is an unknown coefficient matrix, is the designed activation function, is the approximation error. When the number of neurons is large enough, the approximation error can converge to a sufficiently small range. Therefore, the unknown control system can be approximately expressed as:

[0053] where, is the estimated value of the unknown coefficient matrix; The integral parallel learning method uses the numerical integral of historical data over a period of time to online identify the unknown parameters in the model, getting rid of the need for state derivative data in the traditional parallel learning method. For this purpose, collect W groups of historical trajectory data within, denoted as , which represents the set composed of the trajectories of the system and the corresponding control inputs at W different times. For each group of data, calculate:

[0054] where, , define the matrix . When collecting historical trajectory data, it is required that the matrix satisfies ; Define as the estimated value of the system state in the th iteration process, and its adaptation law is designed as follows:

[0055] where, is the learning rate, is the estimated error of the system state in the th iteration process; Define as the estimated value of the coefficient matrix in the th iteration process. Using the historical data set, the adaptation law of the estimated value of the coefficient matrix is designed as follows:

[0056] where is the learning rate, is the identification error of the coefficient matrix in the

[0057] In this embodiment, referring to the vehicle steering control "bicycle model", a neural network with a single hidden layer is selected and expressed as: ; The control inputs of the human driver and the autonomous driving system are randomly given as , where is a random number with a Gaussian distribution, and 200 sets of trajectory data within the randomly selected time interval are used as historical trajectory data.

[0058] Step 104: Based on the control system model, using the quadratic optimization framework, an adaptive iterative learning method is adopted to solve for the optimal control input that meets the requirements of the collision avoidance control task.

[0059] In this embodiment, the specific process of the adaptive iterative algorithm is as follows: S1: Initialize the algorithm parameters , , and the algorithm convergence accuracy , , select the initial control input , and obtain the estimated value of the initial system trajectory and the estimated value of the initial coefficient matrix by calculating and solving the following equations: within:

[0060] where, and and are the initial values of the system state estimation error and the coefficient matrix identification error respectively, is the initial value of the true state trajectory of the system, calculate the initial cost function and let ; S2: Solve for by calculating the following formula:

[0061] where, is the estimated value of the Hamiltonian function, is its partial derivative with respect to the state variable ; S3: Solve for through the quadratic optimization framework:

[0062] where, is the estimated value of the Hamiltonian function with respect to the variable The partial derivative of , let ; S4: Obtain the online real trajectory data according to the new control input , and update the estimated value of the system trajectory and the estimated value of the coefficient matrix by solving the following formula:

[0063] where and are the updated system state estimation error and the coefficient matrix identification error respectively; S5: Calculate the cost function through the updated control input and the online trajectory data , and determine whether the updated control input reduces the cost function: If , then let , , is the iteration step size, and return to S3, otherwise, enter S6; S6: If , then enter S7, otherwise let , , and return to S2; S7: If , stop the iteration process and obtain the corresponding optimal control input, otherwise let , , , , and return to S2.

[0064] In this embodiment, referring to the vehicle steering control "bicycle model", the initial state value of the vehicle system is selected as: , the initial control input is selected as: , the initial parameters of the adaptive iteration algorithm are set as: , the initial coefficient matrix is randomly selected as , specifically as follows:

[0065] Iterate according to the above adaptive iteration algorithm process. After 35 iterations, the algorithm converges. The cost function iteration graph is as shown in Figure 4 . It can be seen that after 33 iterations, the cost function drops sufficiently, and subsequent iterations cannot further reduce the cost value, indicating that the algorithm has converged to the optimal control input; Figure 5This is a schematic diagram of the vehicle path during some iterations. It shows how, given the initial control input, the vehicle gradually adjusts its path during the iterations to approach the target trajectory while avoiding obstacles. It can be seen that when the iterations stop, the vehicle achieves a good obstacle avoidance effect. Figure 6 This is a comparison chart of the actual value of the vehicle state and its estimated value. It can be observed that the error between the identification model and the actual model in CarSim is very close; Figure 7 , Figure 8 , Figure 9 The schematic diagrams of the vehicle paths in CarSim under the initial control input, the control input after the 20th iteration, and the converged control input are shown respectively. It can be seen that the initial trajectory is very different from the target trajectory and cannot effectively avoid obstacles. In addition, there is a large error between the estimated trajectory and the true value. After 20 iterations, the vehicle tracking performance and obstacle avoidance effect are significantly improved, and the identification error is also significantly reduced. When the algorithm converges, it can be seen that the identification error has been reduced to a sufficiently small range, and the vehicle has achieved good results in tracking performance and obstacle avoidance effect in a limited time domain. The obstacle avoidance control method for unmanned vehicles in a limited time domain of the embodiment of the present application is based on the forward invariance to describe the safety of the control system. On this basis, the corresponding obstacle function is constructed according to the obstacle; the constructed obstacle function is integrated into the cost function, and the collision avoidance control problem is converted into an optimal control problem; a parameter identification model of the unknown system is established based on the neural network function approximation method, and an adaptive law is designed using the integral parallel learning method to identify the unknown parameters in the model online; based on the identification model and the identification parameters, a quadratic optimization framework is used to propose an adaptive iterative learning algorithm to solve the optimal control input that meets the requirements of the collision avoidance control task. The present application does not rely on the precise parameter model of the system, takes into account the optimization performance and safety of the control system, and can effectively solve the collision avoidance control problem in a limited time domain.

[0066] In order to implement the above-mentioned embodiments, the present application also proposes a limited time domain unmanned vehicle obstacle avoidance control device.

[0067] Figure 10 A schematic structural diagram of a limited-time-domain unmanned vehicle obstacle avoidance control device provided in an embodiment of the present application.

[0068] like Figure 10 As shown, the limited time domain unmanned vehicle obstacle avoidance control device includes: A first system building module is used to describe the safety of the control system based on forward invariance, and to construct an obstacle function of the control system based on the safety of the control system and according to obstacles; The second system establishment module is used to integrate the obstacle function into the cost function of the control system, and transform the collision avoidance control problem of the control system into an optimal control problem based on the cost function; The system identification module is used to establish a control system model with unknown parameters based on the neural network function approximation method, and design a parameter adaptive law by using the integral parallel learning method to identify the unknown parameters in the model; The iterative solution module is used to solve the optimal control input that meets the requirements of the collision avoidance control task by using the quadratic optimization framework and the adaptive iterative learning method based on the control system model.

[0069] It should be noted that the foregoing explanation of the embodiment of the finite-time domain unmanned vehicle obstacle collision avoidance control method also applies to the finite-time domain unmanned vehicle obstacle collision avoidance control device of this embodiment, and will not be elaborated here.

[0070] To implement the above embodiment, the present application also proposes a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the method described in the above embodiment is implemented.

[0071] In the description of this specification, the description with reference to terms such as "one embodiment", "some embodiments", "example", "specific example" or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0072] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present application, "a plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0073] Any process or method description represented in a flowchart or otherwise described herein may be understood to represent a module, segment, or portion of code including one or more executable instructions for implementing a customized logical function or process. The scope of the preferred embodiments of the present application includes additional implementations in which functions may be executed not in the order shown or discussed, including in a substantially simultaneous manner according to the involved functions or in a reverse order, which should be understood by those skilled in the technical field of the embodiments of the present application.

[0074] The logic and / or steps represented in a flowchart or otherwise described herein, for example, may be considered a sequenced list of executable instructions for implementing a logical function and may be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection having one or more wires (electronic device), a portable computer diskette (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or other suitable medium on which the program can be printed, as the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpretation, or otherwise processing as appropriate, and then storing it in a computer memory.

[0075] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one of the following techniques known in the art or a combination thereof can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0076] Those of ordinary skill in the art can understand that all or part of the steps carried out in the methods of the above embodiments can be completed by instructing relevant hardware through a program. The program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

[0077] In addition, in each of the embodiments of the present application, the functional units can be integrated into a processing module, or each unit can exist physically alone, or two or more units can be integrated into one module. The above integrated module can be implemented in the form of hardware or in the form of a software functional module. When the above integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0078] The storage medium mentioned above can be a read-only memory, a magnetic disk, an optical disc, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. A method for obstacle avoidance control of an unmanned vehicle in a finite time domain, characterized in that, Including: Describe the safety of the control system based on forward invariance, and construct an obstacle function of the control system according to the obstacles based on the safety of the control system; Integrate the obstacle function into the cost function of the control system, and transform the collision avoidance control problem of the control system into an optimal control problem based on the cost function; Establish a control system model with unknown parameters based on the neural network function approximation method, and design a parameter adaptation law using the integral parallel learning method to identify the unknown parameters in the model; Based on the control system model, use the quadratic optimization framework and adopt the adaptive iterative learning method to solve the optimal control input that meets the requirements of the collision avoidance control task.

2. The obstacle avoidance control method for a driverless vehicle in a finite time domain according to claim 1, characterized in that, The description of the safety of the control system based on forward invariance includes: Formulate the control system as: Among them, is the system state variable, is the initial state variable, is the system control input, is the dynamic equation of the control system; Assumptions and its relative status and input The partial derivative of , Satisfy local Lipschitz continuity; Define the system control input The set formed is a convex set , which is expressed as: Among them, is the constraint matrix, is the constraint vector; Assume that for a closed and bounded control input , at the initial value , the solution trajectory of the control system is bounded. The formulation of this assumption is as follows: For , such that Among them, is a finite time domain, is the set composed of states , is the upper bound of the norm of state . The collision avoidance control problem requires that the state of the control system always remains within a given safety set. Define the safety set as follows: Among them, is a continuously differentiable function, and the set is the boundary of, and the set is the interior of, , ; If for any initial state , the solution trajectory always remains within the set , it is determined that the set has forward invariance. If the system state satisfies the forward invariance condition with respect to the set , it is determined that the control system is safe with respect to the set .

3. The obstacle avoidance control method for an unmanned vehicle in a finite time domain according to claim 2, wherein The construction of the obstacle function of the control system according to the obstacles based on the safety of the control system includes: Define the obstacle function based on the safety set as: The obstacle function satisfies: Considering obstacles, define the circular region as the unsafe region of the -th obstacle, denoted as: Among them, is the center of the th obstacle, is the radius of the unsafe area; For a circular region , the obstacle function of the -th obstacle is expressed as: Among them, ; In actual data simulation, the saturation hyperbolic function is used to limit the barrier function so that the barrier function is updated to: 。 4. The obstacle avoidance control method for a driverless vehicle in a finite time domain according to claim 3, characterized in that, The integration of the obstacle function into the cost function of the control system includes: Define a circular sensing area for each obstacle , denoted as: Wherein, is the outer diameter of the annular region, representing the minimum distance at which the control system starts to sense an obstacle, is the inner diameter of the annular region, representing the maximum distance at which the system senses an obstacle; Define the perception level of the control system for obstacles as follows: Among them, , is the distance to the center of the th obstacle; Define the cost function as: Among them, is the instantaneous cost, including system performance and the barrier function, expressed as: Among them, the matrix and the matrix are real symmetric positive definite matrices; In the actual numerical simulation, the hyperbolic function is used to limit the barrier function, so that the cost function is updated to: Among them, is the instantaneous cost, is: 。 5. The obstacle avoidance control method for an unmanned vehicle in a finite time domain according to claim 4, characterized in that, The transformation of the collision avoidance control problem of the control system into an optimal control problem based on the cost function includes: Based on the collision avoidance control problem of the control system, the control objective is to obtain the optimal control input , such that for , the following is satisfied: ; According to the Pontryagin minimum principle, the optimal control input is determined to satisfy the conditions as follows: Among them, , is the partial derivative of the Hamiltonian function with respect to , and is the co-state quantity.

6. The obstacle avoidance control method for a driverless vehicle in a finite time domain according to claim 5, characterized in that The establishment of a control system model with unknown parameters based on the neural network function approximation method includes: Represent the control system with unknown parameters as: Among them, is an unknown coefficient matrix, is an activation function, is an approximation error; When the number of neurons is sufficiently large and the approximation error converges to a sufficiently small range, the control system of the unknown parameters is approximately expressed as: Among them, is the estimated value of the unknown parameter matrix.

7. The obstacle avoidance control method for a finite-time domain driverless vehicle according to claim 6, wherein The establishment of a control system model with unknown parameters based on the neural network function approximation method and the design of a parameter adaptation law using the integral parallel learning method to identify the unknown parameters in the model include: Collect the historical trajectory data within Group W , denoted as , representing a set composed of the trajectories of the system and the corresponding control inputs within different times for each group of data, and calculate for each group of data: Among them, ; Define matrix , the collected historical trajectory data satisfies ; Design the parameter adaptation law during update using the integral parallel learning method, and update the unknown parameters in the control system using the collected historical trajectory double data, where the designed parameter adaptation law is: Definition is the estimated value of the system state in the -th iteration process, and its adaptation law is as follows: Among them, , is the learning rate, is the estimated error of the system state during the -th iteration process; Definition is the estimated value of the coefficient matrix in the -th iteration process, is the identification error of the coefficient matrix in the -th iteration process, , The adaptive law of Among them, , is the learning rate.

8. The obstacle avoidance control method for a finite-time domain driverless vehicle according to claim 7, wherein, Based on the control system model, use the quadratic optimization framework and adopt the adaptive iterative learning method to solve the optimal control input that meets the requirements of the collision avoidance control task includes: Step S1: Initialize the iteration parameters , , and the convergence accuracy , , select the initial control input , and obtain the estimated value of the initial system trajectory and the estimated value of the initial coefficient matrix within by solving the following system of equations: wherein, and are the initial values of the system state estimation error and the coefficient matrix identification error respectively, is the initial value of the true state trajectory of the system; Calculate the initial cost function , is the integral of the instantaneous cost, are the upper and lower limits of the integral respectively, is the integration variable; Let ; Step S2: Calculate by the following formula :[[]]END]] Among them, , representing the estimated value of the Hamiltonian function, is its partial derivative with respect to the state variable ; Step S3: Solve for the control input increment through a secondary optimization method : Among them, , is the partial derivative of the estimated value of the Hamiltonian function with respect to the variable , ; Let ; Step S4: Obtain the online true trajectory from the updated control input , and update the estimated value of the system trajectory and the estimated value of the coefficient matrix by solving the following formula: wherein, and are the updated system state estimation error and the coefficient matrix identification error, respectively; Step S5: Calculate the cost function using the updated control input and the online trajectory data , and determine whether the updated control input reduces the cost function. If , then let , , be the iteration step size, return to step S3. Otherwise, proceed to step S6; Step S6: If , then proceed to Step S7; otherwise, set , , , and return to Step S2; Step S7: If , stop the iterative process to obtain the corresponding optimal control input; otherwise, let , , , , , and return to Step S2.

9. An obstacle avoidance control device for an unmanned vehicle in a finite time domain, characterized in that, Including: The first system establishment module is used to describe the safety of the control system based on forward invariance, and construct an obstacle function of the control system according to the obstacles based on the safety of the control system; The second system establishment module is used to integrate the obstacle function into the cost function of the control system, and transform the collision avoidance control problem of the control system into an optimal control problem based on the cost function; The system identification module is used to establish a control system model with unknown parameters based on the neural network function approximation method, and design a parameter adaptation law using the integral parallel learning method to identify the unknown parameters in the model; The iterative solution module is used to based on the control system model, use the quadratic optimization framework and adopt the adaptive iterative learning method to solve the optimal control input that meets the requirements of the collision avoidance control task.

Images