A system and method for modeling tasks of autonomous driving vehicles in a park

Abstract

The present invention seeks to protect a system and method for modeling tasks for autonomous driving vehicles in a park, including task modeling and task analysis and execution. The task modeling part extracts scene elements by modeling and analyzing the business scenarios and traffic scenarios of autonomous driving vehicles in the park, and forms a set of general task representation standards; for a specific task, a corresponding scene description is generated according to this standard based on the scene sequence involved in the task, and the task model is obtained and stored in the task library. When the vehicle is running, the task analysis and execution part loads the specified task model from the task library, analyzes the corresponding scene description according to the vehicle status, and generates the corresponding information interaction and action sequence by searching the action library, and then calls the vehicle planning control module to control the vehicle operation. The present invention realizes the standardization of the task representation of autonomous driving vehicles in the park, improves the versatility and scalability of the autonomous driving software system, and realizes the rapid development and deployment of the operating functions of autonomous driving vehicles in the park.

Description

Technical Field

[0001] The present invention belongs to the field of automation and computer technology, in particular to the field of task modeling technology for autonomous driving vehicles, and specifically relates to a system and method for task modeling of autonomous driving vehicles in a park. Background Art

[0002] The first area of ​​application of autonomous driving technology is to perform specific functions in specific places. The specific places are mainly scenic spots, ports, factories and other parks. The specific functions include logistics, inspection, and docking.

[0003] In terms of the software architecture of autonomous vehicles, according to the description of its software architecture in Apollo's official "Apollo3.5 Software Architecture", the management of tasks in the Apollo architecture is achieved through path planning combined with a map engine. According to the description of its software architecture in Autoware's official "Autoware-Manuals", the task management function is implemented through global path planning in its planning module. The above two autonomous driving software architectures couple task management with path planning. When doing path planning, the requirements of the task are considered, and the code that implements the task function needs to be solidified in advance, which cannot achieve the effect of modeling and managing abstract tasks, reducing the flexibility of the autonomous driving system. Moreover, the coupling of task management and planning modules leads to poor system compatibility and scalability, and cannot support the addition of new custom tasks.

[0004] Chinese Patent Application: A control method and control device for a park logistics vehicle, and an unmanned vehicle (application number: 202110500189.0) discloses a control method for a park logistics vehicle, which uses path environment information and historical trajectories to estimate the probability of a collision when the park logistics vehicle travels along a global path, and adjusts the travel path of the park logistics vehicle based on the probability value. This method only considers simple factors such as lane width, distance from the lane edge, and number of lanes for vehicle control, and does not involve task modeling, and has insufficient adaptability to complex scenarios.

[0005] Aiming at the operation characteristics of the park's intelligent self-driving vehicles, the present invention designs an intelligent self-driving vehicle software framework based on task modeling, which includes a task library and a task parsing module that represent the tasks of the self-driving vehicle. The software framework separates the task management system of the self-driving vehicle from the decision-making and planning system, thus decoupling the two. Users only need to develop a task library that meets the task representation standards of this article according to the actual application scenario, and then call the action library of this system to implement the corresponding complex business tasks, thereby greatly improving the development efficiency of the system. Summary of the invention

[0006] The present invention aims to solve the above problems of the prior art. A system and method for modeling tasks of autonomous driving vehicles in a park are proposed. The technical solution of the present invention is as follows:

[0007] A method for modeling a task of an autonomous driving vehicle in a park, comprising: a task modeling step and a task parsing and executing step, wherein:

[0008] Task modeling includes the following steps: First, static modeling and dynamic modeling are performed on typical scenarios of autonomous driving vehicles in the park to extract static elements that can describe the scenario and dynamic elements involving information interaction, thus forming a set of general task representation methods. The tasks of autonomous driving vehicles in the park are divided into business scenario tasks. busi , Traffic scene tasks trav And third-party tasks Task other ; Solve the problems obtained from static modeling and dynamic modeling, and design an implementation method for each type of scenario, including: passing through a traffic light intersection, passing through a gate, and parking in a parking space, forming an action library; then, according to this representation method, generate the corresponding scenario description according to the scene sequence involved in the task, obtain the task model and store it in the task library for the actual operating environment of the autonomous driving vehicle;

[0009] The task analysis and execution includes the following steps: when the vehicle is running, the task analysis and execution part loads the task model from the task library, publishes and analyzes the corresponding tasks in real time according to the vehicle status, and then generates the corresponding information interaction and action sequence by searching the action library, and sends it to the task execution module to drive the vehicle to complete the task.

[0010] Furthermore, the business scenario task is a global macro task at the business level, which specifies the task route; busi The element is the starting point Pos for executing a task start 、End pointPos end , the global path segment S of the driving route global ={S1,S2,S3,...S k}、Reference speed V on each path segment refer ={v1,v2,v3…v k}, so the business task Task busi The representation method is as follows:

[0011] Task busi ={Pos start ,Pos end , S global ,V refer};

[0012] The traffic scene task travThe types include traffic lights, gates, and parking spaces. The task model is modeled as follows:

[0013] For the traffic light task, its elements include task interval Task section 、Stop point stop and remaining time of green light green , and use the task type Task type and task number Task id Distinguish between different types of tasks and different individuals of the same type of tasks. light The representation method is as follows:

[0014] Task light ={Task type , Task id , Task section ,Point stop ,Time green}

[0015] Gate Task gate The representation method is as follows:

[0016] Task gate ={Task type ,Task id ,Task section ,Point stop ,Gate id ,Gate state}

[0017] Among them, Task type For task type, Task id Task number, Task section For the task interval, Point stop Stop point, Gate id is the gate number, Gate state It is the gate state.

[0018] Parking space task parking The representation method is as follows:

[0019] Task parking ={Task type ,Task id ,Task section ,Point stop ,C L ,S L ,S W}

[0020] Among them, Tasktype For task type, Task id Task number, Task section For the task interval, Point stop is the stopping point, C L Indicates the road width, S L Indicates the length of the parking space, S w Indicates the width of the parking space.

[0021] Furthermore, the third-party task Task other The modeling is as follows:

[0022] Task other ={Task type ,Task id ,Task section ,Cond,Func}

[0023] Among them, Task type For task type, Task id Task number, Task section is the task interval, Cond is the trigger condition, and Func is the task implementation method provided by the third-party system.

[0024] Furthermore, the task of passing the traffic light intersection is light The implementation methods include:

[0025] ① Determine whether the task interval has been entered based on the current positioning point. If yes, proceed to step ②. If no, repeat step ①.

[0026] ②Get the remaining time of the current green light in the perception interaction module green ;

[0027] ③ Determine the remaining time of the current green light green Is it greater than the threshold value Time? th If yes, go to step ④, if no, go to step ⑤;

[0028] ④ Follow the map path and drive out of the mission area. The mission ends when the current location point is out of the mission area.

[0029] ⑤Stop point stop Stop the car and go to step ②.

[0030] Further, parking space task Task parking The implementation methods include:

[0031] The parking space parking planning problem is described as an optimal control problem to complete the parking task time t f Minimum is the optimization goal;

[0032]

[0033] Where J represents the objective function of minimizing the performance index, z(t) is the differential state variable related to vehicle kinematics, F represents the algebraic function part of the ordinary differential equation, and ζ is (t) represents other algebraic state variables, u (t) represents the control variable, the functional G represents all equality and inequality constraints, t f Indicates the end time of the parking process and is unknown.

[0034] Furthermore, the parking tasks of parking spaces specifically include:

[0035] 1) Parking coordinate system and kinematic modeling: Establish a parking plane coordinate system and set it according to the parking task requirements. For kinematic modeling, use the Ackerman steering model to establish the vehicle kinematic model and obtain the corresponding kinematic differential equations;

[0036] 2) Obtain internal constraints

[0037] Internal constraints refer to the vehicle's own motion constraints, physical structure and motion capability limitations, which should be met by the vehicle. The vehicle's kinematic constraints, steering and speed / acceleration limits are considered to describe the steering and speed / acceleration limits.

[0038] 3) Obtain external constraints

[0039] External constraints include parking space constraints and collision constraints. The vehicle is regarded as a point mass, which should be constrained to move within the parking plane and satisfy the requirement that the parking space boundary and the corner points of the vehicle body do not overlap.

[0040] 4) Obtain boundary value constraints

[0041] Boundary value constraints refer to the motion state constraints that the vehicle meets at the initial and final moments of parking. The initial motion state of the vehicle can be determined based on the measurements of relevant sensors, and it only needs to meet the internal constraints. The final condition indicates the end of the parking process, that is, the vehicle needs to meet the state variable to reach a specific value at the final moment.

[0042] 5) Discrete solution

[0043] The parking continuous-time optimal control problem is discretized, the objective function is solved, and the time-optimal parking trajectory is obtained.

[0044] A task modeling system for autonomous driving vehicles in a park, which includes task loading, task publishing, scenario analysis and task execution modules, wherein:

[0045] The task loading module is used to sequentially retrieve a driving task from the task library.trav ;

[0046] Task release module, used to obtain the current vehicle position Pos myself , calculate the current position Pos myself The straight-line distances Dis1 and Dis2 from the starting point and the end point of the task interval, and the straight-line distance Dis3 from the starting point and the end point of the task interval, determine whether the vehicle position is in the task interval, and if so, publish the task to the task execution analysis module;

[0047] The scene analysis module is used to receive published tasks to be executed, extract static elements from task attributes, and obtain dynamic elements from the environment, such as the remaining time of the green light. green and Gate status state Etc., call the corresponding implementation method in the action library according to the task type. If it is a third-party task, call the task implementation method provided by the third-party system.

[0048] The task execution module is used to call the vehicle control interface to implement the corresponding vehicle motion control.

[0049] Furthermore, the straight-line distance between two points is calculated as follows:

[0050]

[0051] Among them, R is the average radius of the earth, Pi is the pi; lat1, lon1 and lat2, lon2 represent the longitude and latitude coordinates of two points.

[0052] The calculation method to determine whether the current position is in the task interval is as follows:

[0053] ΔDis=|Dis1+Dis2-Dis3|

[0054] ΔDis>d, the current position is not within the task range

[0055] ΔDis<=d, the current position is within the task range

[0056] Where d is the critical value of ΔDis, which is generally half the road width.

[0057] The advantages and beneficial effects of the present invention are as follows:

[0058] Aiming at the operation characteristics of autonomous driving vehicles in the park, the present invention proposes a task modeling system and method for autonomous driving vehicles in the park, which includes two parts: task modeling and task analysis and execution. The task management system and decision-making planning system of the autonomous driving vehicle are independent, thus realizing the decoupling of the two. Users only need to formulate a task library that meets the task representation standards of this article according to the actual application scenario, and then call the action library of this system to perform the corresponding tasks, thereby greatly improving the development efficiency of the system. Through the standardization of the task representation of autonomous driving vehicles and the support for newly added custom tasks, the versatility and scalability of the autonomous driving software system are improved, and the rapid development and deployment of autonomous driving vehicle operation functions can be realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 The present invention provides a preferred embodiment of a method for modeling tasks of a park intelligent autonomous driving vehicle.

[0060] Figure 2 It is a schematic diagram of the state transition of the present invention when passing through a traffic light intersection.

[0061] Figure 3 It is a schematic diagram of the state transition of passing through the gate of the present invention.

[0062] Figure 4 It is a schematic diagram of the parking plane coordinate system of the parking space of the present invention.

[0063] Figure 5 It is a schematic diagram of the vehicle kinematic model of the present invention.

[0064] Figure 6 The present invention provides an intelligent autonomous driving vehicle software framework based on task modeling. DETAILED DESCRIPTION

[0065] The following will describe the technical solutions in the embodiments of the present invention in detail in conjunction with the accompanying drawings in the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention.

[0066] The technical solution of the present invention to solve the above technical problems is:

[0067] like Figure 1 The figure shows a method for modeling tasks of intelligent autonomous driving vehicles in a park, which specifically includes the following stages:

[0068] The task modeling phase includes the following steps: first, static modeling is performed by extracting static elements describing the typical scenarios of autonomous driving vehicles in the park; second, dynamic modeling is performed to extract dynamic elements of information interaction; and then the tasks of autonomous driving vehicles in the park are divided into business scenario tasks. busi , Traffic scene tasks trav And third-party tasks Taskother , forming a set of general task representation methods. Then, for the actual operating environment of the autonomous driving vehicle, according to this representation method, according to the scene sequence involved in the task, the corresponding scenario description is generated, the task model is obtained and stored in the task library; finally, the task model is solved, and for each type of task scenario, an implementation method action and interaction sequence is designed to construct an action library.

[0069] The task analysis and execution phase includes the following steps: when the vehicle is running, the task analysis and execution part loads the task model from the task library, publishes and analyzes the corresponding tasks in real time according to the vehicle status, and then generates the corresponding information interaction and action sequence by searching the action library, and sends it to the task execution module to drive the vehicle to complete the task.

[0070] Figure 2 This is a schematic diagram of the state transition of the present invention passing through a traffic light intersection. Passing through a traffic light intersection is a traffic task. The finite state machine is used to analyze the driving state of the park's autonomous driving vehicle passing through a traffic light intersection. In the figure, the task elements included in the transition conditions between each state mainly include task interval Task section 、Stop point stop and remaining time of green light green At the same time, in order to distinguish different types of tasks and different individuals of the same type of tasks, it is necessary to add task types Task type and task number Task id Explain. Task of passing the traffic light intersection light The representation method is as follows:

[0071] Task light ={Task type , Task id , Task section ,Point stop ,Time green}

[0072] Among them, Task type 、Task id 、Task section 、Point stop Belongs to static elements, Time green It is a dynamic element.

[0073] Traffic light intersection task light The implementation process is as follows:

[0074] ① Determine whether the task interval has been entered based on the current positioning point. If yes, proceed to step ②, otherwise repeat step ①.

[0075] ②Get the remaining time of the current green light in the perception interaction module green .

[0076] ③ Determine the remaining time of the current green light green Is it greater than the threshold value Time? th If yes, go to step ④, if no, go to step ⑤.

[0077] ④ Follow the map path and the mission ends when the current positioning point is out of the mission range.

[0078] ⑤Stop point stop Stop the car and go to step ②.

[0079] Figure 3 This is a schematic diagram of the state transition of the present invention through the gate. The gate passage belongs to the traffic task. The finite state machine is used to analyze the driving state of the park's autonomous driving vehicle through the gate. gate The representation method is as follows:

[0080] Task gate ={Task type ,Task id ,Task section ,Point stop ,Gate id ,Gate state}

[0081] Among them, Task type 、Task id 、Task section 、Point stop 、Gate id It is a static element. state It is a dynamic element.

[0082] Gate passage task gate The implementation steps are as follows:

[0083] ① Determine whether the task interval has been entered based on the current positioning point. If yes, proceed to step ②, otherwise repeat step ①.

[0084] ② According to the gate number of the current task id Get the current gate status Gate in the perception module state .

[0085] ③ Determine the current gate status Gate state Is it in open state? If yes, go to step ④, if no, go to step ⑤.

[0086] ④ Follow the map path and the mission ends when the current positioning point is out of the mission range.

[0087] ⑤Stop point stop Stop the car and go to step ②.

[0088] Figure 4 The figure is a schematic diagram of the parking plane coordinate system of the parking space of the present invention, wherein the origin of the coordinate system is the left front corner point of the parking space when the vehicle is parked correctly, forming a right-hand coordinate system. L is the road width, S L and S w Respectively represent the length and width of the parking space.

[0089] Figure 5 The figure is a schematic diagram of the vehicle kinematic model of the present invention, which adopts the Ackerman steering model, and its corresponding kinematic differential equation is as follows:

[0090]

[0091] Where (x(t), y(t)) represents the coordinates of the current position of the vehicle (the center of the rear axle of the vehicle), v(t) represents the linear velocity of the center of the rear axle of the vehicle, θ(t) represents the heading angle of the vehicle, ɑ(t) is the tangential acceleration of the vehicle, and jert(t) represents the rate of change of acceleration. is the front wheel steering angle, w(t) is the front wheel deflection angular velocity, t f Indicates the completion time of the entire motion process.

[0092] The internal constraints of the vehicle kinematic model are the vehicle's own motion constraints, vehicle physical structure and motion capability limitations, etc., which mainly consider vehicle kinematic constraints, vehicle steering and speed / acceleration limits. The following formula describes the steering and speed / acceleration limit ranges.

[0093]

[0094] The vehicle kinematic constraints should satisfy the kinematic differential equation established in equation (3). The jerk and wheel angular velocity ω in this equation are the control variables of the kinematic equation, and its constraints should satisfy:

[0095]

[0096] in:

[0097]

[0098] Where κ represents the instantaneous curvature, is the corresponding derivative, and It can be set according to the parking task requirements.

[0099] The external constraints of the vehicle kinematic model include parking space constraints and collision constraints. If the vehicle is regarded as a mass point, then this mass point should be constrained to move within the parking plane. Usually, point p in the vehicle kinematic model is used as the mass point. The coordinates of point p (p x (t),p y (t)) should satisfy the constraints in the following formula.

[0100]

[0101] In addition, in order to park the vehicle in a specific parking space, the parking space boundary must not overlap with the corner points of the vehicle body. Usually, the vehicle body outline is regarded as a rectangular box in a two-dimensional Euclidean space, and its four corner points are A, B, C, and D. According to the vehicle kinematic model, the positional relationship between the mass point p and the four corner points can be determined as shown below.

[0102]

[0103] Combination Figure 4 The parking plane coordinate system and the body contour corner points are shown. Successful parking requires satisfying the following inequality constraints.

[0104]

[0105] in:

[0106] f slot (x) = -(H(x) + H(xS L ))S W (10)

[0107] Wherein, the function fslot(x) describes the constraint relationship between the parking space boundary and the vehicle position, CL is the road width, SL and Sw represent the length and width of the parking space respectively, and H(x) is the unit step function.

[0108] The boundary value constraints of the vehicle kinematic model refer to the motion state constraints that the vehicle meets at the initial and final moments of parking. The present invention describes the vehicle motion state as [x(t), y(t), θ(t), a(t), v(t), φ(t), ω(t)]. The initial motion state of the vehicle can be determined based on the measurements of relevant sensors, and the internal constraints can be met. The final condition indicates the end of the parking process, that is, the vehicle needs to meet the state variables to reach a specific value at the final moment, as shown in the following equation.

[0109] [x(t f ),y(t f ),θ(t f ),a(t f ),v(t f ),φ(t f ),ω(tf )]=[x f ,y f ,θ f ,0,0,0,0] (11)

[0110] In the formula, the coordinates of the parking end position (x f ,y f ) and the vehicle heading angle θ f Can be determined by parking task. Parking space parking end point The remaining state variables are all 0.

[0111] Figure 6 The present invention provides an intelligent autonomous driving vehicle software framework based on task modeling, including task loading, task publishing, scenario parsing and task execution modules, wherein:

[0112] The task loading module is used to sequentially retrieve a driving task from the task library. trav ;

[0113] Task release module, used to obtain the current vehicle position Pos myself , calculate the current position Pos myself The straight-line distances Dis1 and Dis2 from the starting point and the end point of the task interval, and the straight-line distance Dis3 from the starting point and the end point of the task interval, determine whether the vehicle position is in the task interval, and if so, publish the task to the task execution analysis module;

[0114] The scene analysis module is used to receive published tasks to be executed, extract static elements from task attributes, and obtain dynamic elements from the environment, such as the remaining time of the green light. green and Gate status state Etc., call the corresponding implementation method in the action library according to the task type. If it is a third-party task, call the task implementation method provided by the third-party system.

[0115] The task execution module is used to call the vehicle control interface to implement the corresponding vehicle motion control.

[0116] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0117] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0118] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0119] The above embodiments should be understood to be only used to illustrate the present invention and not to limit the protection scope of the present invention. After reading the contents of the present invention, technicians can make various changes or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.

Claims

1. A method for modeling tasks of autonomous driving vehicles in a park, characterized in that: include: Task modeling phase and task analysis and execution phase, in which: The process of the task modeling phase is as follows: first, static elements describing the scene are extracted from typical scenes of autonomous driving vehicles in the park, and static modeling is performed; second, dynamic elements of information interaction are extracted through dynamic modeling; and then the tasks of autonomous driving vehicles in the park are divided into business scenario tasks. busi , Traffic scene tasks trav And third-party tasks Task other , forming a set of general task representation methods; then, for the actual operating environment of the autonomous driving vehicle, according to this representation method, according to the scene sequence involved in the task, generate the corresponding scenario description, obtain the task model and store it in the task library; finally, solve the task model, design an implementation method for each type of task scenario, and construct an action library; The process of task analysis and execution phase is as follows: when the vehicle is running, the task model is first loaded from the task library, the corresponding task is issued in real time according to the vehicle status, the scene is analyzed and the corresponding information interaction and action sequence is generated by searching the action library, and then sent to the task execution module to drive the vehicle to complete the task; Divide the tasks of the park's autonomous driving vehicles into business scenario tasks busi , Traffic scene tasks trav And third-party tasks Task other , and construct a general task representation method to represent it; The business scenario task is a global macro task at the business level, which specifies the task route; business task Task busi The element is the starting point Pos for executing a task start 、End pointPos end , the global path segment S of the driving route global ={S1,S2,S3,...S k }、Reference speed V on each path segment refer ={v1,v2,v3…v k }, business task Task busi The representation method is as follows: Task busi ={Pos start ,Pos end ,S global ,V refer } The traffic scene task trav The types include but are not limited to traffic lights, gates, and parking spaces. The task model is modeled as follows: For the traffic light task, its elements include task interval Task section 、Stop point stop and remaining time of green light green , and use the task type Task type and task number Task id Distinguish between different types of tasks and tasks of the same type for different individuals; Traffic light tasks light The representation method is as follows: Task light ={Task type ,Task id ,Task section ,Point stop ,Time green } Gate Task gate The representation method is as follows: Task gate ={Task type ,Task id ,Task section ,Point stop ,Gate id ,Gate state } Among them, Task type For task type, Task id Task number, Task section For the task interval, Point stop For stop point, Gate id is the gate number, Gate state It is the gate state; Parking space task parking The representation method is as follows: Task parking ={Task type ,Task id ,Task section ,Point stop ,C L ,S L ,S W } Among them, Task type For task type, Task id Task number, Task section For the task interval, Point stop is the stopping point, C L Indicates the road width, S L Indicates the length of the parking space, S w Indicates the width of the parking space; The third-party task Task other The modeling is as follows: Task other ={Task type ,Task id ,Task section ,Cond,Func} Among them, Task type For task type, Task id Task number, Task section is the task interval, Cond is the trigger condition, and Func is the task implementation method provided by the third-party system.

2. The method for modeling a task of an autonomous driving vehicle in a park according to claim 1, characterized in that: The traffic light task light The implementation methods include: ① Determine whether the task interval has been entered based on the current positioning point. If yes, proceed to step ②. If no, repeat step ①. ②Get the current remaining time of the green light from the environment perception module green ; ③ Determine the remaining time of the current green light green Is it greater than the threshold value Time? th If yes, go to step ④, if no, go to step ⑤; ④ Follow the map path and drive out of the mission area. The mission ends when the current location point is out of the mission area. ⑤Stop point stop Stop the car and go to step ②.

3. The method for modeling a task of an autonomous driving vehicle in a park according to claim 1, characterized in that: Terminal parking task parking The implementation methods include: The parking planning problem is described as an optimal control problem to complete the parking task time t f Minimum is the optimization goal; min J=t f Where J represents the objective function of minimizing the performance index, z(t) is the differential state variable related to vehicle kinematics, F represents the algebraic function part of the ordinary differential equation, ζ(t) represents other algebraic state variables, u(t) represents the control variable, the functional G represents all the equality and inequality constraints, and t f Indicates the end time of the parking process and is unknown.

4. The method for modeling a task of an autonomous driving vehicle in a park according to claim 3, characterized in that: Terminal parking tasks include: 1) Parking coordinate system and kinematic modeling: Establish a parking plane coordinate system and set it according to the requirements of the parking task. For kinematic modeling, use the Ackerman steering model to establish a vehicle kinematic model and obtain the corresponding kinematic differential equations; 2) Obtain internal constraints Internal constraints refer to the vehicle's own motion constraints, vehicle physical structure and motion capability limitations, which mainly consider vehicle kinematic constraints, vehicle steering and speed / acceleration limits, and describe the steering and speed / acceleration limit ranges; 3) Obtain external constraints External constraints include parking space constraints and collision constraints. If the vehicle is regarded as a point mass, it should be constrained to move within the parking plane. In order to park the vehicle in a specific parking space, it is also necessary to ensure that the parking space boundary and the corner points of the vehicle body do not overlap. 4) Obtain boundary value constraints Boundary value constraints refer to the motion state constraints that the vehicle meets at the initial and final moments of parking. The initial motion state of the vehicle can be determined based on the measurements of relevant sensors, and it only needs to meet the internal constraints. The final condition indicates the end of the parking process, that is, the vehicle needs to meet the state variable to reach a specific value at the final moment. 5) Discrete solution process.

5. The method for modeling a task of an autonomous driving vehicle in a park according to claim 4, characterized in that: The discrete solution process of step 5) specifically includes: using the orthogonal configuration finite element method to discretize the parking continuous time optimal control problem into a nonlinear solution optimization problem, and then using the interior point method to solve the objective function to obtain the time optimal parking trajectory.

6. A campus autonomous driving vehicle task modeling system according to any one of claims 1 to 5, characterized in that: It includes task loading, task publishing, scene parsing and task execution modules, among which: The task loading module is used to sequentially retrieve a driving task from the task library. trav ; Task release module, used to obtain the current vehicle position Pos myself , calculate the current position Pos myself The straight-line distances Dis1 and Dis2 from the starting point and the end point of the task interval, and the straight-line distance Dis3 from the starting point and the end point of the task interval, determine whether the vehicle position is in the task interval, and if so, publish the task to the task analysis and execution module; The scene analysis module is used to receive published tasks to be executed, extract static elements from task attributes, and obtain dynamic elements from the environment, such as the remaining time of the green light. green and Gate status state Etc., call the corresponding implementation method in the action library according to the task type. If it is a third-party task, call the task implementation method provided by the third-party system; The task execution module is used to call the vehicle control interface to implement the corresponding vehicle motion control.

Images