A global path planning method and system for hybrid scenarios compatible with multiple vehicle types

Abstract

The present invention relates to the field of automatic driving technology, and in particular to a global path planning control system in a mixed scenario compatible with multiple vehicle models, including a data perception module, a rule setting module, a scene classification module, a path planning module, a control execution module and a communication integration module. The present invention also discloses a global path planning control method using the above system, and the steps are as follows: S1: the data perception module obtains road information in a region through multiple sensors; S2: scene classification road information, obtains current scene information and classifies it; S3: the rule setting module obtains the vehicle model of the path to be planned, and generates path planning rules for the corresponding vehicle model; S4: the path planning module generates a driving path; S5: the control execution module controls / guides the vehicle to move according to the path generated by the path planning module. The path planning system and method of the present invention can adapt to different vehicle models and scenarios, and improve the adaptability and safety of automatic driving in complex environments.

Description

Technical field:

[0001] The present invention relates to the field of autonomous driving technology, and in particular to a global path planning control system in a hybrid scenario compatible with multiple vehicle models. Background technology:

[0002] With the rapid development of autonomous driving technology, the design of global path planning control systems has become increasingly important. However, existing global path planning control systems are usually only designed for a single vehicle model or a specific scenario, and cannot adapt to the compatible control of multiple vehicle models in mixed scenarios.

[0003] As a hub for sea and land transportation, ports require port vehicles to load, unload and transfer goods. Usually, there are multiple types of vehicles operating in the same port area, such as container trucks, straddle carriers, IGVs, etc. The above types of vehicles have different travel characteristics when planning the paths of operations. Planning the paths for each type of vehicle separately will cause serious waste of resources. Therefore, it has become an urgent need to develop a global path planning control system in mixed scenarios that is compatible with multiple models to improve the adaptability and safety of autonomous driving in complex environments. Summary of the invention:

[0004] The present invention aims to solve the problems existing in the above-mentioned prior art and provides a global path planning control method and system in a hybrid scenario compatible with multiple vehicle types.

[0005] A global path planning control system in a hybrid scenario compatible with multiple vehicle types, characterized by:

[0006] A data sensing module, wherein the data sensing module includes a plurality of sensors for acquiring road information in real time;

[0007] A rule setting module, wherein the rule setting module obtains the vehicle type of the vehicle to be routed and sets corresponding route planning rules according to the vehicle type;

[0008] A scene classification module, wherein the scene classification module obtains the road information collected by the data perception module, obtains the current scene information and classifies it, and distinguishes different types and levels of roads in the scene;

[0009] A path planning module, which performs global path planning based on the scene information obtained by the scene classification module and the vehicle path planning rules set by the rule setting module to generate a vehicle driving path;

[0010] A control execution module, which converts the driving path generated by the global path planning module into a control instruction to control the vehicle in real time;

[0011] The communication integration module is responsible for the communication and data exchange between the modules.

[0012] A global path planning control method in a hybrid scenario compatible with multiple vehicle models is characterized in that the path planning is performed using the above system, including the following steps:

[0013] The S1 data perception module obtains road information in the area through multiple sensors;

[0014] The S2 scene classification module obtains and classifies the current scene information based on the road information obtained by the data perception module, distinguishes different types and levels of roads in the scene, and matches the above roads with the map data;

[0015] The S3 rule setting module obtains the vehicle type of the route to be planned and generates the path planning rules for the corresponding vehicle type;

[0016] The S4 path planning module generates a driving path according to the map data and the path planning rules set by the rule setting module, and sends the driving path to the control execution module;

[0017] S5 controls the execution module to control / guide the vehicle to move according to the path generated by the path planning module.

[0018] Furthermore, in step S1, the road information includes road attributes, obstacle information and pedestrian information.

[0019] Furthermore, in step S2, the map data is obtained by pre-input or generated by a scene classification module according to road information.

[0020] Furthermore, in step S3, the path planning rule is represented by a heuristic function, and the heuristic functions corresponding to different vehicle models are respectively

[0021] Straddle carrier: h(n) = (abs(current.x - goal.x) + abs(current.y - goal.y));

[0022] With hanging card:

[0023] h(n)=sqrt(2)min(|goal.x-current.x|,|goal.y-current.y|)+|goal.x-current.x|+|goal.y-current.y|;

[0024] IGV: h(n)=sqrt((goal.x-current.x)^2+(goal.y-current.y)^2);

[0025] Among them, current.x, current.y are the current coordinates of the vehicle, and goal.x, goal.y are the coordinates of the destination point.

[0026] Furthermore, in step S4, the method in which the path planning module generates the driving path is as follows:

[0027] S41 starts from the initial point of the vehicle, and according to the map data, the points around the initial point that can be used for the next path planning are listed in the open list, and the unreachable points are listed in the closed list, indicating that the points cannot be paid attention to and cannot be iterated;

[0028] S42 selects a point from the open list, generates a heuristic function and a cost function between the initial point and the selected point, traverses all the points in the open list, selects a point with the smallest sum of the heuristic function and the cost function as the next point on the path, and generates a path between the initial point and the point;

[0029] S43 uses the point selected in step S42 as a new initial point, generates a new open list, and repeats steps S41 and S42 until the end point of the path to be planned appears in the open list. The connection path between all the points selected in the above steps and the end point is the planned path;

[0030] S44 performs a reverse calculation on the planned path obtained in step S43 until it returns to the initial point of the vehicle. If the path evaluation passes, the path result is sent to the control execution module. If the reverse calculation fails, the result is returned to the path planning module, the map data is re-acquired, and the path planning is performed again.

[0031] Furthermore, the cost function between the initial point and the selected point is

[0032] Cost(n)=a*(Slope+Distance(current_node,n) / Vmax)+b*

[0033] (Curvature+Traffic_Congestion);

[0034] The Slope is the average slope of the road between the initial point and the selected point, Distance(current_node,n) is the distance between the selected point and the end point, Curvature is the road curvature, Traffic_Congestion is the traffic congestion of the road between the initial point and the selected point, and a and b are constant coefficients.

[0035] Furthermore, when the planned path is reversely calculated in step S44, if the maximum turning curvature of the planned vehicle type is less than the road curvature between any adjacent points, it is determined that the path is impassable.

[0036] The present invention has the following beneficial effects:

[0037] The global path planning control system in the mixed scenario compatible with multiple vehicle models of the present invention can adapt to different vehicle models and scenarios, and improve the adaptability and safety of automatic driving in complex environments.

[0038] Through the collaborative work of modules such as data collection, rule setting, scenario classification, path planning and control execution, adaptive global path planning control can be achieved, thereby improving the efficiency and safety of autonomous driving.

[0039] This system has good scalability and can easily add new car models and scenarios to adapt to the evolving needs of autonomous driving technology. Description of the drawings:

[0040] Figure 1 A flow chart of the global path planning control method in a hybrid scenario compatible with multiple vehicle models; Specific implementation method:

[0041] The present invention will be further described below in conjunction with the accompanying drawings.

[0042] Embodiment 1:

[0043] This embodiment discloses a global path planning control system in a hybrid scenario compatible with multiple vehicle models, including the following modules:

[0044] Data perception module, the data perception module includes multiple sensors for obtaining road information in real time; the multiple sensors for obtaining road information in real time are video / image sensors, infrared sensors, etc. The video / image sensors are used to obtain real-time video / images of the road and send the video / images to the scene classification module. The infrared sensors are set at fixed points such as port intersections to obtain road condition information (whether there are obstacles, pedestrians, etc.).

[0045] The rule setting module obtains the vehicle type of the to-be-planned path and sets the corresponding path planning rules according to the vehicle type. The vehicle types involved in this embodiment include container trucks with hangings, straddle carriers, and IGVs. The above three vehicle types have different sizes and steering angles, which will generate different path planning rules.

[0046] Scene classification module: The scene classification module obtains the road information collected by the data perception module, obtains the current scene information and classifies it, distinguishing the different types and levels of roads in the scene; the scene classification module identifies the road information in the scene based on image recognition and semantic segmentation technology, and the classification of roads includes: single-phase / two-way road, road slope, obstacles on the road, vehicle driving / stopping conditions on the road, pedestrian conditions on the road, etc., and generates corresponding road condition information.

[0047] The path planning module performs global path planning based on the scene information obtained by the scene classification module and the vehicle path planning rules set by the rule setting module to generate a vehicle driving path;

[0048] A control execution module converts the driving path generated by the global path planning module into control instructions to control the vehicle in real time;

[0049] Communication integration module, the communication integration module is responsible for the communication and data exchange between modules.

[0050] Embodiment 2:

[0051] This embodiment discloses a multi-vehicle global path planning control method using the system described in Embodiment 1:

[0052] The S1 data perception module obtains road information in the area through multiple sensors;

[0053] The S2 scene classification module obtains and classifies the current scene information based on the road information obtained by the data perception module, distinguishes different types and levels of roads in the scene, and matches the above roads with the map data;

[0054] The S3 rule setting module obtains the vehicle type of the path to be planned and generates the path planning rules for the corresponding vehicle type; the path planning rules are represented by heuristic functions, and the heuristic functions corresponding to different vehicle types are

[0055] Straddle carrier: h(n) = (abs(current.x - goal.x) + abs(current.y - goal.y));

[0056] With hanging card:

[0057] h(n)=sqrt(2)min(|goal.x-current.x|,|goal.y-current.y|)+|goal.x-current.x|+|goal.y-current.y|;

[0058] IGV: h(n)=sqrt((goal.x-current.x)^2+(goal.y-current.y)^2);

[0059] Among them, current.x, current.y are the current coordinates of the vehicle, and goal.x, goal.y are the coordinates of the destination point.

[0060] The S4 path planning module generates a driving path according to the map data and the path planning rules set by the rule setting module, and sends the driving path to the control execution module;

[0061] S5 controls the execution module to control / guide the vehicle to move according to the path generated by the path planning module.

[0062] The road information in steps S1 and S2 specifically includes single-phase / two-way road, road direction, road curvature, road slope, obstacles on the road, vehicle driving / staying on the road, pedestrians on the road, etc., which are obtained by the data perception module and sent to the scene classification module to generate the above information. It should be noted that in this embodiment, the road information is obtained in real time, and the scene classification module keeps the scene information updated at a fixed time frequency. After obtaining the above information, the scene classification module needs to match it with the map data. The map data here can be imported into a digital map, or it can be generated by the scene classification module by extracting all the data related to the road. The specific way to match the information with the map data is to bind the road conditions on the road between adjacent points with the section of road. The road conditions here include single-phase / two-way road, road direction, road curvature, road slope, obstacles on the road, vehicle driving / staying on the road, pedestrians on the road, etc.

[0063] The cost function of the corresponding vehicle model in step S3 is related to the characteristics of the vehicle model (vehicle movement mode). At the same time, the maximum travel speed and maximum steering curvature coefficient of the corresponding vehicle model need to be obtained. In this embodiment, for the straddle carrier, Vmax=6km / h, the maximum steering curvature coefficient is 4.7, for the trailer with hanging, Vmax=20km / h, the maximum steering curvature coefficient is 7.9, for the IGV, Vmax=30km / h, the maximum steering curvature coefficient is 5.3.

[0064] The method of generating a path in step S4 is to divide the map into multiple reachable points, connect the points through roads, generate a cost function corresponding to the vehicle type, and construct a heuristic function at each point to determine the relative distance and difficulty of reaching the final target location from each point, select the forward route one by one, and finally reach the target location. The specific generation method is as follows:

[0065] Taking the generated osm map as an example, each point on the map is a grid that the vehicle can travel, and also the path point information that the vehicle can travel. Suppose on the osm map, a drivable path needs to be planned with point A as the starting point and point B as the end point. The following is the process of gradually searching for the path using the node evaluation function and the cost function.

[0066] S41 starts from vehicle point A, and according to the map data, the points around the initial point that can be used for the next path planning are listed in an open list, and the unreachable points are listed in a closed list, indicating that the points cannot be paid attention to and cannot be iterated;

[0067] S42 selects a point from the open list, generates the heuristic function and cost function between the initial point A and the selected point, traverses all the points in the open list, calculates the sum of the heuristic function and cost function of the vehicle model corresponding to the point, denoted as f, and there will be different f values ​​around point A. The point with the smallest f is selected as the next point on the path, denoted as point C, and generates a path between the initial point and the point; the cost function between the initial point A and the selected point C is

[0068] Cost(n)=a*(Slope+Distance(current_node,n) / Vmax)+b*

[0069] (Curvature+Traffic_Congestion);

[0070] Slope is the average slope of the road between the initial point and the selected point. The slope has a great influence on the driving speed of the cargo vehicle.

[0071] Distance(current_node,n) is the distance between the selected point and the end point to ensure that each selection is close to point B.

[0072] Vmax is the maximum travel speed of the corresponding vehicle type.

[0073] Curvature is the curvature of the road. Different types of vehicles have different maximum turning curvatures when turning. Therefore, there are some roads in the port area that can only pass some types of vehicles. When generating a path, it is necessary to avoid cargo vehicles passing through intersections that cannot turn, which may cause traffic accidents. Therefore, the road curvature in this embodiment is not only the basis for the path passing time, but also one of the bases for whether the path can be passed.

[0074] Traffic_Congestion is the traffic congestion situation of the road between the initial point and the selected point. This value is obtained by comparing the road idle image with the real-time image analysis through the scene classification module, and setting a threshold for the difference value of the above two images. When the difference value exceeds the threshold, the road section is fed back as impassable.

[0075] a and b are constant coefficients used to adjust the influence of the heuristic function and the cost function on the f value.

[0076] S43 uses the point C selected in step S42 as a new initial point, generates a new open list, and repeats steps S41 and S42 until the end point B of the path to be planned appears in the open list. The path formed by all the points between point A and point B selected in the above steps is the planned path;

[0077] S44 performs a reverse calculation on the planned path obtained in step S43 until it returns to the initial point of the vehicle. If the path evaluation passes, the path result is sent to the control execution module. If the reverse calculation fails, the result is returned to the path planning module, the map data is re-acquired and the path planning is performed again. Specifically, when performing the path reverse calculation, the path between each point needs to be matched with the road attributes to determine whether the path is passable, such as:

[0078] When the maximum turning curvature of the vehicle model to be planned is less than the road curvature between any adjacent points, the path is determined to be impassable; when the traffic congestion of the road section exceeds the threshold, the path is determined to be impassable; when the path direction of the road section does not match the road attributes, the path is determined to be impassable... Road passability is determined when the path is reversed. Although the speed of path planning is sacrificed, on the one hand, it reduces the number of times the system calls up the scene classification module information, and on the other hand, it reduces the system's computational load.

[0079] For the inaccessible paths that appear during the reverse process (such as the section from point B to point C), the path planning module returns the corresponding results. When arriving at point B again, point C is added to the closed list to avoid the inaccessible path again. When a traversable path appears, it is sent to the control execution module as a planned path, and the control execution module controls / guides the vehicle to move according to the path.

[0080] The above generally describes the present invention in detail, but it is obvious to a person skilled in the art that some modifications or improvements can be made to the present invention. Therefore, modifications or improvements that do not depart from the spirit of the present invention are within the scope of protection of the present invention.

Claims

1. A global path planning control method in a hybrid scenario compatible with multiple vehicle models, characterized in that: The steps include: The S1 data perception module obtains road information in the area through multiple sensors; The S2 scene classification module obtains and classifies the current scene information based on the road information obtained by the data perception module, distinguishes different types and levels of roads in the scene, and matches the above roads with the map data; The S3 rule setting module obtains the vehicle type of the route to be planned and generates the path planning rules for the corresponding vehicle type; The path planning rules are represented by heuristic functions, and the heuristic functions corresponding to different vehicle models are: Straddle carrier: h(n)=(abs(current.x-goal.x)+abs(current.y-goal.y)); With hanging card: h(n)=sqrt(2)min(|goal.x-current.x|,|goal.y-current.y|)+|goal.x-current.x|+|goal.y-current.y|; IGV: h(n)=sqrt((goal.x-current.x)^2+(goal.y-current.y)^2); Where current.x, current.y are the current coordinates of the vehicle, goal.x, goal.y are the coordinates of the destination point; The S4 path planning module generates a driving path according to the map data and the path planning rules set by the rule setting module, and sends the driving path to the control execution module; The method for the path planning module to generate a driving path is: S41 starts from the initial point of the vehicle, and according to the map data, the points around the initial point that can be used for the next path planning are listed in the open list, and the unreachable points are listed in the closed list, indicating that the points cannot be paid attention to and cannot be iterated; S42 selects a point from the open list, generates a heuristic function and a cost function between the initial point and the selected point, traverses all the points in the open list, selects the point with the smallest sum of the heuristic function and the cost function as the next point on the path, and generates a path between the initial point and the point. The cost function between the initial point and the selected point is: Cost(n)=a*(Slope+Distance(current_node,n) / Vmax)+b*(Curvature+Traffic_Congestion); Slope is the average slope of the road between the initial point and the selected point, Distance(current_node,n) is the distance between the selected point and the end point, Vmax is the maximum speed of the corresponding vehicle model, Curvature is the road curvature, Traffic_Congestion is the traffic congestion of the road between the initial point and the selected point, and a and b are constant coefficients; S43 uses the point selected in step S42 as a new initial point, generates a new open list, and repeats steps S41 and S42 until the end point of the path to be planned appears in the open list. The connection path between all the points selected in the above steps and the end point is the planned path; S44 performs a reverse calculation on the planned path obtained in step S43 until the path returns to the initial point of the vehicle. If the path evaluation passes, the path result is sent to the control execution module. If the reverse calculation fails, the result is returned to the path planning module, the map data is re-acquired, and the path planning is performed again. S5 controls the execution module to control / guide the vehicle to move according to the path generated by the path planning module.

2. The global path planning control method according to claim 1, characterized in that: In step S1, the road information includes road attributes, obstacle information and pedestrian information.

3. The global path planning control method according to claim 1, characterized in that: In step S2, the map data is obtained by pre-input or generated by a scene classification module according to road information.

4. The global path planning control method according to claim 1, characterized in that: When the planned path is reversely calculated in step S44, if the maximum turning curvature of the planned vehicle type is less than the road curvature between any adjacent points, it is determined that the path is impassable.

5. A global path planning control system in a hybrid scenario compatible with multiple vehicle types, used to implement the method described in any one of claims 1 to 4, characterized in that: include: A data sensing module, wherein the data sensing module includes a plurality of sensors for acquiring road information in real time; A rule setting module, wherein the rule setting module obtains the vehicle type of the vehicle to be routed and sets corresponding route planning rules according to the vehicle type; A scene classification module, wherein the scene classification module obtains the road information collected by the data perception module, obtains the current scene information and classifies it, and distinguishes different types and levels of roads in the scene; A path planning module, which performs global path planning based on the scene information obtained by the scene classification module and the vehicle path planning rules set by the rule setting module to generate a vehicle driving path; A control execution module, which converts the driving path generated by the global path planning module into a control instruction to control the vehicle in real time; The communication integration module is responsible for the communication and data exchange between the modules.

Images