Local path planning method for automatic driving system
By combining the information obtained by Frenet coordinates and multiple sensors in local path planning, the vehicle behavior before and after is evaluated and the path dynamically adjusts, the problem of inaccurate obstacle avoidance in the existing technology is solved, and the obstacle avoidance ability and path of the autonomous driving system are improved. The adaptability of choice.
Patent Information
- Application Number
- CN202411000947.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-24
- Publication Date
- 2024-11-08
AI Technical Summary
The existing local path planning algorithm cannot consider driving experience and dynamic changes in the environment, resulting in insufficient dynamic obstacle avoidance and prone to collision or scratches.
Through Frenet coordinate-based path planning, combining vehicle positioning module, visual camera and lidar to obtain geographical information, establish a driving trajectory model, evaluate the driving behavior and route prediction of front and rear vehicles, and form a dynamic driving strategy and path planning.
It improves the dynamic obstacle avoidance capability in local areas, reduces the risk of collision, and optimizes path selection to adapt to actual road conditions, reducing computational complexity.
Smart Images

Figure CN118915752A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of path planning, and in particular to a method for local path planning of an automatic driving system. Background Art
[0002] Path planning refers to the selection of a route from one location to another, which includes local path planning and global path planning. The global path planning algorithm refers to selecting the shortest overall route from the starting point to the end point; local path planning is the process of planning a passable local route from the current point to a certain local range according to the surrounding obstacle environment.
[0003] At present, local path planning is usually based on Frenet coordinates, which is mainly converted into Frenet coordinates in the Cartesian coordinate system, and the lateral and longitudinal path planning is completed by deducing polynomial coefficients; however, in local path planning, the conventional algorithm process is still the obstacle avoidance algorithm, that is, calculating the obstacles in front and behind the vehicle, and completing the path planning effect by avoiding the obstacle points. However, the complexity of road driving is far greater than the calculation efficiency. Because the driving habits and driving habits of novices and veterans, as well as the degree of fatigue of the driver are very different, it is obviously impossible to completely form the optimal path selection in the local path. The optimal path selected by the algorithm is only a choice under the algorithm system, and does not take into account the driving experience of the front and rear vehicle conditions. The final dynamic obstacle avoidance planning cannot form an accurate judgment of the wind direction field and the judgment mechanism of following the vehicle. For example, there will still be situations such as emergency avoidance of the front vehicle or vehicles or pedestrians that suddenly appear in the blind spot of vision. Summary of the invention
[0004] The technical problem to be solved by the present invention is to overcome the defects of the prior art and provide a method for local path planning of an autonomous driving system.
[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:
[0006] The present invention provides a method for local path planning of an automatic driving system, comprising the following steps:
[0007] S1. Based on the positioning module signal of the driving system, determine the geographic information within the target area, the geographic information includes elevation difference and local road sign information and local map information;
[0008] S2. Establish a coordinate system based on the path planning of Frenet coordinates, split the long-distance path planning constructed by the Cartesian coordinate system into local paths from the center line segments, form a planning cycle, and collect current vehicle information;
[0009] S3. Establish a driving trajectory model within a local area based on the current speed, angular velocity and acceleration, and construct solutions for multiple foreseeable driving trajectories;
[0010] S4. Confirm the environment around the vehicle, evaluate the target driving behavior of the front and rear vehicles, and predict the driving routes of the front and rear vehicles;
[0011] S5. Output a specific control plan, form a corresponding driving strategy in the local area according to the driver's selection requirements, and quickly output the corresponding driving path.
[0012] As a preferred technical solution of the present invention, the positioning module includes vehicle positioning data and images obtained by a visual camera and / or a laser radar and a remote sensing satellite.
[0013] As a preferred technical solution of the present invention, the elevation difference and local road landmark information are obtained through lidar, or the local road landmark information is based on a visible camera, and the elevation difference is based on a three-dimensional map model constructed by a remote sensing satellite according to elevation DEM data, wherein the vehicle is positioned within the model by a positioning module signal.
[0014] As a preferred technical solution of the present invention, the driving trajectory model is based on the output of the current dynamic position data and is simply calculated according to the following formula:
[0015]
[0016] Dis is the distance and Angle is the angle, that is, the distance traveled by the current local path is collected based on the vehicle itself as the origin, as well as the acceleration and angular velocity during the forward process, and the vehicle's brake usage information; after completing the current local information planning, the lateral path planning and longitudinal path planning are completed according to the Frenet coordinates.
[0017] As a preferred technical solution of the present invention, in step S4, according to the vehicle information obtained in step S3, the distance detection of the front and rear vehicles is completed based on the visual camera or lidar, and based on the Tensorflow deep learning framework, an LSTM network model is established to predict the driving behavior of the front and rear vehicles and evaluate the danger.
[0018] As a preferred technical solution of the present invention, in step S5, the final local path planning is completed according to the driver's autonomous selection and judgment, the following distance of the front vehicle, and the driving distance to the rear vehicle based on the selected solution trend.
[0019] Compared with the prior art, the present invention has the following beneficial effects:
[0020] The present invention establishes an evaluation system for the front and rear vehicles, cooperates with the driving status of the driving vehicle, and based on the final driving plan decision, completes the decision on the following intensity of the vehicle in front, the overtaking priority and the vehicle's driving speed. Before an accident occurs or an obstacle appears, a larger dynamic obstacle avoidance space can be given, reducing the collision phenomenon of vehicles within the local planning range.
[0021] The present invention further provides a method for adjusting the priority of paths during vehicle dynamic planning. Based on more factors, the path selection that can be formed in the lateral and longitudinal directions can be more in line with the actual road conditions without repeatedly increasing the calculation intensity. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:
[0023] Figure 1 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION
[0024] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.
[0025] Example 1
[0026] like Figure 1 As shown, the present invention provides a method for local path planning of an automatic driving system, comprising the following steps:
[0027] S1. Based on the positioning module signal of the driving system, determine the geographic information within the target area, the geographic information includes elevation difference and local road sign information and local map information;
[0028] S2. Establish a coordinate system based on the path planning of Frenet coordinates, split the long-distance path planning constructed by the Cartesian coordinate system into local paths from the center line segments, form a planning cycle, and collect current vehicle information;
[0029] S3. Establish a driving trajectory model within a local area based on the current speed, angular velocity and acceleration, and construct solutions for multiple foreseeable driving trajectories;
[0030] S4. Confirm the environment around the vehicle, evaluate the target driving behavior of the front and rear vehicles, and predict the driving routes of the front and rear vehicles;
[0031] S5. Output a specific control plan, form a corresponding driving strategy in the local area according to the driver's selection requirements, and quickly output the corresponding driving path.
[0032] The positioning module contains vehicle positioning data and images obtained by visual cameras and / or lidar and remote sensing satellites.
[0033] The elevation difference and local road sign information are obtained through lidar, or the local road sign information is based on a visible camera. The elevation difference is based on a three-dimensional map model constructed by a remote sensing satellite based on elevation DEM data, and the vehicle is positioned within the model using the positioning module signal.
[0034] The driving trajectory model is based on the output of the current dynamic position data and makes a simple calculation according to the following formula:
[0035]
[0036] Dis is the distance and Angle is the angle, that is, the distance traveled by the current local path is collected based on the vehicle itself as the origin, as well as the acceleration and angular velocity during the forward process, and the vehicle's brake usage information; after completing the current local information planning, the lateral path planning and longitudinal path planning are completed according to the Frenet coordinates.
[0037] In step S4, according to the vehicle information obtained in step S3, the distance detection of the front and rear vehicles is completed based on the visual camera or lidar, and based on the Tensorflow deep learning framework, an LSTM network model is established to predict the driving behavior of the front and rear vehicles and evaluate the danger.
[0038] In step S5, the final local path planning is completed according to the driver's autonomous choice and judgment, the following distance of the front vehicle, and the driving distance to the rear vehicle based on the selected solution trend.
[0039] Specifically, the vehicle evaluation factors mainly complete the driving evaluation based on the driver's driving habits, mainly based on the visual camera or laser radar to complete the distance detection of the vehicles in front and behind, calculate the actual speed of other vehicles according to the speed and time, and then complete the evaluation of the target vehicles in front and behind. The evaluation factors include the downhill and uphill rate, the number of emergency stops or emergency avoidances, the driving speed, whether the driving is in a straight line, and the overtaking time, etc., and the road spectrum data, including altitude and vehicle speed, are sorted and normalized, and other vehicles are predicted according to the driving trajectory of step S3.
[0040] Under the evaluation system of the present invention, in the selection of the optimal trajectory, if Figure 1As shown, when overtaking and looking for a safe overtaking distance, first determine whether there are other vehicles with the same trend, and verify their past overtaking habits. For example, when observing the overtaking of the following vehicle, check the safe distance, overtaking speed and lane change time left by the following vehicle for overtaking, as well as the calculation of the overtaking distance left by the following vehicle. When the overtaking habit of the target vehicle is determined to be safe, a vehicle following can be formed, and the target vehicle can be followed according to the following intensity and following distance scheme selected in step S5, thereby forming the same path planning when moving forward in the local area;
[0041] When no vehicle with the same destination appears, the driving risk field in the scene is calculated under normal conditions, and the path planning based on Frenet coordinates is completed. At this time, the obstacle avoidance route is further selected based on the driving evaluation. For example, when facing a vehicle with a dangerous driving tendency, an obstacle avoidance route away from the vehicle is adopted instead of the optimal route to complete the obstacle avoidance. If the target dangerous vehicle does not appear at a safe distance for a long time, the route is replanned.
[0042] Under the planning system of the present invention, an evaluation system for other vehicles is unconventionally established, and this evaluation system is used to form a safe driving route for multiple vehicles on the driving route; because under the route planning system, the transportation space occupied by the algorithm system for following the vehicle is smaller, it can further collect more information about the road conditions and surrounding vehicles, and can complete the selection of the following vehicle for the next time; at the same time, because under different road conditions, the existing algorithm only focuses on the planning of the forward route, but under this method, the driving evaluation of the rear vehicle can be completed, and the evaluation of the obstacles in front can be completed according to the actions of the rear vehicle. For example, when a major vehicle accident occurs in front, the rear vehicle with safe driving habits suddenly brakes suddenly, accelerates abnormally or reverses, etc., and the driven vehicle will also add the abnormal situation to the evaluation of the road condition in front, and combine the focal length of the front camera to judge the accident, and the judgment of the vehicle evaluation to complete the effects of emergency stop, reverse, turn to the emergency lane to pass quickly, and turn around to stay away from the danger source, so as to make up for the judgment of the frenet algorithm on the scene of driving wind that suddenly appears far away
[0043] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A method for local path planning of an autonomous driving system, It is characterized in that The steps include: S1. Based on the positioning module signal of the driving system, determine the geographic information within the target area, the geographic information includes elevation difference and local road sign information and local map information; S2. Establish a coordinate system based on the path planning of Frenet coordinates, split the long-distance path planning constructed by the Cartesian coordinate system into local paths from the center line segments, form a planning cycle, and collect current vehicle information; S3. Establish a driving trajectory model within a local area based on the current speed, angular velocity and acceleration, and construct solutions for multiple foreseeable driving trajectories; S4. Confirm the environment around the vehicle, evaluate the target driving behavior of the front and rear vehicles, and predict the driving routes of the front and rear vehicles; S5. Output a specific control plan, form a corresponding driving strategy in the local area according to the driver's selection requirements, and quickly output the corresponding driving path.
2. A method for local path planning of an autonomous driving system according to claim 1, It is characterized in that The positioning module includes vehicle positioning data and images obtained by visual cameras and / or laser radars and remote sensing satellites.
3. The method for local path planning of an automatic driving system according to claim 2, It is characterized in that The elevation difference and local road sign information are obtained through laser radar, or the local road sign information is based on a visual camera, and the elevation difference is based on a three-dimensional map model constructed by a remote sensing satellite according to elevation DEM data, wherein the vehicle is positioned within the model by a positioning module signal.
4. The method for local path planning of an automatic driving system according to claim 3, It is characterized in that The driving trajectory model is based on the output of the current dynamic position data and is simply calculated according to the following formula: Dis is the distance and Angle is the angle, that is, the distance traveled by the current local path is collected based on the vehicle itself as the origin, as well as the acceleration and angular velocity during the forward process, and the vehicle's brake usage information; after completing the current local information planning, the lateral path planning and longitudinal path planning are completed according to the Frenet coordinates.
5. The method for local path planning of an automatic driving system according to claim 4, It is characterized in that In step S4, according to the vehicle information obtained in step S3, the distance detection of the front and rear vehicles is completed based on the visual camera or lidar, and based on the Tensorflow deep learning framework, an LSTM network model is established to predict the driving behavior of the front and rear vehicles and evaluate the danger.
6. A method for local path planning of an automatic driving system according to claim 5, It is characterized in that In step S5, the final local path planning is completed according to the driver's autonomous choice and judgment, the following distance of the front vehicle, and the driving distance to the rear vehicle based on the selected solution trend.
