Vehicle autonomous obstacle avoidance method and system based on vision and laser radar

Abstract

The invention relates to a vehicle autonomous obstacle avoidance method and system based on vision and laser radar, and belongs to the technical field of intelligent vehicles. The method comprises the following steps: acquiring illumination intensity data, and automatically switching an acquisition mode according to the illumination intensity; the method comprises the following steps: acquiring real-time position data of a vehicle, calculating an optimal driving route of the vehicle, dividing the optimal driving route according to a preset distance threshold to obtain a target point sequence, and determining real-time target points according to the target point sequence; determining a vehicle advancing direction and a real-time vehicle safety domain, dividing identification subareas, and obtaining the obstacle size, the real-time state quantity and the obstacle historical track of the obstacle in the identification subareas; according to the historical path of the obstacle, calculating through a prediction model to obtain a prediction point coordinate, and calculating through a planning model to obtain an evasion decision. The automatic switching of the vehicle data acquisition mode is realized, and the response speed and the data accuracy of automatic obstacle avoidance are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent automobiles, and in particular relates to a vehicle autonomous obstacle avoidance method and system based on vision and laser radar. Background Art

[0002] With the continuous advancement of science and technology, vehicles are becoming more and more automated and intelligent. Autonomous obstacle avoidance, as one of the key technologies for vehicle intelligence, is of great significance for improving vehicle safety and reliability.

[0003] In traditional vehicle driving, drivers mainly rely on vision and experience to judge the surrounding environment and perform obstacle avoidance operations. However, this method has many limitations. First, human vision has certain limitations. In bad weather conditions (such as fog, heavy rain, etc.) or insufficient light, the driver's vision will be seriously affected, making it difficult to accurately judge the surrounding obstacles. Secondly, the driver's reaction time is limited, and in an emergency, he may not be able to make the correct obstacle avoidance decision in time. In addition, factors such as driver fatigue and distraction will also increase the risk of vehicle collision.

[0004] In existing autonomous obstacle avoidance systems, the vehicle's data collection unit relies on lidar, visual sensors, or a combination of lidar and visual sensors to collect data. However, all three data collection modes have certain problems:

[0005] (1) When the light intensity is high, the data collected by LiDAR will be inaccurate because the laser is affected by the external light intensity.

[0006] (2) When the light intensity is poor, when using visual sensors for data collection, the collected image data is unclear due to the weak light, and obstacles cannot be identified in time.

[0007] (3) Although the obstacle recognition accuracy is improved when data is collected by combining lidar and visual sensors, more data is processed, the data processing time is longer, and higher-precision hardware and software equipment is required. However, in current car driving, the main driving mode is manual driving with automatic driving as a supplement. In this driving mode, there is no need for high-precision obstacle collection equipment. It is only necessary to quickly measure the obstacle position, size, speed and movement direction so that the on-board processor can quickly make obstacle avoidance decisions. Summary of the invention

[0008] In order to solve the above problems existing in the prior art, the present invention provides a vehicle autonomous obstacle avoidance method and system based on vision and laser radar.

[0009] The purpose of the present invention can be achieved through the following technical solutions:

[0010] The light intensity data is collected by the vehicle-mounted photosensor, and the vehicle-mounted control unit generates an adaptive signal according to the light intensity data to control the collection unit to switch the collection mode;

[0011] The real-time position data of the vehicle is obtained through the vehicle positioning unit, and the map data, the driving end point, and the driving starting point are obtained. The vehicle motion planning unit calculates the optimal path through the Dijkstra algorithm according to the map data, the driving end point, and the driving starting point, presets a distance threshold, divides the optimal path according to the distance threshold to obtain a target point sequence, and determines the real-time target point according to the target point sequence and the real-time position data;

[0012] Determine the direction of vehicle travel according to the real-time position data and the real-time target point, preset a vehicle safety distance, determine a real-time vehicle safety domain according to the vehicle safety distance and the real-time position data, establish a coordinate system according to the real-time position data, divide the identification partition according to the coordinate system, obtain the obstacle size and real-time state quantity of the obstacle in the identification partition through the acquisition unit, add a timestamp according to the real-time state quantity to obtain a time series state quantity, and obtain the obstacle history trajectory according to the time series state quantity;

[0013] The predicted point coordinates are calculated by the prediction model according to the historical trajectory of the obstacle, the real-time grid map is obtained by the on-board mapping unit, and the avoidance decision is calculated by the planning model according to the real-time grid map, the vehicle travel direction, the predicted point coordinates, the real-time vehicle safety domain, the real-time state quantity, the obstacle size, and the real-time target point.

[0014] Specifically, the acquisition mode includes a lidar acquisition mode and a visual acquisition mode; the real-time target point is the target point with the shortest distance from the real-time position data in the target point sequence; the real-time state quantity includes the real-time coordinates, speed, and movement direction of the obstacle.

[0015] Specifically, the identified partitions include partition number one, partition number two, partition number three, and partition number four.

[0016] Specifically, the vehicle-mounted control unit generates an adaptive signal according to the light intensity data to control the acquisition unit to switch the acquisition mode, including:

[0017] A light intensity threshold is preset, and the light intensity data is judged according to the light intensity threshold. If the light intensity data is greater than the light intensity threshold, the acquisition mode is the visual acquisition mode; if the light intensity data is less than the light intensity threshold, the acquisition mode is the lidar acquisition mode.

[0018] Specifically, the specific calculation steps of the prediction model include:

[0019] S201: discretizing the obstacle history trajectory to obtain an attribute block, constructing a one-hot vector according to the attribute block, presetting a sparsity constraint, obtaining a high-dimensional embedding vector, and correlating the one-hot vector with the high-dimensional embedding vector according to the sparsity constraint to obtain an association vector;

[0020] S202: Calculate a temporal feature using a temporal convolution model according to the association vector, and calculate a spatial feature using a Transformer model according to the association vector;

[0021] S203: Calculate predicted time features and predicted space features through an adaptive pooling layer according to the time features and the space features, and perform feature fusion through a CNN-LSTM-Attention model according to the predicted time features and the predicted space features to obtain predicted spatiotemporal features;

[0022] S204: Perform a pseudo-inverse operation on the predicted spatiotemporal features to obtain the predicted coordinate points.

[0023] Specifically, the specific calculation steps of the planning model are:

[0024] S301: determining a real-time coordinate according to the real-time state quantity, and determining an obstacle moving direction according to the real-time coordinate and the predicted point coordinate;

[0025] S302: judging according to the predicted point coordinates and the real-time vehicle safety domain, if the predicted point coordinates are within the real-time vehicle safety domain, executing step S303; if the predicted point coordinates are not within the real-time vehicle safety domain, executing the first decision;

[0026] S303: judging according to the obstacle moving direction and the vehicle moving direction, if the obstacle moving direction is parallel to the vehicle moving direction, executing the first decision; if the obstacle moving direction is not parallel to the vehicle moving direction, executing step S304;

[0027] S304: acquiring the real-time coordinates of the obstacle according to the real-time state quantity, and determining an avoidance direction according to the real-time coordinates of the obstacle and the real-time vehicle safety domain; if the obstacle coordinates are within the real-time vehicle safety domain, the avoidance direction is the direction in which the obstacle moves; if the obstacle coordinates are not within the real-time vehicle safety domain, the avoidance direction is the opposite direction of the direction in which the obstacle moves;

[0028] S305: Marking the real-time grid map according to the obstacle size and the real-time coordinates of the obstacle to obtain a marked obstacle map, and calculating a traversal result through a path traversal model according to the marked obstacle map, the real-time target point, the real-time location data, and the travel direction, wherein the traversal result includes whether a path exists or not;

[0029] S306: Making a judgment based on the traversal result, if the traversal result is that the path exists, executing the second decision; if the traversal result is that the path does not exist, executing the third decision.

[0030] Specifically, the path traversal model specifically includes:

[0031] S401: Initialize the avoidance route, set the real-time target point as the traversal starting point, and set the real-time position data as the traversal end point;

[0032] S402: judging the pass factor of the grid in the obstacle map by a heuristic algorithm according to the traversal starting point and the avoidance direction; if the pass factor is 0, storing the grid in the avoidance route and executing step S403; if the pass factor is 1, outputting the traversal result as path non-existence;

[0033] S403: Determine whether the grid is the traversal end point. If the grid is the traversal end point, output the avoidance route and output the traversal result as path existence. If the grid is not the traversal end point, repeat steps S401-S403.

[0034] A vehicle autonomous obstacle avoidance system based on vision and laser radar, comprising: a mode switching module, a data acquisition module, a data processing module, and an avoidance decision module;

[0035] The mode switching module is used to obtain light intensity data through the vehicle-mounted photosensor, and the vehicle-mounted control unit generates an adaptive signal according to the light intensity data to control the acquisition unit to switch the acquisition mode;

[0036] The data acquisition module is used to obtain the real-time position data of the vehicle through the vehicle positioning unit, obtain the map data, the driving end point, and the driving starting point, the vehicle motion planning unit calculates the optimal path through the Dijkstra algorithm according to the map data, the driving end point, and the driving starting point, presets a distance threshold, divides the optimal path according to the distance threshold to obtain a target point sequence, and determines the real-time target point according to the target point sequence and the real-time position data;

[0037] The data processing module is used to determine the vehicle's direction of travel according to the real-time position data and the real-time target point, preset a vehicle safety distance, determine a real-time vehicle safety domain according to the vehicle safety distance and the real-time position data, establish a coordinate system according to the real-time position data, divide the identification partition according to the coordinate system, obtain the obstacle size and real-time state quantity of the obstacle in the identification partition through the acquisition unit, add a timestamp according to the real-time state quantity to obtain a time series state quantity, and obtain the obstacle history trajectory according to the time series state quantity;

[0038] The avoidance decision module is used to calculate the predicted point coordinates through the prediction model according to the obstacle historical trajectory, obtain the real-time grid map through the on-board mapping unit, and calculate the avoidance decision through the planning model according to the real-time grid map, the vehicle travel direction, the predicted point coordinates, the real-time vehicle safety domain, the real-time state quantity, the obstacle size, and the real-time target point.

[0039] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the vehicle autonomous obstacle avoidance method based on vision and laser radar as described above is implemented.

[0040] A storage medium containing computer executable instructions, which, when executed by a computer processor, are used to perform any of the above-mentioned vehicle autonomous obstacle avoidance methods based on vision and laser radar.

[0041] The beneficial effects of the present invention are:

[0042] By setting up a vehicle-mounted photosensor to obtain the light intensity data of the external environment in real time, and adjusting the vehicle-mounted collection mode in time according to the light intensity data, it is possible to collect data through the visual sensor when the light intensity is good, and collect data through the lidar when the light intensity is weak. This avoids data errors caused by data collection through a single collection mode under different lighting environments, and also avoids the problem of long data processing time and slow obstacle avoidance response in the data collection mode that combines lidar and visual sensors. It improves the reaction speed of automatic obstacle avoidance of the car under the existing automobile hardware technology while improving the accuracy of collected data. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0044] Figure 1 The present invention is a flowchart of a method for autonomous obstacle avoidance of a vehicle based on vision and laser radar. DETAILED DESCRIPTION

[0045] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0046] See also Figure 1 , a vehicle autonomous obstacle avoidance method and system based on vision and laser radar,

[0047] The light intensity data is acquired by the vehicle-mounted photosensor, and the vehicle-mounted control unit generates an adaptive signal according to the light intensity data to control the acquisition unit to switch the acquisition mode;

[0048] The real-time position data of the vehicle is obtained through the vehicle positioning unit, and the map data, the driving end point, and the driving starting point are obtained. The vehicle motion planning unit calculates the optimal path through the Dijkstra algorithm according to the map data, the driving end point, and the driving starting point, presets a distance threshold, divides the optimal path according to the distance threshold to obtain a target point sequence, and determines the real-time target point according to the target point sequence and the real-time position data;

[0049] Determine the direction of vehicle travel according to the real-time position data and the real-time target point, preset a vehicle safety distance, determine a real-time vehicle safety domain according to the vehicle safety distance and the real-time position data, establish a coordinate system according to the real-time position data, divide the identification partition according to the coordinate system, obtain the obstacle size and real-time state quantity of the obstacle in the identification partition through the acquisition unit, add a timestamp according to the real-time state quantity to obtain a time series state quantity, and obtain the obstacle history trajectory according to the time series state quantity;

[0050] The predicted point coordinates are calculated by the prediction model according to the historical trajectory of the obstacle, the real-time grid map is obtained by the on-board mapping unit, and the avoidance decision is calculated by the planning model according to the real-time grid map, the vehicle travel direction, the predicted point coordinates, the real-time vehicle safety domain, the real-time state quantity, the obstacle size, and the real-time target point.

[0051] The acquisition mode includes a laser radar acquisition mode and a visual acquisition mode; the real-time target point is the target point with the smallest distance from the real-time position data in the target point sequence; the real-time state quantity includes the real-time coordinates, speed, and movement direction of the obstacle.

[0052] In this embodiment, the visual acquisition mode includes:

[0053] Image data is acquired through a visual sensor, preprocessing is performed according to the image data to obtain preprocessed data, feature extraction is performed according to the preprocessed data by an edge detection method to obtain image features, obstacle recognition is performed according to the image features by a convolutional neural network to obtain an obstacle recognition result, the speed and the moving direction are calculated according to the obstacle recognition result by an optical flow method, obstacle material points are determined according to the obstacle recognition result, pixel coordinates of the obstacle recognition result are converted by coordinates to obtain real-time coordinates of the obstacle, and the obstacle size is calculated according to the edge point coordinates of the obstacle recognition result.

[0054] Specifically, the coordinate transformation calculation formula is:

[0055] S*P i =A[R|T]P w ,

[0056] Where S is the distance between the obstacle and the camera, P i is the pixel coordinate, A is the camera intrinsic matrix, R is the rotation matrix, T is the translation vector, P w The real-time coordinates of obstacles.

[0057] It should be noted that the preprocessing includes noise reduction, data enhancement, grayscale, and binarization.

[0058] Specifically, the laser radar acquisition mode includes:

[0059] Point cloud data is acquired through a laser radar, and de-noised point cloud data is calculated based on the point cloud data through a through-filtering algorithm. A region of interest is calculated based on the de-noised point cloud data through an ROI algorithm. An obstacle point cloud is obtained by performing ground segmentation based on the region of interest through a plane fitting equation. A point cloud cluster is obtained based on the obstacle point cloud through Euclidean clustering. Point cloud features are calculated based on the point cloud cluster through a DGCNN algorithm. Two consecutive frames of obstacle point clouds are associated based on the point cloud features to obtain associated data. The obstacle size, the real-time coordinates of the obstacle, and the movement direction are obtained based on the associated data through an EKF filtering algorithm. The speed is calculated based on the real-time coordinates of the obstacle in two consecutive frames and a frame interval.

[0060] Specifically, the identified partitions include partition No. 1, partition No. 2, partition No. 3, and partition No. 4. Partition No. 1 is the [60°, 120°] interval area on the coordinate system, partition No. 2 is the [120°, 240°] interval area on the coordinate system, partition No. 3 is the [240°, 300°] interval area on the coordinate system, and partition No. 4 is the [300°, 60°] interval area on the coordinate system.

[0061] Specifically, the vehicle-mounted control unit generates an adaptive signal according to the light intensity data to control the acquisition unit to switch the acquisition mode, including:

[0062] A light intensity threshold is preset, and the light intensity data is judged according to the light intensity threshold. If the light intensity data is greater than the light intensity threshold, the acquisition mode is the visual acquisition mode; if the light intensity data is less than the light intensity threshold, the acquisition mode is the lidar acquisition mode.

[0063] It should be noted that both the laser radar acquisition mode and the visual acquisition mode are used to acquire the real-time coordinates of the obstacle, and the light intensity threshold is used to determine whether the vehicle driving environment is a bright environment.

[0064] Specifically, the specific calculation steps of the prediction model include:

[0065] S201: discretizing the obstacle history trajectory to obtain an attribute block, constructing a one-hot vector according to the attribute block, presetting a sparsity constraint, obtaining a high-dimensional embedding vector, and correlating the one-hot vector with the high-dimensional embedding vector according to the sparsity constraint to obtain an association vector;

[0066] It should be noted that the attribute block includes a horizontal coordinate attribute block, a vertical coordinate attribute block, a speed attribute block, and a motion direction attribute block;

[0067] S202: Calculate a temporal feature using a temporal convolution model according to the association vector, and calculate a spatial feature using a Transformer model according to the association vector;

[0068] S203: Calculate predicted time features and predicted space features through an adaptive pooling layer according to the time features and the space features, and perform feature fusion using a CNN-LSTM-Attention model on the predicted time features and the predicted space features to obtain predicted spatiotemporal features;

[0069] S204: Perform a pseudo-inverse operation on the predicted spatiotemporal features to obtain the predicted coordinate points.

[0070] The pseudo-inverse operation expression is:

[0071] X=(A T A) -1 A T b,

[0072] Wherein, X is the predicted coordinate point, A is the predicted spatiotemporal feature, and b is the mapping relationship vector.

[0073] Specifically, the specific calculation steps of the planning model are:

[0074] S301: determining a real-time coordinate according to the real-time state quantity, and determining an obstacle moving direction according to the real-time coordinate and the predicted point coordinate;

[0075] S302: judging according to the predicted point coordinates and the real-time vehicle safety domain, if the predicted point coordinates are within the real-time vehicle safety domain, executing step S303; if the predicted point coordinates are not within the real-time vehicle safety domain, executing the first decision;

[0076] S303: judging according to the obstacle moving direction and the vehicle moving direction, if the obstacle moving direction is parallel to the vehicle moving direction, executing the first decision; if the obstacle moving direction is not parallel to the vehicle moving direction, executing step S304;

[0077] S304: acquiring the obstacle coordinates according to the real-time state quantity, and determining an avoidance direction according to the obstacle coordinates and the real-time vehicle safety domain; if the obstacle coordinates are within the real-time vehicle safety domain, the avoidance direction is the obstacle travel direction; if the obstacle coordinates are not within the real-time vehicle safety domain, the avoidance direction is the opposite direction of the obstacle travel direction;

[0078] S305: Marking the real-time grid map according to the obstacle size and the real-time coordinates of the obstacle to obtain a marked obstacle map, and calculating a traversal result through a path traversal model according to the marked obstacle map, the real-time target point, the real-time location data, and the travel direction;

[0079] S306: Making a judgment based on the traversal result, if the traversal result is that the path exists, executing the second decision, and if the traversal result is that the path does not exist, executing the third decision.

[0080] Specifically, the avoidance decision includes the first decision, the second decision, and the third decision. The first decision is to maintain the current condition and continue to operate, the second decision is to operate according to the planned route, and the third decision is to park.

[0081] Specifically, the path traversal model specifically includes:

[0082] S401: Initialize the avoidance route, set the real-time target point as the traversal starting point, and set the real-time position data as the traversal end point;

[0083] S402: judging the pass factor of the grid in the obstacle map by a heuristic algorithm according to the traversal starting point and the avoidance direction; if the pass factor is 0, storing the grid in the avoidance route and executing step S403; if the pass factor is 1, outputting the traversal result as path non-existence;

[0084] S403: Determine whether the grid is the traversal end point. If the grid is the traversal end point, output the avoidance route and output the traversal result as path existence. If the grid is not the traversal end point, repeat steps S401-S403.

[0085] It should be noted that the heuristic algorithm includes but is not limited to a genetic algorithm, a particle swarm optimization algorithm, and an ant colony algorithm greedy algorithm.

[0086] A vehicle autonomous obstacle avoidance system based on vision and laser radar, comprising: a mode switching module, a data acquisition module, a data processing module, and an avoidance decision module;

[0087] The mode switching module is used to obtain light intensity data through the vehicle-mounted photosensor, and the vehicle-mounted control unit generates an adaptive signal according to the light intensity data to control the acquisition unit to switch the acquisition mode;

[0088] The data acquisition module is used to obtain the real-time position data of the vehicle through the vehicle positioning unit, obtain the map data, the driving end point, and the driving starting point, the vehicle motion planning unit calculates the optimal path through the Dijkstra algorithm according to the map data, the driving end point, and the driving starting point, presets a distance threshold, divides the optimal path according to the distance threshold to obtain a target point sequence, and determines the real-time target point according to the target point sequence and the real-time position data;

[0089] The data processing module is used to determine the vehicle's direction of travel according to the real-time position data and the real-time target point, preset a vehicle safety distance, determine a real-time vehicle safety domain according to the vehicle safety distance and the real-time position data, establish a coordinate system according to the real-time position data, divide the identification partition according to the coordinate system, obtain the obstacle size and real-time state quantity of the obstacle in the identification partition through the acquisition unit, add a timestamp according to the real-time state quantity to obtain a time series state quantity, and obtain the obstacle history trajectory according to the time series state quantity;

[0090] The avoidance decision module is used to calculate the predicted point coordinates through the prediction model according to the obstacle historical trajectory, obtain the real-time grid map through the on-board mapping unit, and calculate the avoidance decision through the planning model according to the real-time grid map, the vehicle travel direction, the predicted point coordinates, the real-time vehicle safety domain, the real-time state quantity, the obstacle size, and the real-time target point.

[0091] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the vehicle autonomous obstacle avoidance method based on vision and laser radar as described above is implemented.

[0092] A storage medium containing computer executable instructions, which, when executed by a computer processor, are used to perform any of the above-mentioned vehicle autonomous obstacle avoidance methods based on vision and laser radar.

[0093] The computer storage medium of the embodiment of the present invention may adopt any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device.

[0094] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0095] The program code included on the computer readable medium can be transmitted with any appropriate medium, including but not limited to wireless, electric wire, optical cable, RF, etc., or any suitable combination of the above. The computer program code for performing the operation of the present invention can be written in one or more programming languages ​​or their combinations, and the programming language includes object-oriented programming languages-such as Java, Smalltalk, C++, and also includes conventional procedural programming languages-such as "C" language or similar programming languages. The program code can be executed completely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on the remote computer, or completely on the remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (for example, using an Internet service provider to connect through the Internet).

[0096] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A vehicle autonomous obstacle avoidance method based on vision and laser radar, characterized in that: include: The light intensity data is collected by the vehicle-mounted photosensor, and the vehicle-mounted control unit generates an adaptive signal according to the light intensity data to control the collection unit to switch the collection mode; The real-time position data of the vehicle is obtained through the vehicle positioning unit, and the map data, the driving end point, and the driving starting point are obtained. The vehicle motion planning unit calculates the optimal path through the Dijkstra algorithm according to the map data, the driving end point, and the driving starting point, presets a distance threshold, divides the optimal path according to the distance threshold to obtain a target point sequence, and determines the real-time target point according to the target point sequence and the real-time position data; Determine the direction of vehicle travel according to the real-time position data and the real-time target point, preset a vehicle safety distance, determine a real-time vehicle safety domain according to the vehicle safety distance and the real-time position data, establish a coordinate system according to the real-time position data, divide the identification partition according to the coordinate system, obtain the obstacle size and real-time state quantity of the obstacle in the identification partition through the acquisition unit, add a timestamp according to the real-time state quantity to obtain a time series state quantity, and obtain the obstacle history trajectory according to the time series state quantity; The predicted point coordinates are calculated by the prediction model according to the historical trajectory of the obstacle, the real-time grid map is obtained by the on-board mapping unit, and the avoidance decision is calculated by the planning model according to the real-time grid map, the vehicle travel direction, the predicted point coordinates, the real-time vehicle safety domain, the real-time state quantity, the obstacle size, and the real-time target point.

2. The vehicle autonomous obstacle avoidance method based on vision and laser radar according to claim 1 is characterized in that: The acquisition mode includes a laser radar acquisition mode and a visual acquisition mode; the real-time target point is the target point with the smallest distance from the real-time position data in the target point sequence; the real-time state quantity includes the real-time coordinates, speed, and movement direction of the obstacle.

3. The vehicle autonomous obstacle avoidance method based on vision and laser radar according to claim 1 is characterized in that: The identified partitions include partition number one, partition number two, partition number three, and partition number four.

4. The vehicle autonomous obstacle avoidance method based on vision and laser radar according to claim 1 is characterized in that: The vehicle-mounted control unit generates an adaptive signal according to the light intensity data to control the acquisition unit to switch the acquisition mode, including: A light intensity threshold is preset, and the light intensity data is judged according to the light intensity threshold. If the light intensity data is greater than the light intensity threshold, the acquisition mode is the visual acquisition mode; if the light intensity data is less than the light intensity threshold, the acquisition mode is the lidar acquisition mode.

5. The vehicle autonomous obstacle avoidance method based on vision and laser radar according to claim 1 is characterized in that: The specific calculation steps of the prediction model include: S201: discretizing the obstacle history trajectory to obtain an attribute block, constructing a one-hot vector according to the attribute block, presetting a sparsity constraint, obtaining a high-dimensional embedding vector, and correlating the one-hot vector with the high-dimensional embedding vector according to the sparsity constraint to obtain an association vector; S202: Calculate a temporal feature using a temporal convolution model according to the association vector, and calculate a spatial feature using a Transformer model according to the association vector; S203: Calculate predicted time features and predicted space features through an adaptive pooling layer according to the time features and the space features, and perform feature fusion through a CNN-LSTM-Attention model according to the predicted time features and the predicted space features to obtain predicted spatiotemporal features; S204: Perform a pseudo-inverse operation on the predicted spatiotemporal features to obtain the predicted coordinate points.

6. The vehicle autonomous obstacle avoidance method based on vision and laser radar according to claim 1 is characterized in that: The specific calculation steps of the planning model are: S301: determining a real-time coordinate according to the real-time state quantity, and determining an obstacle moving direction according to the real-time coordinate and the predicted point coordinate; S302: judging according to the predicted point coordinates and the real-time vehicle safety domain, if the predicted point coordinates are within the real-time vehicle safety domain, executing step S303; if the predicted point coordinates are not within the real-time vehicle safety domain, executing the first decision; S303: judging according to the obstacle moving direction and the vehicle moving direction, if the obstacle moving direction is parallel to the vehicle moving direction, executing the first decision; if the obstacle moving direction is not parallel to the vehicle moving direction, executing step S304; S304: acquiring the real-time coordinates of the obstacle according to the real-time state quantity, and determining an avoidance direction according to the real-time coordinates of the obstacle and the real-time vehicle safety domain; if the obstacle coordinates are within the real-time vehicle safety domain, the avoidance direction is the direction in which the obstacle moves; if the obstacle coordinates are not within the real-time vehicle safety domain, the avoidance direction is the opposite direction of the direction in which the obstacle moves; S305: Marking the real-time grid map according to the obstacle size and the real-time coordinates of the obstacle to obtain a marked obstacle map, and calculating a traversal result through a path traversal model according to the marked obstacle map, the real-time target point, the real-time location data, and the travel direction, wherein the traversal result includes whether a path exists or not; S306: Making a judgment based on the traversal result, if the traversal result is that the path exists, executing the second decision; if the traversal result is that the path does not exist, executing the third decision.

7. The vehicle autonomous obstacle avoidance method based on vision and laser radar according to claim 6 is characterized in that: The path traversal model specifically includes: S401: Initialize the avoidance route, set the real-time target point as the traversal starting point, and set the real-time position data as the traversal end point; S402: judging the pass factor of the grid in the obstacle map by a heuristic algorithm according to the traversal starting point and the avoidance direction; if the pass factor is 0, storing the grid in the avoidance route and executing step S403; if the pass factor is 1, outputting the traversal result as path non-existence; S403: Determine whether the grid is the traversal end point. If the grid is the traversal end point, output the avoidance route and output the traversal result as path existence. If the grid is not the traversal end point, repeat steps S401-S403.

8. A vehicle autonomous obstacle avoidance system based on vision and laser radar, characterized in that: include: Mode switching module, data acquisition module, data processing module, avoidance decision module; The mode switching module is used to obtain light intensity data through the vehicle-mounted photosensor, and the vehicle-mounted control unit generates an adaptive signal according to the light intensity data to control the acquisition unit to switch the acquisition mode; The data acquisition module is used to obtain the real-time position data of the vehicle through the vehicle positioning unit, obtain the map data, the driving end point, and the driving starting point, the vehicle motion planning unit calculates the optimal path through the Dijkstra algorithm according to the map data, the driving end point, and the driving starting point, presets a distance threshold, divides the optimal path according to the distance threshold to obtain a target point sequence, and determines the real-time target point according to the target point sequence and the real-time position data; The data processing module is used to determine the vehicle's direction of travel according to the real-time position data and the real-time target point, preset a vehicle safety distance, determine a real-time vehicle safety domain according to the vehicle safety distance and the real-time position data, establish a coordinate system according to the real-time position data, divide the identification partition according to the coordinate system, obtain the obstacle size and real-time state quantity of the obstacle in the identification partition through the acquisition unit, add a timestamp according to the real-time state quantity to obtain a time series state quantity, and obtain the obstacle history trajectory according to the time series state quantity; The avoidance decision module is used to calculate the predicted point coordinates through the prediction model according to the obstacle historical trajectory, obtain the real-time grid map through the on-board mapping unit, and calculate the avoidance decision through the planning model according to the real-time grid map, the vehicle travel direction, the predicted point coordinates, the real-time vehicle safety domain, the real-time state quantity, the obstacle size, and the real-time target point.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the vehicle autonomous obstacle avoidance method based on vision and lidar as described in any one of claims 1-7 is implemented.

10. A storage medium containing computer executable instructions, characterized in that: The computer executable instructions, when executed by a computer processor, are used to execute the vehicle autonomous obstacle avoidance method based on vision and lidar as described in any one of claims 1 to 7.

Images