Unmanned vehicle autonomous obstacle avoidance method and system, unmanned vehicle and storage medium
By using sensor detection and deep learning algorithms to identify dynamic obstacles and generate replanning paths, the problem of obstacle avoidance lag in complex environments for unmanned vehicles is solved, enabling them to accurately reach the target location.
Patent Information
- Application Number
- CN202510870719.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-26
- Publication Date
- 2025-10-31
AI Technical Summary
Unmanned vehicles cannot effectively identify and predict the movement trends of dynamic obstacles in complex road environments, resulting in delayed obstacle avoidance measures, safety hazards, and inability to accurately reach the target location.
By detecting environmental information and vehicle motion data through sensors, dynamic obstacles are identified using deep learning object detection algorithms and data fusion technology, a replanning path is generated, and path optimization is performed by combining dynamic constraints.
It improves the obstacle avoidance capabilities of unmanned vehicles in complex environments, ensures accurate arrival at the target location, and reduces the risk of delayed obstacle avoidance measures.
Smart Images
Figure CN120871846A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of unmanned vehicles, and more particularly to an autonomous obstacle avoidance method, system, unmanned vehicle, and storage medium for unmanned vehicles. Background Technology
[0002] Unmanned vehicles have been initially applied in urban logistics and park delivery. Currently, common control methods for unmanned vehicles are generally based on static maps and simple control rules. In complex road environments, such as when temporary obstacles, road construction, or pedestrians suddenly appear, there will be response lags. In addition, unmanned vehicles lack flexibility. Furthermore, traditional obstacle avoidance and path planning methods often do not deeply model the future movement trends of obstacles and cannot identify changes in obstacle trajectories, resulting in delayed obstacle avoidance measures and safety hazards. This can cause unmanned vehicles to stop abnormally on the road and fail to reach the target location.
[0003] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention
[0004] The main objective of this invention is to provide an autonomous obstacle avoidance method, system, unmanned vehicle, and storage medium for unmanned vehicles, aiming to improve the obstacle avoidance capability of unmanned vehicles and thereby increase the probability of reaching the target location.
[0005] To achieve the above objectives, the present invention provides an autonomous obstacle avoidance method for unmanned vehicles, the method comprising the following steps:
[0006] The control sensors detect the environmental information of the unmanned vehicle at its current location and the first motion data of the unmanned vehicle.
[0007] Based on the environmental information and the target detection algorithm, dynamic obstacles and their second motion data are determined, wherein the dynamic obstacles are those that are not currently stationary.
[0008] A replanning path is generated based on the first motion data, the second motion data, the target endpoint, and the dynamic constraints.
[0009] Optionally, the environmental information includes: point cloud data collected by lidar, image data collected by camera, and echo data collected by millimeter-wave radar. The step of determining the dynamic obstacle and its second motion data based on the environmental information and target detection algorithm includes:
[0010] The target fused data is obtained by fusing the point cloud data, the image data, and the echo data.
[0011] The dynamic obstacle is identified based on the target fusion data and the target detection algorithm;
[0012] The second motion data is determined based on the fusion data of the dynamic obstacles and the target.
[0013] Optionally, the target detection algorithm is a deep learning target detection algorithm, and the step of identifying the dynamic obstacle based on the target fusion data and the target detection algorithm includes:
[0014] The target fusion data is identified according to the deep learning target detection algorithm, and the object recognition result is output, which includes: object type, boundary coordinates and confidence score.
[0015] Dynamic obstacles are determined based on the object recognition results.
[0016] Optionally, the number of target fusion data is multiple, and the step of determining the second motion data based on the dynamic obstacle and the target fusion data includes:
[0017] Real-time motion data is determined based on the positional change information of the dynamic obstacle in the fused data of multiple targets;
[0018] The corresponding motion prediction model is determined based on the obstacle type of the dynamic obstacle;
[0019] Predicted motion data is determined based on the motion prediction model and the real-time motion data;
[0020] The real-time motion data and the predicted motion data are used as the second motion data.
[0021] Optionally, after determining the second motion data based on the fused data of the dynamic obstacle and the target, the method further includes:
[0022] A deep learning model is trained based on the positional change information of the dynamic obstacle to obtain a candidate motion prediction model;
[0023] The motion prediction model is adjusted based on the candidate motion prediction model.
[0024] Optionally, before the step of generating a replanning path based on the first motion data, the second motion data, the target endpoint, and the dynamic constraints, the method further includes:
[0025] The corresponding obstacle identification result is determined based on the first motion data, the second motion data, and the original planned path;
[0026] When the obstruction identification result indicates that the original planned path is blocked, the step of generating a replanned path is determined by using the first motion data, the second motion data, the target endpoint, and the dynamic constraints.
[0027] Optionally, the step of generating a replanning path based on the first motion data, the second motion data, the target endpoint, and the dynamic constraints includes:
[0028] Generate a path map model based on the first motion data and the second motion data;
[0029] The target endpoint is mapped to the path map model to determine the map endpoint, which corresponds to the target endpoint;
[0030] Based on the dynamic constraints, a first path decision model corresponding to path search is generated in the path map model;
[0031] The replanning path is determined based on the path map model, the map endpoint, the search algorithm, and the first path decision model.
[0032] Furthermore, to achieve the above objectives, the present invention also provides an autonomous obstacle avoidance system for unmanned vehicles, the autonomous obstacle avoidance system for unmanned vehicles comprising:
[0033] The sensor module is used to control the environmental information of the unmanned vehicle at its current location and the first motion data of the unmanned vehicle detected by the sensors.
[0034] The perception module is used to determine dynamic obstacles and their second motion data based on the environmental information and the target detection algorithm, wherein the dynamic obstacles are obstacles that are not currently stationary.
[0035] The path decision module is used to generate a replanning path based on the first motion data, the second motion data, the target endpoint, and dynamic constraints.
[0036] In addition, to achieve the above objectives, the present invention also provides an unmanned vehicle, the unmanned vehicle comprising: a memory, a processor, and an unmanned vehicle autonomous obstacle avoidance program stored in the memory and executable on the processor, the unmanned vehicle autonomous obstacle avoidance program being configured to implement the steps of the unmanned vehicle autonomous obstacle avoidance method described in any of the above claims.
[0037] In addition, to achieve the above objectives, the present invention also provides a storage medium storing an autonomous obstacle avoidance program for unmanned vehicles, wherein when the autonomous obstacle avoidance program for unmanned vehicles is executed by a processor, it implements the steps of the autonomous obstacle avoidance method for unmanned vehicles described in any of the above claims.
[0038] This invention proposes an autonomous obstacle avoidance method for unmanned vehicles. This method utilizes control sensors to detect environmental information at the current location of the unmanned vehicle and its first motion data. Based on the environmental information and a target detection algorithm, it determines dynamic obstacles and their second motion data. Compared to unmanned vehicle control methods based on static maps, this method can effectively detect dynamic objects in the environment. Furthermore, it generates a replanning path based on the first motion data, the second motion data, the target endpoint, and dynamic constraints. This replanning path enables obstacle avoidance, preventing the vehicle from failing to reach its target location due to obstacles. Attached Figure Description
[0039] Figure 1 This is a schematic diagram of the structure of the unmanned vehicle in the hardware operating environment involved in the embodiments of the present invention;
[0040] Figure 2 This is a flowchart illustrating the first embodiment of the autonomous obstacle avoidance method for unmanned vehicles of the present invention.
[0041] Figure 3 This is a flowchart illustrating the second embodiment of the autonomous obstacle avoidance method for unmanned vehicles of the present invention;
[0042] Figure 4 This is a flowchart illustrating the third embodiment of the autonomous obstacle avoidance method for unmanned vehicles of the present invention.
[0043] Figure 5 This is a flowchart illustrating the fourth embodiment of the autonomous obstacle avoidance method for unmanned vehicles of the present invention.
[0044] Figure 6 This is a system example diagram of the autonomous obstacle avoidance system for unmanned vehicles of the present invention.
[0045] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation
[0046] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
[0047] Reference Figure 1 , Figure 1 This is a schematic diagram of the unmanned vehicle structure in the hardware operating environment of the embodiment of the present invention.
[0048] like Figure 1As shown, the unmanned vehicle may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, an interaction device 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The interaction device 1003 may include a display screen or an input unit such as a keyboard. Optionally, the interaction device 1003 may also be connected to the communication bus via standard wired or wireless interfaces. The network interface 1004 may optionally include standard wired or wireless interfaces (such as a Wi-Fi interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.
[0049] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on autonomous vehicles and may include more or fewer components than shown, or combine certain components, or have different component arrangements.
[0050] like Figure 1 As shown, the memory 1005, which serves as a storage medium, may include an operating system, a data storage module, a network communication module, a user interface module, and an autonomous obstacle avoidance program for unmanned vehicles.
[0051] exist Figure 1 In the unmanned vehicle shown, the network interface 1004 is mainly used for data communication with other devices; the interaction device 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the unmanned vehicle of the present invention can be set in the unmanned vehicle, and the unmanned vehicle calls the unmanned vehicle autonomous obstacle avoidance program stored in the memory 1005 through the processor 1001 and executes the unmanned vehicle autonomous obstacle avoidance method provided in the embodiment of the present invention.
[0052] This invention provides an autonomous obstacle avoidance method for unmanned vehicles, referring to... Figure 2 , Figure 2 This is a flowchart illustrating the first embodiment of an autonomous obstacle avoidance method for unmanned vehicles according to the present invention.
[0053] In this embodiment, the autonomous obstacle avoidance method for unmanned vehicles includes:
[0054] Step S1: Control the environmental information of the unmanned vehicle at its current location and the first motion data of the unmanned vehicle detected by the control sensor;
[0055] In this embodiment, the sensors may include one or more of a visual sensor, LiDAR, and millimeter-wave radar. The number of each type of sensor is not limited. Specifically, the detected environmental information is the environmental information of the current location collected by the aforementioned sensors, which may include point cloud data, image data, and echo data. Before the sensors collect environmental information, a correction algorithm is used to correct the collected environmental data. Furthermore, the speed, direction, and acceleration data of the unmanned vehicle during its operation are also recorded.
[0056] Step S2: Determine the dynamic obstacle and its second motion data based on the environmental information and the target detection algorithm. The dynamic obstacle is an obstacle that is not currently stationary.
[0057] Standardization involves collecting point cloud data, image data, and echo data from sensors, and fusing these data. Specifically, it determines the correspondence between the point cloud data, image data, and echo data, and identifies obstacle types using target detection algorithms. Since sensor data is generally continuous, there can be multiple pieces of environmental information, each corresponding to a specific moment. Sensors typically have a fixed data acquisition frequency, resulting in multiple pieces of environmental information. Therefore, dynamic obstacles are identified by comparing changes in obstacle position. It should be noted that in this embodiment, buildings and traffic facilities, such as traffic lights and green belts, can be directly ruled out as dynamic obstacles. For vehicles, multiple pieces of environmental information are needed to determine if they are dynamic obstacles. For pedestrians, pedestrians can be considered as dynamic obstacles.
[0058] Step S3: Generate a replanning path based on the first motion data, the second motion data, the target endpoint, and the dynamic constraints.
[0059] In this embodiment, it is determined whether the currently executed original planned path will be blocked based on the first motion data and the second motion data. When an obstruction is detected, a replanned path is generated based on the first motion data, the second motion data, the target endpoint, and dynamic constraints. Optionally, reinforcement learning can be used to generate the replanned path based on the first motion data, the second motion data, the target endpoint, and dynamic constraints. After obtaining the replanned path, the unmanned vehicle is controlled to adjust its real-time operating speed, direction, and acceleration according to the replanned path, so that the unmanned vehicle travels according to the replanned path.
[0060] In this embodiment, the system controls the environmental information of the unmanned vehicle at its current location and the vehicle's first motion data detected by sensors. Based on the environmental information and a target detection algorithm, dynamic obstacles and their second motion data are determined. Compared to unmanned vehicle control methods based on static maps, this approach effectively detects dynamic objects in the environment. Furthermore, a replanning path is generated based on the first motion data, the second motion data, the target endpoint, and dynamic constraints. This replanning path enables obstacle avoidance, preventing the vehicle from failing to reach its target location due to obstacles.
[0061] Furthermore, based on the first embodiment, a second embodiment of the autonomous obstacle avoidance method for unmanned vehicles of the present invention is proposed. In this embodiment, reference is made to... Figure 3 The environmental information includes: point cloud data collected by lidar, image data collected by camera, and echo data collected by millimeter-wave radar. The step of determining the dynamic obstacle and the second motion data of the dynamic obstacle based on the environmental information and the target detection algorithm includes:
[0062] Step S21: Perform data fusion based on the point cloud data, the image data, and the echo data to obtain target fused data;
[0063] Optionally, a feature-based data fusion method specifically extracts data features from point cloud data, image data, and echo data respectively. These data features can be: shape features of point cloud data, texture and color features of image data, and echo intensity variation features of echo data. By concatenating these data features, a corresponding vector is generated, storing the corresponding feature data. Commonly, the fused data can also be mapped to the corresponding point cloud data, with each point cloud coordinate associated with data such as color, texture, and echo intensity. This embodiment does not limit the data fusion method. In other embodiments, a decision-level fusion method can be used to fuse the point cloud data, the image data, and the echo data.
[0064] Step S22: Identify the dynamic obstacle based on the target fusion data and the target detection algorithm;
[0065] Optionally, the target detection algorithm here can be a template matching method. Specifically, the target fusion data is segmented to obtain multiple groups of fusion data. Each group of fusion data is matched with preset data, and the type of obstacle is determined based on the matching result. For example, pedestrians are directly identified as dynamic obstacles, while buildings are identified as static obstacles.
[0066] Step S23: Determine the second motion data based on the fused data of the dynamic obstacle and the target.
[0067] Optionally, the target fusion data corresponds to a time point, so the second motion data can be calculated from multiple target fusion data. It should be noted that the data type of the second motion data is the same as that of the first motion data. However, since the acquisition methods of the first and second motion data are different, the precision of the first and second motion data may be different.
[0068] In this embodiment, target fusion data is obtained by fusing the point cloud data, the image data, and the echo data; the dynamic obstacle is identified based on the target fusion data and the target detection algorithm; and the second motion data is determined based on the dynamic obstacle and the target fusion data. This allows for accurate identification of dynamic obstacles from the data detected by the sensors, thereby achieving the identification and detection of moving objects and improving the accuracy of subsequent obstacle avoidance functions.
[0069] Furthermore, the target detection algorithm is a deep learning target detection algorithm, and the step of identifying the dynamic obstacle based on the target fusion data and the target detection algorithm includes:
[0070] The target fusion data is identified according to the deep learning target detection algorithm, and the object recognition result is output, which includes: object type, boundary coordinates and confidence score.
[0071] Dynamic obstacles are determined based on the object recognition results.
[0072] The method is not limited to the type of deep learning-based object detection algorithm. It involves inputting fused data of the target and outputting the recognition result. Commonly, YOLOv5 or Faster R-CNN can be used. The method identifies obstacles and determines whether they are dynamic obstacles based on the identification results. Optionally, the method determines whether an object is a dynamic obstacle based on its type.
[0073] In this embodiment, the target fusion data is identified according to the deep learning target detection algorithm, and the object recognition result is output. The object recognition result includes: object type, boundary coordinates and confidence level. Dynamic obstacles are determined based on the object recognition result, thereby enabling the identification of multiple obstacle data and efficiently determining the obstacles present in the scene.
[0074] Furthermore, based on the first or second embodiment, a third embodiment of the autonomous obstacle avoidance method for unmanned vehicles of the present invention is proposed. In this embodiment, referring to FIG4, the number of target fusion data is multiple, and the step of determining the second motion data based on the dynamic obstacle and the target fusion data includes:
[0075] Step S231: Determine real-time motion data based on the position change information of the dynamic obstacle in the multiple target fusion data;
[0076] Optionally, a position coordinate system is constructed based on the target fusion data, and the dynamic obstacle is mapped to the position coordinate system. Specifically, the real-time motion data can be represented in the position coordinate system.
[0077] Step S232: Determine the corresponding motion prediction model based on the obstacle type of the dynamic obstacle;
[0078] In this embodiment, a corresponding motion prediction model is determined based on different dynamic obstacles. This motion prediction model can receive motion data and predict its motion over a period of time. For example, different motion prediction models exist for pedestrians, bicycles, and cars.
[0079] Step S233: Determine the predicted motion data based on the motion prediction model and the real-time motion data;
[0080] Specifically, the real-time motion data is input into the corresponding motion prediction model, and a corresponding prediction result is generated. The predicted motion data is then determined based on the prediction result.
[0081] Step S234: Use the real-time motion data and the predicted motion data as the second motion data.
[0082] In this embodiment, real-time motion data is determined by the position change information of the dynamic obstacle in multiple target fusion data; and a corresponding motion prediction model is determined according to the obstacle type of the dynamic obstacle; different predictions can be made for different types of obstacles, and predicted motion data is determined according to the motion prediction model and the real-time motion data, and the real-time motion data and the predicted motion data are used as the second motion data.
[0083] Furthermore, after the step of determining the second motion data based on the fused data of the dynamic obstacle and the target, the method further includes:
[0084] A deep learning model is trained based on the positional change information of the dynamic obstacle to obtain a candidate motion prediction model;
[0085] The motion prediction model is adjusted based on the candidate motion prediction model.
[0086] Specifically, when the model evaluation index of the candidate motion prediction model is better than that of the motion prediction model, the candidate motion prediction model is adopted as the motion prediction model.
[0087] Furthermore, based on any of the above embodiments, a fourth embodiment of the autonomous obstacle avoidance method for unmanned vehicles of the present invention is proposed. In this embodiment, reference is made to... Figure 5 The step of generating a replanning path based on the first motion data, the second motion data, the target endpoint, and the dynamic constraints includes:
[0088] Step S31: Generate a path map model based on the first motion data and the second motion data;
[0089] In this embodiment, the path map model can be the position coordinate system of the third embodiment. Preferably, the position coordinate system can be constructed based on point cloud data to obtain the path map model. The path area is divided into the smallest units according to a preset size. Furthermore, it should be noted that a limiting condition is set for the smallest unit occupied by the movement of the dynamic obstacle based on the second motion data; the limiting condition can be time.
[0090] Optionally, the smallest unit is a 1-meter cube, which can also be set according to the size of the unmanned vehicle.
[0091] Step S32: Map the target endpoint to the path map model to determine the map endpoint, which corresponds to the target endpoint;
[0092] Step S33: Generate a first path decision model corresponding to path search in the path map model based on the dynamic constraints;
[0093] The dynamic constraint here is the turning radius, which represents all reachable minimum units based on the current position and the autonomous vehicle's direction of movement during path search. For example, conventional path search would not find the minimum unit adjacent to the autonomous vehicle in a direction perpendicular to its current direction of movement. In this embodiment, the search range can be effectively reduced.
[0094] Step S34: Determine the replanning path based on the path map model, map endpoint, search algorithm, and the first path decision model.
[0095] In this embodiment, the search algorithm can be a breadth-first search algorithm, or an A* or DWA (Dynamic Window Approach) algorithm. In this embodiment, the time stamp is updated at each step during the search. Some minimum units have a defined passable time range. The time stamp needs to be compared with this passable time range, and the comparison result determines whether the minimum unit is passable. Specifically, based on the path map model, map endpoint, and search algorithm, when multiple paths are found, the replanning path is determined from among the multiple paths according to the first path decision model.
[0096] Furthermore, before the step of generating a replanning path based on the first motion data, the second motion data, the target endpoint, and the dynamic constraints, the method further includes:
[0097] The corresponding obstacle identification result is determined based on the first motion data, the second motion data, and the original planned path;
[0098] When the obstruction identification result indicates that the original planned path is blocked, the step of generating a replanned path is determined by using the first motion data, the second motion data, the target endpoint, and the dynamic constraints.
[0099] Furthermore, this invention also proposes an autonomous obstacle avoidance system for unmanned vehicles, the autonomous obstacle avoidance system for unmanned vehicles comprising:
[0100] The sensor module is used to control the environmental information of the unmanned vehicle at its current location and the first motion data of the unmanned vehicle detected by the sensors.
[0101] The perception module is used to determine dynamic obstacles and their second motion data based on the environmental information and the target detection algorithm, wherein the dynamic obstacles are obstacles that are not currently stationary.
[0102] The path decision module is used to generate a replanning path based on the first motion data, the second motion data, the target endpoint, and dynamic constraints.
[0103] In addition, a control and execution module is provided to handle the actual movement of the vehicle. Alternatively, refer to... Figure 6 The autonomous obstacle avoidance system for unmanned vehicles may include hardware such as lidar, cameras, and millimeter-wave radar.
[0104] Furthermore, this invention also proposes an unmanned vehicle, which includes: a memory, a processor, and an unmanned vehicle autonomous obstacle avoidance program stored in the memory and executable on the processor. The unmanned vehicle autonomous obstacle avoidance program is configured to implement the steps of any of the above-described embodiments of the unmanned vehicle autonomous obstacle avoidance method.
[0105] Furthermore, this embodiment of the invention also proposes a storage medium storing an autonomous obstacle avoidance program for unmanned vehicles, wherein when the autonomous obstacle avoidance program for unmanned vehicles is executed by a processor, it implements the steps of any of the embodiments of the autonomous obstacle avoidance method for unmanned vehicles described above.
[0106] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.
[0107] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.
[0108] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.
[0109] The above are merely preferred embodiments of the present invention and do not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.
Claims
1. A method for autonomous obstacle avoidance by unmanned vehicles, characterized in that, The autonomous obstacle avoidance method for unmanned vehicles includes the following steps: The control sensors detect the environmental information of the unmanned vehicle at its current location and the first motion data of the unmanned vehicle. Based on the environmental information and the target detection algorithm, dynamic obstacles and their second motion data are determined, wherein the dynamic obstacles are those that are not currently stationary. A replanning path is generated based on the first motion data, the second motion data, the target endpoint, and the dynamic constraints.
2. The autonomous obstacle avoidance method for unmanned vehicles as described in claim 1, characterized in that, The environmental information includes: point cloud data collected by lidar, image data collected by camera, and echo data collected by millimeter-wave radar. The step of determining the dynamic obstacle and the second motion data of the dynamic obstacle based on the environmental information and the target detection algorithm includes: The target fused data is obtained by fusing the point cloud data, the image data, and the echo data. The dynamic obstacle is identified based on the target fusion data and the target detection algorithm; The second motion data is determined based on the fusion data of the dynamic obstacles and the target.
3. The autonomous obstacle avoidance method for unmanned vehicles as described in claim 2, characterized in that, The target detection algorithm is a deep learning target detection algorithm, and the step of identifying the dynamic obstacle based on the target fusion data and the target detection algorithm includes: The target fusion data is identified according to the deep learning target detection algorithm, and the object recognition result is output, which includes: object type, boundary coordinates and confidence score. Dynamic obstacles are determined based on the object recognition results.
4. The autonomous obstacle avoidance method for unmanned vehicles as described in claim 2, characterized in that, The number of target fusion data is multiple, and the step of determining the second motion data based on the dynamic obstacle and the target fusion data includes: Real-time motion data is determined based on the positional change information of the dynamic obstacle in the fused data of multiple targets; The corresponding motion prediction model is determined based on the obstacle type of the dynamic obstacle; Predicted motion data is determined based on the motion prediction model and the real-time motion data; The real-time motion data and the predicted motion data are used as the second motion data.
5. The autonomous obstacle avoidance method for unmanned vehicles as described in claim 4, characterized in that, After the step of determining the second motion data based on the fused data of the dynamic obstacle and the target, the method further includes: A deep learning model is trained based on the positional change information of the dynamic obstacle to obtain a candidate motion prediction model; The motion prediction model is adjusted based on the candidate motion prediction model.
6. The autonomous obstacle avoidance method for unmanned vehicles as described in claim 1, characterized in that, Before the step of generating a replanning path based on the first motion data, the second motion data, the target endpoint, and the dynamic constraints, the method further includes: The corresponding obstacle identification result is determined based on the first motion data, the second motion data, and the original planned path; When the obstruction identification result indicates that the original planned path is blocked, the step of generating a replanned path is determined by using the first motion data, the second motion data, the target endpoint, and the dynamic constraints.
7. The autonomous obstacle avoidance method for unmanned vehicles as described in any one of claims 1 to 6, characterized in that, The step of generating a replanning path based on the first motion data, the second motion data, the target endpoint, and the dynamic constraints includes: Generate a path map model based on the first motion data and the second motion data; The target endpoint is mapped to the path map model to determine the map endpoint, which corresponds to the target endpoint; Based on the dynamic constraints, a first path decision model corresponding to path search is generated in the path map model; The replanning path is determined based on the path map model, the map endpoint, the search algorithm, and the first path decision model.
8. An autonomous obstacle avoidance system for unmanned vehicles, characterized in that, The autonomous obstacle avoidance system for unmanned vehicles includes: The sensor module is used to control the environmental information of the unmanned vehicle at its current location and the first motion data of the unmanned vehicle detected by the sensors. The perception module is used to determine dynamic obstacles and their second motion data based on the environmental information and the target detection algorithm, wherein the dynamic obstacles are obstacles that are not currently stationary. The path decision module is used to generate a replanning path based on the first motion data, the second motion data, the target endpoint, and dynamic constraints.
9. An unmanned vehicle, characterized in that, The unmanned vehicle includes: a memory, a processor, and an unmanned vehicle autonomous obstacle avoidance program stored in the memory and executable on the processor, the unmanned vehicle autonomous obstacle avoidance program being configured to implement the steps of the unmanned vehicle autonomous obstacle avoidance method as described in any one of claims 1 to 7.
10. A storage medium, characterized in that, The storage medium stores an autonomous obstacle avoidance program for unmanned vehicles, which, when executed by a processor, implements the steps of the autonomous obstacle avoidance method for unmanned vehicles as described in any one of claims 1 to 7.