Method for emergency obstacle avoidance of a robot during ball picking

Abstract

This application relates to the field of robot control technology and discloses a method for handling emergency obstacle avoidance during ball retrieval by a robot. The method includes real-time obstacle detection to obtain obstacle categories, obtaining different obstacle avoidance levels based on obstacle categories, obtaining different obstacle avoidance strategies based on obstacle avoidance levels, and performing emergency obstacle avoidance during ball retrieval based on the obstacle avoidance strategies. This application classifies obstacles and dynamically adjusts the obstacle avoidance level based on distance, speed, direction, and category, making risk assessment more closely aligned with real-world scenarios. Through hierarchical obstacle avoidance strategies and multi-obstacle priority ranking, it achieves targeted and efficient continuous obstacle avoidance, avoiding over-avoidance or response lag. Dynamic updates of the local cost map and a behavior tree-driven task recovery mechanism enhance the robot's adaptability and robustness in complex environments, ensuring rapid recovery of the ball retrieval task after obstacle avoidance, significantly improving task execution efficiency and autonomous operation reliability.

Description

Technical Field

[0001] This application relates to the field of robot control technology, specifically a method for handling emergency obstacle avoidance by a robot during ball retrieval. Background Technology

[0002] During the autonomous ball-picking task, robots often need to deal with various obstacles in dynamic environments, and their obstacle avoidance capabilities directly affect task efficiency and operational safety. In existing technologies, robot obstacle avoidance mostly relies on a single sensor for obstacle detection, and the classification of obstacles is relatively simple, such as only distinguishing between static and dynamic obstacles. It is difficult to accurately identify special obstacles such as slow-moving, high-speed moving, and human-shaped obstacles, resulting in insufficient targeting of obstacle avoidance strategies.

[0003] At the same time, the existing obstacle avoidance level classification is crude and does not combine obstacle categories for dynamic risk assessment, which can easily lead to problems such as over-avoidance or untimely obstacle avoidance. In particular, there is a lack of special response mechanisms for objects that should be avoided first, such as human-shaped obstacles.

[0004] In addition, the poor connection between path replanning after obstacle avoidance and the resumption of the original task often leads to the interruption of the ball retrieval task, reducing the overall execution efficiency.

[0005] Chinese invention patent application CN120386360A discloses a method, device, equipment and medium for obstacle avoidance of sanitation robots in complex scenes, but the obstacle avoidance performance of this patent application is not good when responding to different obstacles.

[0006] In conclusion, there is an urgent need for a new technical solution for robots to avoid obstacles in the process of picking up balls. Summary of the Invention

[0007] The purpose of this application is to provide a method for emergency obstacle avoidance by a robot during ball retrieval, so as to solve the technical problems mentioned in the background art.

[0008] To achieve the above objectives, this application discloses the following technical solution: a method for handling emergency obstacle avoidance by a robot during ball retrieval, the method comprising:

[0009] Multiple obstacle detection devices are used to detect obstacles in front of the ball in real time during the ball retrieval process and perform data fusion to obtain obstacle categories, which include stationary obstacles, slow-moving obstacles, high-speed moving obstacles, and humanoid obstacles.

[0010] Based on the results of real-time detection of obstacles in front during the ball retrieval process and the obstacle category, different obstacle avoidance levels are obtained, including low-risk, medium-risk, and high-risk levels; wherein, a basic obstacle avoidance level is obtained based on the results of real-time detection of obstacles in front during the ball retrieval process, and then the basic obstacle avoidance level is adjusted based on the obstacle category to obtain the final obstacle avoidance level.

[0011] Based on the obstacle avoidance level, different obstacle avoidance strategies are obtained. Each obstacle avoidance strategy responds to each obstacle in front during the ball retrieval process, and there is a corresponding obstacle avoidance sub-strategy for each obstacle. The obstacle avoidance strategy is obtained based on the obstacle avoidance sub-strategy. Based on the obstacle avoidance strategy, emergency obstacle avoidance is performed during the ball retrieval process.

[0012] Preferably, the obstacle detection device includes a lidar and a camera;

[0013] When the obstacle detection device detects obstacles in front of the ball during the ball-picking process in real time, it performs multi-source data fusion on the data collected by the lidar and camera to obtain the obstacle category.

[0014] Preferably, the process of fusing multi-source data from the LiDAR and camera data to obtain the obstacle category includes:

[0015] Based on the data collected by lidar, information on the distance, outline, and speed of movement of obstacles is obtained;

[0016] Based on data collected by the camera, visual features and shape information of obstacles are obtained;

[0017] The data collected by LiDAR and cameras will be processed and subjected to spatiotemporal matching and data fusion to obtain obstacle categories.

[0018] Preferably, the different obstacle avoidance levels obtained based on the obstacle category include:

[0019] A basic obstacle avoidance level is obtained based on the distance, speed, and direction of movement of the obstacle;

[0020] When the obstacle category is the humanoid obstacle, the risk level is increased by one level based on the basic obstacle avoidance level;

[0021] When the obstacle category is a high-speed moving obstacle and the basic obstacle avoidance level is the low-risk level, it is adjusted to the medium-risk level;

[0022] When the obstacle category is either a stationary obstacle or a slow-moving obstacle, the basic obstacle avoidance level remains unchanged.

[0023] Preferably, the basic obstacle avoidance level is obtained based on the distance, speed, and direction of movement of the obstacle, including:

[0024] Obtain a preset distance range, and determine a distance factor based on the distance range in which the actual distance between the obstacle and the robot falls;

[0025] Obtain a preset speed range, and determine a speed factor based on the speed range in which the actual speed of the obstacle falls;

[0026] Obtain a preset directional angle range, and determine the direction factor based on the actual angle between the obstacle's movement direction and the robot's travel direction within the directional angle range;

[0027] A comprehensive score is obtained by weighting the distance factor, the speed factor, and the direction factor using a preset weighting model. Based on the preset score range in which the comprehensive score falls, the basic obstacle avoidance level is determined to be the low-risk level, the medium-risk level, or the high-risk level.

[0028] Preferably, the different obstacle avoidance strategies derived based on the obstacle avoidance level include:

[0029] When the obstacle avoidance level is the low-risk level, the obstacle avoidance strategy is: control the robot to make a lateral offset within a preset angle range along the original path to form a detour path, while maintaining the original travel speed at a preset ratio;

[0030] When the obstacle avoidance level is the medium risk level, the obstacle avoidance strategy is as follows: control the robot to stop moving, start environmental scanning to update the local cost map, generate a new path that avoids all obstacles based on a preset path planning algorithm, and then continue moving.

[0031] When the obstacle avoidance level is the high-risk level, the obstacle avoidance strategy is as follows: trigger emergency braking to stop the robot from moving, and at the same time issue an alarm to send an intervention request. If no intervention instruction is received in response to the intervention request within a preset time, then obstacle scanning is performed according to a preset cycle until the obstacle is cleared and the ball picking task is restarted.

[0032] Preferably, the obstacle avoidance strategy responds to each obstacle in front during the ball retrieval process, including:

[0033] Multiple detected obstacles are prioritized according to the obstacle avoidance level, with high-risk obstacles having a higher priority than medium-risk obstacles, and medium-risk obstacles having a higher priority than low-risk obstacles.

[0034] When multiple obstacles are at the same obstacle avoidance level, they are sorted from closest to furthest from the robot.

[0035] Based on the sorting results, obstacle avoidance sub-strategies are generated sequentially for each obstacle, and multiple obstacle avoidance sub-strategies are integrated into a continuous obstacle avoidance path.

[0036] Preferably, the local cost map is updated at a preset frequency, and the cost coefficient of the obstacle area in the map is adjusted based on the obstacle category; wherein, the cost coefficient of the humanoid obstacle is greater than that of the high-speed moving obstacle, and the cost coefficient of the high-speed moving obstacle is greater than that of the stationary obstacle and the slow-moving obstacle.

[0037] Preferably, the method further includes:

[0038] After emergency obstacle avoidance is completed, a task recovery process is executed through a behavior tree structure, which includes:

[0039] Path correction node: Calculates the deviation between the position after obstacle avoidance and the original ball-picking path. When the deviation is less than a preset deviation threshold, the robot is controlled to return to the original path.

[0040] Task reordering node: When the deviation is greater than or equal to the deviation threshold or the original path is blocked, the ball picking priority is reassigned based on the position and quantity of the balls to be picked up and the robot's remaining battery power;

[0041] Execution node: Continue executing the ball-picking task based on the revised path or the new priority order.

[0042] Preferably, when the number of consecutive path replanning failures exceeds a preset failure threshold, the behavior tree triggers a recovery node to perform the following operations:

[0043] Clear historical obstacle data from the local cost map;

[0044] Control the robot to rotate in place at a preset speed and re-collect environmental data to update the map;

[0045] If a valid path still cannot be generated, adjust the travel speed to a preset ratio of the original speed, and adjust the obstacle avoidance safety distance to a preset multiple of the original distance before re-planning the path.

[0046] Beneficial effects: The emergency obstacle avoidance method for the robot during ball retrieval in this application achieves accurate obstacle classification by integrating a multi-source obstacle detection device, and dynamically adjusts the obstacle avoidance level by combining distance, speed, direction and category, making the risk assessment more in line with the actual scenario. In particular, the special handling of humanoid obstacles and high-speed moving obstacles improves obstacle avoidance safety. Through hierarchical obstacle avoidance strategy and multi-obstacle priority ranking, targeted and efficient continuous obstacle avoidance is achieved, avoiding over-avoidance or response lag. The dynamic updating of the local cost map and the behavior tree-driven task recovery mechanism enhance the robot's adaptability and robustness in complex environments, ensuring rapid recovery of the ball retrieval task after obstacle avoidance, and significantly improving task execution efficiency and autonomous operation reliability. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 A flowchart illustrating the emergency obstacle avoidance method for a robot during ball retrieval, as provided in this embodiment of the application. Detailed Implementation

[0049] The technical solutions in the embodiments of this application will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0050] In this document, the term "comprising" is intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus 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 apparatus. Without further limitation, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0051] Existing obstacle avoidance methods for robot ball retrieval suffer from problems such as coarse obstacle classification and insufficient correlation between obstacle avoidance levels and categories, resulting in poor targeting of obstacle avoidance strategies. In a simple example, when the robot is performing a ball retrieval task on a tennis court, collisions with athletes or other high-speed moving objects could cause the game to stop or unnecessary injuries. Therefore, this embodiment improves the accuracy of emergency obstacle avoidance during ball retrieval by clearly defining four types of obstacle detection and determining three levels of obstacle avoidance and corresponding obstacle avoidance strategies based on these categories. This enhances the safety and task continuity of the robot in complex environments.

[0052] Specifically, this embodiment discloses as follows: Figure 1 The method shown is a method for handling emergency obstacle avoidance by a robot during ball retrieval, the method including:

[0053] Multiple obstacle detection devices are used to detect obstacles in front of the ball in real time during the ball retrieval process and perform data fusion to obtain obstacle categories, which include stationary obstacles, slow-moving obstacles, high-speed moving obstacles, and humanoid obstacles.

[0054] Based on the real-time detection results of obstacles in front during the ball retrieval process and the obstacle categories, different obstacle avoidance levels are obtained, including low-risk, medium-risk, and high-risk levels. Among them, the basic obstacle avoidance level is obtained based on the real-time detection results of obstacles in front during the ball retrieval process, and then the basic obstacle avoidance level is adjusted based on the obstacle category to obtain the final obstacle avoidance level.

[0055] Based on the obstacle avoidance level, different obstacle avoidance strategies are obtained. The obstacle avoidance strategy responds to each obstacle in front during the ball retrieval process, and there is a corresponding obstacle avoidance sub-strategy for each obstacle. The obstacle avoidance strategy is obtained based on the obstacle avoidance sub-strategy. Based on the obstacle avoidance strategy, emergency obstacle avoidance is performed during the ball retrieval process.

[0056] When retrieving balls, existing robots often rely on a single sensor for obstacle detection, which is susceptible to environmental interference leading to inaccurate classification. This embodiment utilizes existing LiDAR and cameras, and through existing multi-source data fusion processing, fully leverages the advantages of each sensor to significantly improve the accuracy of obstacle category recognition, providing a reliable basis for subsequent obstacle avoidance decisions.

[0057] Specifically, when the obstacle detection device detects obstacles in front of the ball during the ball-picking process in real time, it performs multi-source data fusion on the data collected by the lidar and camera to obtain the obstacle category.

[0058] For data fusion during ball retrieval, existing multi-source data fusion methods lack targeted processing procedures and suffer from poor data matching. This embodiment refines the processing of data from each sensor, using distance information obtained from LiDAR and visual features obtained from cameras, and then performs spatiotemporal matching fusion to further improve the accuracy and reliability of obstacle classification.

[0059] Specifically, obstacle categories are obtained through the following methods:

[0060] Based on the data collected by lidar, information on the distance, outline, and speed of movement of obstacles is obtained;

[0061] Based on data collected by the camera, visual features and shape information of obstacles are obtained;

[0062] The data collected by LiDAR and cameras will be processed and subjected to spatiotemporal matching and data fusion to obtain obstacle categories.

[0063] In addition, the obstacle detection device may also include an infrared sensor, and when the obstacle detection device detects obstacles in front of the ball in real time during the ball picking process, it performs multi-source data fusion on the data collected by the lidar, camera and infrared sensor to obtain the obstacle category.

[0064] Secondly, when classifying obstacles in the lake area, data collected by infrared sensors is used to help identify obstacles with temperature characteristics; and the processing results of data collected by lidar, cameras and infrared sensors are spatiotemporally matched and fused to obtain the obstacle category.

[0065] Furthermore, as those skilled in the art can conceive, in order to improve obstacle recognition accuracy and thus improve the final obstacle avoidance effect, any suitable error correction technology in the prior art can be used to correct sensor errors. It is feasible to use the error correction technology based on IMU to correct errors as described in application number CN202211027562.6. Those skilled in the art can adapt the error correction technology to the emergency obstacle avoidance processing method of the robot in the ball picking process of this embodiment based on the description of the technical solution of this application. This text will not elaborate further.

[0066] Existing obstacle avoidance levels are generally determined directly based on the movement characteristics of obstacles, often ignoring the influence of obstacle category, resulting in risk assessments that deviate from reality. This embodiment, based on a basic level, raises the level for humanoid obstacles and adjusts the low-risk to medium-risk level for high-speed moving obstacles, making the obstacle avoidance levels more closely match the actual threat posed by obstacles and improving the accuracy of risk response.

[0067] Specifically, based on the type of obstacle, different obstacle avoidance levels are obtained, including:

[0068] A basic obstacle avoidance level is obtained based on the distance, speed, and direction of movement of the obstacle;

[0069] When the obstacle is classified as a humanoid obstacle, the risk level is increased by one level from the basic obstacle avoidance level;

[0070] When the obstacle category is a high-speed moving obstacle and the basic obstacle avoidance level is low risk, it is adjusted to medium risk.

[0071] When the obstacle category is a stationary obstacle or a slow-moving obstacle, the basic obstacle avoidance level remains unchanged.

[0072] When analyzing basic obstacle avoidance levels, there are issues such as relying on a single determining factor or setting unreasonable weights, leading to biased level determinations. This embodiment calculates a comprehensive score using a weighted model built with distance, speed, and direction factors and existing machine learning techniques, and combines this with a preset range to determine the basic level, making the basic level assessment more scientific and laying a reliable foundation for subsequent level adjustments.

[0073] Specifically, based on the distance, speed, and direction of movement of the obstacle, a basic obstacle avoidance level is obtained, including:

[0074] Obtain a preset distance range and determine the distance factor based on the actual distance range between the obstacle and the robot;

[0075] Obtain a preset speed range, and determine the speed factor based on the actual speed range of the obstacle;

[0076] Obtain a preset range of directional angles, and determine the direction factor based on the actual angle between the obstacle's movement direction and the robot's travel direction within that range;

[0077] A comprehensive score is obtained by weighting the distance factor, speed factor, and direction factor using a preset weighting model. The basic obstacle avoidance level is determined as low-risk, medium-risk, or high-risk based on the preset score range in which the comprehensive score falls.

[0078] It should be noted that the distance range, speed range, and directional angle range in this embodiment can be empirical values ​​set based on the experience of those skilled in the art, or thresholds obtained through analysis of historical ball-picking data. In this embodiment, the weight model is constructed based on existing deep learning technology, aiming to achieve weighted calculation of the distance factor, speed factor, and direction factor.

[0079] In practical applications, robots often use a uniform obstacle avoidance pattern, which is difficult to adapt to different risk scenarios. This embodiment designs lateral offset, stop-and-sweep replanning, and emergency braking strategies for low, medium, and high risk levels respectively, to avoid over-avoidance or response lag, and improve the adaptability and efficiency of obstacle avoidance.

[0080] Specifically, based on the obstacle avoidance level, different obstacle avoidance strategies are derived, including:

[0081] When the obstacle avoidance level is low risk, the obstacle avoidance strategy is: control the robot to make a lateral offset within a preset angle range along the original path to form an bypass path, while maintaining the original travel speed at a preset ratio.

[0082] When the obstacle avoidance level is medium risk, the obstacle avoidance strategy is as follows: control the robot to stop moving, start environmental scanning to update the local cost map, generate a new path that avoids all obstacles based on the preset path planning algorithm, and then continue moving.

[0083] When the obstacle avoidance level is high-risk, the obstacle avoidance strategy is as follows: trigger emergency braking to stop the robot from moving, and at the same time send an alarm to send an intervention request. If no intervention command is received within a preset time, the obstacle will be scanned according to a preset cycle until the obstacle is cleared and the ball picking task will be restarted.

[0084] Furthermore, obstacle avoidance with multiple obstacles is prone to strategy conflicts due to the lack of priority processing. This embodiment sorts obstacle avoidance sub-strategies by obstacle avoidance level and distance, and integrates them into a continuous path to ensure orderly and efficient obstacle avoidance in multi-obstacle scenarios and avoid path chaos.

[0085] Specifically, the obstacle avoidance strategy responds to each obstacle in front of the player during the ball retrieval process, including:

[0086] Multiple detected obstacles are prioritized according to their obstacle avoidance level, with high-risk obstacles having a higher priority than medium-risk obstacles, and medium-risk obstacles having a higher priority than low-risk obstacles.

[0087] When multiple obstacles are at the same obstacle avoidance level, they are sorted from closest to furthest from the robot.

[0088] Based on the sorting results, obstacle avoidance sub-strategies are generated sequentially for each obstacle, and multiple obstacle avoidance sub-strategies are integrated into a continuous obstacle avoidance path.

[0089] Existing local cost maps suffer from delayed updates or fixed cost coefficients, resulting in low path planning accuracy. This embodiment updates the map at a preset frequency and dynamically adjusts the cost coefficients based on obstacle categories, giving higher weight to human figures and high-speed obstacles, thereby improving the safety and rationality of path planning.

[0090] Specifically, the local cost map is updated at a preset frequency, and the cost coefficient of obstacle areas in the map is adjusted based on the obstacle category; among them, the cost coefficient of humanoid obstacles is greater than that of high-speed moving obstacles, and the cost coefficient of high-speed moving obstacles is greater than that of stationary obstacles and slow-moving obstacles.

[0091] After the robot completes obstacle avoidance, the suddenness of the obstacle avoidance can lead to a lack of smooth transition in task resumption, which can easily cause the ball retrieval task to be interrupted or even cause the robot to lose control on the field. This embodiment, based on existing behavior tree technology, uses path correction, task reordering, and execution node adjustments to ensure that the original task can be quickly resumed or priorities adjusted after obstacle avoidance, thus guaranteeing the continuity and efficiency of the ball retrieval task.

[0092] Specifically, the method also includes:

[0093] After emergency obstacle avoidance is completed, the task recovery process is executed through a behavior tree structure, which includes:

[0094] Path correction node: Calculates the deviation between the position after obstacle avoidance and the original ball picking path. When the deviation is less than the preset deviation threshold, the robot is controlled to return to the original path.

[0095] Task reordering node: When the deviation is greater than or equal to the deviation threshold or the original path is blocked, the ball picking priority is reassigned based on the position and quantity of the balls to be picked up and the robot's remaining battery power.

[0096] Execution node: Continue executing the ball-picking task based on the revised path or the new priority order.

[0097] Furthermore, the lack of an effective recovery mechanism after path replanning failures can cause the robot to become stuck, i.e., the robot becomes out of control on the field. This embodiment clears historical data, rotates and updates the map, and adjusts speed and safe distance when there are consecutive failures, significantly improving the robustness of the robot's navigation recovery ability in complex environments.

[0098] Specifically, when the number of consecutive path replanning failures exceeds a preset failure threshold, the behavior tree triggers a recovery node to perform the following operations:

[0099] Clear historical obstacle data from the local cost map;

[0100] Control the robot to rotate in place at a preset speed and re-collect environmental data to update the map;

[0101] If a valid path still cannot be generated, adjust the travel speed to a preset ratio of the original speed and adjust the obstacle avoidance safety distance to a preset multiple of the original distance before trying to plan again.

[0102] In summary, the emergency obstacle avoidance method for the robot during ball retrieval in this embodiment achieves accurate obstacle classification by integrating a multi-source obstacle detection device. It dynamically adjusts the obstacle avoidance level based on distance, speed, direction, and category, making the risk assessment more realistic. The special handling of humanoid obstacles and high-speed moving obstacles enhances obstacle avoidance safety. Through a hierarchical obstacle avoidance strategy and multi-obstacle priority ranking, it achieves targeted and efficient continuous obstacle avoidance, preventing over-avoidance or response lag. Dynamic updates of the local cost map and a behavior tree-driven task recovery mechanism enhance the robot's adaptability and robustness in complex environments, ensuring rapid recovery of the ball retrieval task after obstacle avoidance, significantly improving task execution efficiency and autonomous operation reliability.

[0103] In the embodiments provided in this application, it should be understood that the embodiments described herein can be implemented in hardware, software, firmware, middleware, code, or any suitable combination thereof. For hardware implementation, the processor may be implemented in one or more of the following: application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, other electronic units designed to implement the functions described herein, or combinations thereof. For software implementation, some or all of the processes of the embodiments may be performed by a computer program instructing the associated hardware. During implementation, the program may be stored in a computer-readable storage medium or transmitted as one or more instructions or code on a computer-readable storage medium. Computer-readable storage media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of a computer program from one place to another. Storage media may be any available medium accessible to a computer. Computer-readable storage media may include, but are not limited to, RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code having the form of instructions or data structures and accessible to a computer.

[0104] Finally, it should be noted that the above description is only a preferred embodiment of this application and is not intended to limit this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for handling emergency obstacle avoidance by a robot during ball retrieval, characterized in that, The method includes: Multiple obstacle detection devices are used to detect obstacles in front of the ball in real time during the ball retrieval process and perform data fusion to obtain obstacle categories, which include stationary obstacles, slow-moving obstacles, high-speed moving obstacles, and humanoid obstacles. Based on the results of real-time detection of obstacles in front during the ball retrieval process and the obstacle category, different obstacle avoidance levels are obtained, including low-risk, medium-risk, and high-risk levels; wherein, a basic obstacle avoidance level is obtained based on the results of real-time detection of obstacles in front during the ball retrieval process, and then the basic obstacle avoidance level is adjusted based on the obstacle category to obtain the final obstacle avoidance level. Based on the obstacle avoidance level, different obstacle avoidance strategies are obtained. Each obstacle avoidance strategy responds to each obstacle in front of the ball during the ball retrieval process, and there is a corresponding obstacle avoidance sub-strategy for each obstacle. The obstacle avoidance strategy is obtained based on the obstacle avoidance sub-strategy. Based on the obstacle avoidance strategy, emergency obstacle avoidance is performed during the ball retrieval process. The different obstacle avoidance strategies derived based on the obstacle avoidance level include: When the obstacle avoidance level is the low-risk level, the obstacle avoidance strategy is: control the robot to make a lateral offset within a preset angle range along the original path to form a detour path, while maintaining the original travel speed at a preset ratio; When the obstacle avoidance level is the medium risk level, the obstacle avoidance strategy is as follows: control the robot to stop moving, start environmental scanning to update the local cost map, generate a new path that avoids all obstacles based on a preset path planning algorithm, and then continue moving. When the obstacle avoidance level is the high-risk level, the obstacle avoidance strategy is as follows: trigger emergency braking to stop the robot from moving, and at the same time issue an alarm to send an intervention request. If no intervention instruction responding to the intervention request is received within a preset time, then obstacle scanning is performed according to a preset cycle until the obstacle is cleared and the ball picking task is restarted. The method also includes: After emergency obstacle avoidance is completed, a task recovery process is executed through a behavior tree structure, which includes: Path correction node: Calculates the deviation between the position after obstacle avoidance and the original ball-picking path. When the deviation is less than a preset deviation threshold, the robot is controlled to return to the original path. Task reordering node: When the deviation is greater than or equal to the deviation threshold or the original path is blocked, the ball picking priority is reassigned based on the position and quantity of the balls to be picked up and the robot's remaining battery power; Execution node: Continue executing the ball-picking task based on the revised path or the new priority order; When the number of consecutive path replanning failures exceeds a preset failure threshold, the behavior tree triggers a recovery node to perform the following operations: Clear historical obstacle data from the local cost map; Control the robot to rotate in place at a preset speed and re-collect environmental data to update the map; If a valid path still cannot be generated, adjust the travel speed to a preset ratio of the original speed, and adjust the obstacle avoidance safety distance to a preset multiple of the original distance before re-planning the path.

2. The method for handling emergency obstacle avoidance by a robot during ball retrieval according to claim 1, characterized in that, The obstacle detection device includes a lidar and a camera; When the obstacle detection device detects obstacles in front of the ball during the ball-picking process in real time, it performs multi-source data fusion on the data collected by the lidar and camera to obtain the obstacle category.

3. The method for handling emergency obstacle avoidance by a robot during ball retrieval according to claim 2, characterized in that, The obstacle category is obtained by performing multi-source data fusion on the data collected by lidar and camera, including: Based on the data collected by lidar, information on the distance, outline, and speed of movement of obstacles is obtained; Based on data collected by the camera, visual features and shape information of obstacles are obtained; The data collected by LiDAR and cameras will be processed and subjected to spatiotemporal matching and data fusion to obtain obstacle categories.

4. The method for handling emergency obstacle avoidance by a robot during ball retrieval according to claim 1, characterized in that, Based on the obstacle category, different obstacle avoidance levels are obtained, including: A basic obstacle avoidance level is obtained based on the distance, speed, and direction of movement of the obstacle; When the obstacle category is the humanoid obstacle, the risk level is increased by one level based on the basic obstacle avoidance level; When the obstacle category is a high-speed moving obstacle and the basic obstacle avoidance level is the low-risk level, it is adjusted to the medium-risk level; When the obstacle category is either a stationary obstacle or a slow-moving obstacle, the basic obstacle avoidance level remains unchanged.

5. The method for handling emergency obstacle avoidance by a robot during ball retrieval according to claim 4, characterized in that, The basic obstacle avoidance level is obtained based on the distance, speed, and direction of movement of the obstacle, including: Obtain a preset distance range, and determine a distance factor based on the distance range in which the actual distance between the obstacle and the robot falls; Obtain a preset speed range, and determine a speed factor based on the speed range in which the actual speed of the obstacle falls; Obtain a preset directional angle range, and determine the direction factor based on the actual angle between the obstacle's movement direction and the robot's travel direction within the directional angle range; A comprehensive score is obtained by weighting the distance factor, the speed factor, and the direction factor using a preset weighting model. Based on the preset score range in which the comprehensive score falls, the basic obstacle avoidance level is determined to be the low-risk level, the medium-risk level, or the high-risk level.

6. The method for handling emergency obstacle avoidance by a robot during ball retrieval according to claim 1, characterized in that, The obstacle avoidance strategy responds to each obstacle in front of the player during the ball retrieval process, including: Multiple detected obstacles are prioritized according to the obstacle avoidance level, with high-risk obstacles having a higher priority than medium-risk obstacles, and medium-risk obstacles having a higher priority than low-risk obstacles. When multiple obstacles are at the same obstacle avoidance level, they are sorted from closest to furthest from the robot. Based on the sorting results, obstacle avoidance sub-strategies are generated sequentially for each obstacle, and multiple obstacle avoidance sub-strategies are integrated into a continuous obstacle avoidance path.

7. The method for handling emergency obstacle avoidance by a robot during ball retrieval according to claim 1, characterized in that, The local cost map is updated at a preset frequency, and the cost coefficient of the obstacle area in the map is adjusted based on the obstacle category; wherein, the cost coefficient of the humanoid obstacle is greater than that of the high-speed moving obstacle, and the cost coefficient of the high-speed moving obstacle is greater than that of the stationary obstacle and the slow-moving obstacle.

Images