Obstacle avoidance moving method and device, equipment, storage medium and program product

Abstract

The invention discloses an obstacle avoidance moving method, device and equipment, a storage medium and a program product, the method is applied to a robot, and the method comprises the following steps: carrying out feature analysis on collected sensor data to obtain obstacle contour information, static obstacle information and dynamic obstacle information in a current environment; generating a plurality of candidate movement strategies according to the obstacle information and the current position of the robot; performing value evaluation on the plurality of candidate movement strategies to obtain an evaluation result; screening out a target movement strategy from the plurality of candidate movement strategies based on the evaluation result, and moving in the current environment according to the target movement strategy; due to the fact that the dynamic obstacle information and the static obstacle information in the current environment are obtained, different types of obstacles in the dynamic environment can be accurately recognized, the target movement strategy is screened out from the candidate movement strategy evaluation result, the optimal movement decision is made, and the accuracy of the movement decision is improved. The intelligent and accurate obstacle avoidance movement of the robot in a dynamic and complex environment is realized.

Description

Technical Field

[0001] The present invention relates to the field of robotics technology, and in particular to an obstacle avoidance movement method, device, equipment, storage medium and program product. Background Art

[0002] Currently, there is an increasing demand for deploying mobile service robots in indoor environments, but there are many problems from simulation to actual application. Many navigation methods are difficult to cope with real-world challenges, such as being unable to effectively avoid dynamic obstacles, unable to avoid C-shaped obstacles, etc. They have poor adaptability in different environments, are difficult to generalize to the real world, and rely on complex sensor fusion and heavy post-processing. Traditional navigation methods also have difficulty in obstacle avoidance and adjustment. Summary of the invention

[0003] The main purpose of the present invention is to provide an obstacle avoidance movement method, device, equipment, storage medium and program product, aiming to solve the technical problem that the prior art cannot accurately avoid obstacles and move in a dynamic environment, resulting in poor environmental adaptability of the robot and inability to move flexibly.

[0004] To achieve the above object, the present invention provides an obstacle avoidance movement method, the method comprising the following steps:

[0005] Performing feature analysis on the collected sensor data to obtain obstacle information in the current environment, wherein the obstacle information includes obstacle contour information, static obstacle information, and dynamic obstacle information;

[0006] generating a plurality of candidate movement strategies according to the obstacle information and the current position of the robot;

[0007] Performing value evaluation on the multiple candidate mobility strategies to obtain evaluation results;

[0008] A target movement strategy is selected from the plurality of candidate movement strategies based on the evaluation result, and movement is performed in the current environment according to the target movement strategy.

[0009] Optionally, the performing feature analysis on the collected sensor data to obtain obstacle information in the current environment includes:

[0010] Preprocess the collected sensor data to obtain the data to be analyzed in each area of ​​the current environment;

[0011] Performing feature analysis on the data to be analyzed to obtain obstacle feature information of each area, wherein the obstacle feature information includes time distribution feature information and space distribution feature information;

[0012] Get the attention weight of each region;

[0013] Feature fusion is performed based on the attention weight and the obstacle feature information to obtain obstacle information in the current environment.

[0014] Optionally, performing feature analysis on the data to be analyzed to obtain feature information of obstacles in each area includes:

[0015] Determining obstacle coordinate information and obstacle speed information in each area based on the data to be analyzed;

[0016] Determining static obstacles and dynamic obstacles in the current environment according to the obstacle speed information;

[0017] Acquire first relative motion information between the robot and the static obstacle, and second relative motion information between the robot and the dynamic obstacle based on the obstacle coordinate information and the obstacle speed information;

[0018] Predicting a trajectory of the dynamic obstacle according to the first relative motion information and the second relative motion information to obtain an obstacle motion trajectory of the dynamic obstacle;

[0019] Feature analysis is performed based on the obstacle motion trajectory and the obstacle coordinate information to obtain obstacle feature information of each area.

[0020] Optionally, generating a plurality of candidate movement strategies according to the obstacle information and the current position of the robot includes:

[0021] Get future path point information;

[0022] Inputting the obstacle information, the current position of the robot and the future path point information into a pre-built path planning model for path planning to obtain a candidate movement strategy;

[0023] Among them, the path planning model includes an actor network, which is configured to determine the current state based on the obstacle information, the current position of the robot and the future path point information, determine the action instructions corresponding to the current state based on a preset mapping relationship library, and generate multiple candidate movement strategies based on the action instructions.

[0024] Optionally, performing value evaluation on the multiple candidate mobility strategies to obtain evaluation results includes:

[0025] Inputting the plurality of candidate movement strategies and the obstacle information into a pre-built path planning model for value evaluation to obtain an evaluation result, wherein the evaluation result includes a path value parameter of each candidate movement strategy;

[0026] Wherein, the path planning model includes an evaluation network, and the evaluation network is configured to perform action value evaluation on the action instructions in each candidate mobile strategy based on the obstacle information, and determine the path value parameters of each candidate mobile strategy based on the action value evaluation.

[0027] Optionally, after selecting a target mobility strategy from the plurality of candidate mobility strategies based on the evaluation result and moving in the current environment according to the target mobility strategy, the method further includes:

[0028] Monitor the current moving path and the distance between it and obstacles in the current environment;

[0029] Determine whether the current movement path deviates from the navigation path of the target movement strategy based on the monitoring result of the current movement path, and obtain a first judgment result;

[0030] Based on the monitoring result of the interval distance, determine whether the interval distance between the obstacle in the current environment meets the preset safety distance condition, and obtain a second judgment result;

[0031] Determine reward and punishment parameters during the movement according to the preset reward and punishment strategy, the first judgment result and the second judgment result;

[0032] The target movement strategy is updated based on the reward and punishment parameters, the target movement strategy is moved in the current environment based on the updated target movement strategy, and the step of monitoring the current movement path and the interval distance between the target movement path and the obstacles in the current environment is returned to be executed.

[0033] In addition, to achieve the above-mentioned purpose, the present invention further proposes an obstacle avoidance mobile device, the obstacle avoidance mobile device comprising:

[0034] A feature analysis module is used to perform feature analysis on the collected sensor data to obtain obstacle information in the current environment, wherein the obstacle information includes obstacle contour information, static obstacle information, and dynamic obstacle information;

[0035] A movement planning module, used to generate a plurality of candidate movement strategies according to the obstacle information and the current position of the robot;

[0036] A strategy evaluation module, used to evaluate the value of the multiple candidate mobile strategies and obtain an evaluation result;

[0037] The mobile control module is used to select a target mobile strategy from the multiple candidate mobile strategies based on the evaluation result, and move in the current environment according to the target mobile strategy.

[0038] In addition, to achieve the above-mentioned purpose, the present application also proposes an obstacle avoidance mobile device, which includes: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the computer program is configured to implement the steps of the obstacle avoidance movement method described above.

[0039] In addition, to achieve the above-mentioned purpose, the present application also proposes a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the obstacle avoidance movement method described above are implemented.

[0040] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps of the obstacle avoidance movement method described above are implemented.

[0041] The present invention obtains obstacle information in the current environment by performing feature analysis on the collected sensor data, and the obstacle information includes obstacle contour information, static obstacle information and dynamic obstacle information; generates multiple candidate movement strategies according to the obstacle information and the current position of the robot; performs value evaluation on the multiple candidate movement strategies to obtain evaluation results; screens out a target movement strategy from the multiple candidate movement strategies based on the evaluation results, and moves in the current environment according to the target movement strategy; because the present invention obtains obstacle contour information, dynamic obstacle information and static obstacle information in the current environment by performing feature analysis on the collected sensor data, the static obstacles, dynamic obstacles and the shapes of each obstacle in the environment are accurately identified, multiple candidate movement strategies are generated based on the obstacle information and the current position of the robot, and the target movement strategy is screened out from the candidate movement strategy evaluation results, so as to make the optimal movement decision, effectively avoiding the problem of easy collision with different types of obstacles such as dynamic obstacles and irregularly shaped obstacles during the movement of the robot, and realizing the robot to accurately avoid obstacles and move flexibly in a dynamic and complex environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0043] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0044] Figure 1It is a structural diagram of an obstacle avoidance mobile device in a hardware operating environment involved in an embodiment of the present invention;

[0045] Figure 2 It is a schematic diagram of the flow chart of the first embodiment of the obstacle avoidance movement method of the present invention;

[0046] Figure 3 It is a flow chart of the second embodiment of the obstacle avoidance movement method of the present invention;

[0047] Figure 4 It is a flowchart of a third embodiment of the obstacle avoidance movement method of the present invention;

[0048] Figure 5 It is a structural block diagram of the first embodiment of the obstacle avoidance mobile device of the present invention.

[0049] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0050] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0051] Reference Figure 1 , Figure 1 A schematic diagram of the structure of an obstacle avoidance mobile device in a hardware operating environment involved in an embodiment of the present invention.

[0052] like Figure 1 As shown, the obstacle avoidance mobile device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a wireless fidelity (Wireless-Fidelity, WI-FI) interface). The memory 1005 may be a high-speed random access memory (Random Access Memory, RAM), or a stable non-volatile memory (Non-Volatile Memory, NVM), such as a disk storage. The memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0053] Those skilled in the art will understand that Figure 1The structure shown in the figure does not constitute a limitation on the obstacle avoidance mobile device, and may include more or less components than shown in the figure, or combine certain components, or arrange the components differently.

[0054] like Figure 1 As shown, the memory 1005 as a computer-readable storage medium may include an operating system, a network communication module, a user interface module, and an obstacle avoidance movement program.

[0055] exist Figure 1 In the obstacle avoiding mobile device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the obstacle avoiding mobile device of the present invention can be set in the obstacle avoiding mobile device, and the obstacle avoiding mobile device calls the obstacle avoidance movement program stored in the memory 1005 through the processor 1001, and executes the obstacle avoidance movement method provided by the embodiment of the present invention.

[0056] The embodiment of the present invention provides an obstacle avoidance movement method, referring to Figure 2 , Figure 2 FIG. 4 is a flow chart of the first embodiment of the obstacle avoidance movement method of the present invention.

[0057] In this embodiment, the obstacle avoidance movement method includes the following steps:

[0058] Step S10: Perform feature analysis on the collected sensor data to obtain obstacle information in the current environment.

[0059] It should be understood that the execution subject of this embodiment can be a movable robot with data processing, network communication and program running functions, such as a sweeping robot, a cleaning robot, etc., or an electronic device capable of realizing the above functions, etc. The following takes a robot as an example to illustrate this embodiment and the following embodiments.

[0060] It should be noted that the above obstacle information includes obstacle profile information, static obstacle information and dynamic obstacle information. The above static obstacle information may be the distance information, coordinate / position information, speed information, etc. of obstacles in a static or relatively static state in the current environment. The above dynamic obstacle information may be the distance information, coordinate / position information, speed information, etc. of obstacles in a moving state in the current environment.

[0061] The above obstacle contour information may be contour shape information of dynamic obstacles / static obstacles in the current environment. The rough shape or contour of the obstacle may be inferred from the lidar data through point cloud segmentation and clustering technology.

[0062] The above distance information can be calculated by the robot through a sensor (e.g., lidar) by emitting laser and measuring the time it takes for it to reflect back. The distance information can be used to determine the relative distance between the obstacle and the robot.

[0063] The above coordinate / position information can be the coordinates and position of the obstacle. The robot can convert the lidar data into position information (i.e. coordinates) in the robot coordinate system. By combining the distance and angle information, the specific position of the obstacle in the robot's surrounding environment can be determined. When performing path planning or obstacle avoidance, it is usually necessary to convert these distance and angle information into specific coordinate points for further processing and analysis.

[0064] The above speed information can be the movement speed information of dynamic obstacles. The robot can analyze the movement speed and direction of obstacles by continuously observing the lidar data. Through TAGD technology, the movement of obstacles is inferred by comparing the changes in data points in continuous lidar scans.

[0065] In some embodiments, the robot can collect sensor data through a lidar sensor to provide the robot with accurate obstacle information within the plane. The sensor is independent of lighting conditions and can operate in day and night environments.

[0066] It can be understood that the robot can determine the motion state of obstacles in the current environment by processing and feature analyzing continuous LiDAR scanning data, thereby distinguishing between dynamic obstacles and static obstacles, and determining the movement trend of dynamic obstacles relative to static obstacles, thereby offsetting the impact of the robot's own motion, helping to distinguish between static and dynamic obstacles, and providing a more robust solution for learning-based navigation.

[0067] It should be noted that an important goal of the feature extraction process is to distinguish between static and dynamic obstacles, so as to avoid misjudgment due to the robot's own motion. The feature extraction results can also include information related to the robot's navigation path, such as the location of the nearest obstacle, the movement trend of the obstacle, possible avoidance paths, etc.

[0068] Step S20: Generate multiple candidate movement strategies according to the obstacle information and the current position of the robot.

[0069] It should be noted that the candidate movement strategy may include the navigation movement path of the robot and the robot action instructions, and the candidate movement strategy may be planned based on the obstacle contour information, dynamic obstacle information and static obstacle information in the current environment.

[0070] It should be noted that the navigation movement path in the candidate movement strategy may be the navigation path of the robot moving from the current position to the expected target position (i.e., the expected position of the movement instruction received by the robot). The above-mentioned action instruction may be the action and posture taken by the robot during the movement along the navigation movement path, for example, the action instruction may include the linear and angular velocity of the robot during the movement process.

[0071] It can be understood that the robot can determine the edge contour of each obstacle based on the obstacle contour information in the current environment, thereby determining the shape of the obstacle, identifying irregularly shaped obstacles, distinguishing between static obstacles and dynamic obstacles in the current environment based on static obstacle information and dynamic obstacle information, and determining the position and distance of static obstacles and dynamic obstacles, and planning multiple candidate movement strategies based on the identified irregularly shaped obstacles, dynamic obstacles and static obstacles, as well as the position and distance of each obstacle.

[0072] Furthermore, in order to improve the accuracy of the mobile planning, the above step S20 may include:

[0073] Step S21: Obtaining future path point information;

[0074] Step S22: inputting the obstacle information, the current position of the robot and the future path point information into a pre-built path planning model to perform path planning and obtain a candidate movement strategy;

[0075] Among them, the path planning model includes an actor network, which is configured to determine the current state based on the obstacle information, the current position of the robot and the future path point information, determine the action instructions corresponding to the current state based on a preset mapping relationship library, and generate multiple candidate movement strategies based on the action instructions.

[0076] It should be noted that the future path point information may be the path and point information planned for the robot in the movement command received by the robot, that is, the expected path point information planned for the robot in advance, which helps the robot determine the direction and target during navigation. For example, the future path point information may include the end point position information and the waypoint position information of the robot's movement.

[0077] It should be noted that an action instruction can be a decision made by the robot based on the current state, that is, the behavior that the agent chooses to perform. In robot navigation tasks, actions usually refer to the robot's control instructions, such as linear speed, angular velocity, or specific turning and forward instructions. These actions directly affect the robot's motion trajectory and state changes.

[0078] The above state can be the environment observed by the robot at a certain point in time. The state contains all the information needed for the agent to make decisions. In the case of a robot navigation task, the state includes the readings of the lidar sensor (such as the distance, position, speed, etc. of obstacles), the current position of the robot, global path information (such as future path points), and any other data related to the task.

[0079] It is understandable that the preset mapping relationship library can be a mapping relationship library constructed based on the state-action pairs obtained by pre-associating action instructions with states. The state-action pair refers to the action taken by the agent in a specific state. It is a basic unit used to describe the interaction process between the agent and the environment. In reinforcement learning, the agent continuously tries different actions, observes the state changes caused by the actions, and adjusts its own strategy according to these changes to optimize long-term benefits.

[0080] In some embodiments, the path planning model may be a neural network model pre-constructed by reinforcement learning, for example, it may be constructed based on the DDPG (Deep Deterministic Policy Gradient) reinforcement algorithm, which is an off-policy reinforcement learning algorithm. The state-action pairs are stored in the experience replay buffer for batch policy updates, with a continuous action space, allowing the robot to be smoothly controlled. It is composed of an actor and a critic. Through the mutual cooperation of the actor and the critic, the robot can learn the optimal navigation strategy to achieve the goal of navigating along a given path freely in unknown dynamic obstacles. Among them, the above-mentioned actor may be an actor network in the path planning model, and the actor network can generate candidate strategies based on the state-action pairs.

[0081] It should be noted that the state-action pair can be the basis for the robot to make decisions. In a given state, the robot can choose the most appropriate action based on the learned strategy. This process is the core of the reinforcement learning algorithm.

[0082] It should be understood that in the reinforcement learning process, the robot learns the relationship between state and action by continuous trial and observation. Each attempt will produce a new state-action pair, which is stored in the experience replay buffer. By repeatedly replaying these pairs, the agent can continuously optimize its strategy and thus improve navigation performance.

[0083] It should be noted that the robot selects actions through the learned strategy and observes the impact of these actions on the state. Based on the observed results (i.e. rewards), the agent can evaluate the quality of the action and adjust the strategy accordingly. This iterative optimization process enables the agent to gradually adapt to complex and changing environments.

[0084] It can be understood that in robot navigation tasks, state-action pairs enable the robot to select the most appropriate action (such as turning, accelerating or decelerating) according to the real-time observed environmental state (such as the position and dynamics of obstacles), thereby achieving intelligent and safe navigation.

[0085] Step S30: performing value evaluation on the multiple candidate mobility strategies to obtain evaluation results.

[0086] It should be noted that the evaluation result may include a path value evaluation result of the navigation movement path in the candidate movement strategy and an action value evaluation result of the action instruction in the candidate movement strategy.

[0087] It can be understood that, in this embodiment, the path value of the navigation movement path and the action value of the action instruction can be determined by performing value evaluation on the navigation movement path and the action instruction in each candidate movement strategy.

[0088] In some embodiments, the robot can evaluate the value of candidate mobile strategies through a pre-built neural network model, taking the navigation movement path and action instructions in the candidate mobile strategies as input, and outputting the path value of the navigation movement path and the action value of the action instructions through calculation of the neural network.

[0089] Furthermore, in order to accurately evaluate the value of the candidate mobile strategies and thus select high-value strategies, the above step S30 may include:

[0090] Step S31: inputting the plurality of candidate movement strategies and the obstacle information into a pre-built path planning model for value evaluation to obtain an evaluation result, wherein the evaluation result includes a path value parameter of each candidate movement strategy;

[0091] Wherein, the path planning model includes an evaluation network, and the evaluation network is configured to perform action value evaluation on the action instructions in each candidate mobile strategy based on the obstacle information, and determine the path value parameters of each candidate mobile strategy based on the action value evaluation.

[0092] It should be noted that the evaluation network can be a critic network in the path planning model, and the critic network can be used to evaluate the value of the action taken by the actor. It also receives information in the state space and the action output by the actor as input, and outputs an estimate of the value of the action through the calculation of the neural network. The actor network in the path planning model can be used to output the linear and angular velocity of the robot. It receives information in the state space as input, including 2D lidar sensor data, TAGD derived from the lidar, and upcoming waypoints. Through the calculation of the neural network, the actor network determines the action that the robot should take.

[0093] In some embodiments, the path planning model can be built based on DDPG. First, through the dual-stream spatiotemporal attention network, by focusing on the sensor data in space and time, the collision-related features are highlighted from the feature data, and selective scene perception is achieved, the overall navigation performance is improved, and more targeted information is provided for the DDPG architecture. Then, the deep deterministic policy gradient (DDPG) architecture based on reinforcement learning is used for navigation, and the neural network is used for modeling, weighing the importance of the upcoming waypoints, and outputting information such as linear and angular velocities. The state space and linear space provide a basis for the training and decision-making of the reinforcement learning algorithm. The state space includes the lidar sensor data and the upcoming waypoints for global guidance. The action space consists of linear and angular velocities, and the robot moves according to these speed instructions.

[0094] It can be understood that the robot obtains sensor data, performs feature analysis on the sensor data, obtains spatial feature data and temporal feature data respectively, and then inputs the spatial feature data and temporal feature data into the path planning model for path planning to obtain the target movement strategy.

[0095] It should be noted that the time feature data can be time stream data. The input of the time stream data can come from the continuous scanning of the LiDAR sensor. The feature information is the key data extracted from these time stream data that can help the robot understand and respond to dynamic changes in the environment. The input of time stream data usually comes from the LiDAR sensor. The LiDAR continuously scans the environment at a certain time interval and generates LiDAR point cloud data at multiple time points. These point cloud data will form a time series over time, which is called time stream data. The robot captures environmental information in real time through the LiDAR sensor and inputs these continuous LiDAR scanning data into the navigation system. The key data extracted from the time stream data can help the robot understand the dynamic changes in the environment.

[0096] Step S40: Filtering out a target mobility strategy from the plurality of candidate mobility strategies based on the evaluation result, and moving in the current environment according to the target mobility strategy.

[0097] It should be noted that the target movement strategy can be the movement strategy with the highest path value and / or action value among the candidate movement strategies. The robot can effectively avoid collisions with obstacles in the current environment by adopting the target movement strategy, and greatly improve movement efficiency and shorten the time required to move to the desired position.

[0098] In some embodiments, the robot is a cleaning robot, which moves with a target movement strategy to improve cleaning efficiency, quickly reach the target cleaning position for cleaning, and avoid collisions with obstacles during the cleaning and moving processes, thereby shortening the time it takes for the cleaning robot to reach the next cleaning position and improving cleaning efficiency.

[0099] It can be understood that this embodiment determines the path value and action value of each candidate mobile strategy by performing value evaluation on each candidate mobile strategy, and performs value evaluation on the navigation mobile path and action instructions in each candidate mobile strategy, and screens out the target mobile strategy with the highest value from multiple candidate mobile strategies based on the path value and action value.

[0100] In some embodiments, during the movement of the robot, a reward and punishment strategy can be set based on the movement behavior during the movement to dynamically optimize the target movement strategy, thereby continuously collecting experience data (state, action, reward, etc.) and storing these data in the experience playback buffer. Then, these data are used to perform batch strategy updates to optimize the decision-making strategy of the intelligent agent so that it can better adapt to the environment and achieve the navigation goal.

[0101] This embodiment obtains obstacle information in the current environment by performing feature analysis on the collected sensor data, and the obstacle information includes obstacle contour information, static obstacle information and dynamic obstacle information; generates multiple candidate movement strategies according to the obstacle information and the current position of the robot; performs value evaluation on the multiple candidate movement strategies to obtain evaluation results; screens out a target movement strategy from the multiple candidate movement strategies based on the evaluation results, and moves in the current environment according to the target movement strategy; because this embodiment obtains obstacle contour information, dynamic obstacle information and static obstacle information in the current environment by performing feature analysis on the collected sensor data, the static obstacles, dynamic obstacles and the shapes of each obstacle in the environment are accurately identified, multiple candidate movement strategies are generated based on the obstacle information and the current position of the robot, and the target movement strategy is screened out from the candidate movement strategy evaluation results, so as to make the optimal movement decision, effectively avoiding the problem of easy collision with different types of obstacles such as dynamic obstacles and irregularly shaped obstacles during the movement of the robot, and realizing the robot to accurately avoid obstacles and move flexibly in a dynamic and complex environment.

[0102] refer to Figure 3 , Figure 3 FIG. 4 is a flow chart of a second embodiment of the obstacle avoidance movement method of the present invention.

[0103] Based on the above first embodiment, in this embodiment, step S10 further includes:

[0104] Step S11: pre-process the collected sensor data to obtain the data to be analyzed in each area of ​​the current environment.

[0105] It should be noted that the data to be analyzed may be the sensor data collected in each area after being pre-processed.

[0106] It should be noted that the sensor data collected by the robot through sensors may be original sensor data such as original lidar data and millimeter wave radar data.

[0107] It is understandable that the robot can improve data quality by performing pre-processing operations such as denoising and filtering on the raw lidar data.

[0108] In some embodiments, the robot can pre-divide the current environment into space to improve the efficiency of data processing and path analysis, such as dividing the scanning range of the lidar into several sectors or grids so that each area can be processed separately, which helps to simplify calculations and reduce the amount of calculations.

[0109] Step S12: performing feature analysis on the data to be analyzed to obtain feature information of obstacles in each area.

[0110] It should be noted that the above obstacle feature information includes time distribution feature information and space distribution feature information. This embodiment focuses on sensor data in space and time, thereby highlighting collision-related features from feature data in both time and space dimensions, achieving selective scene perception and improving overall navigation performance.

[0111] It should be noted that the above-mentioned spatial distribution feature information may be information describing the distribution of obstacles in the robot's surrounding environment, such as density, degree of aggregation, etc. These features are helpful in understanding the complexity of the environment and potential navigation challenges.

[0112] The above-mentioned temporal dynamic feature information can capture the movement trend and change pattern of obstacles by analyzing the time series of LiDAR data, which is crucial for predicting the future position of dynamic obstacles.

[0113] It can be understood that this embodiment can reveal the movement trend of mobile obstacles relative to static obstacles through feature extraction and processing of continuous laser radar scanning data. It can offset the impact of the robot's own movement, help distinguish static and dynamic obstacles, and provide a more robust solution for learning-based navigation.

[0114] Furthermore, in order to accurately perform feature analysis on the data, the above step S12 may include:

[0115] Step S121: determining the obstacle coordinate information and obstacle speed information in each area based on the data to be analyzed;

[0116] Step S122: determining static obstacles and dynamic obstacles in the current environment according to the obstacle speed information;

[0117] Step S123: acquiring first relative motion information between the robot and the static obstacle, and second relative motion information between the robot and the dynamic obstacle based on the obstacle coordinate information and the obstacle speed information;

[0118] Step S124: predicting the trajectory of the dynamic obstacle according to the first relative motion information and the second relative motion information to obtain the obstacle motion trajectory of the dynamic obstacle;

[0119] Step S125: performing feature analysis based on the obstacle motion trajectory and the obstacle coordinate information to obtain obstacle feature information of each area.

[0120] It should be noted that the obstacle coordinate information can be the coordinate position of the obstacle. The robot can convert the distance and angle measured by the laser radar into coordinate points in the robot coordinate system. These coordinate points are used to describe the specific spatial position of the obstacle, and then combined with the current position of the robot to determine the current distance information between the obstacle and the robot.

[0121] The above obstacle speed information can be estimated by comparing the changes in data points in continuous LiDAR scans, for example, by using optical flow, feature point tracking or other spatiotemporal analysis methods.

[0122] It is understandable that the robot can distinguish between static obstacles and dynamic obstacles in the current environment by determining the speed of each obstacle in the current environment.

[0123] It should be understood that the robot can obtain the first relative motion information between the robot itself and the static obstacle and the second relative motion information of the dynamic obstacle, and then predict the motion trajectory of the dynamic obstacle based on the first relative motion information and the second relative motion information.

[0124] Step S13: Obtain the attention weight of each region.

[0125] It should be noted that the attention weight can be the weight of the importance and attention degree of each area. In this embodiment, the attention mechanism can be used to perform weighted processing on the key areas in the lidar data to highlight the features related to the collision. This is achieved through a neural network, where the network learns how to assign different attention weights to different lidar sectors or time steps.

[0126] Step S14: performing feature fusion based on the attention weight and the obstacle feature information to obtain obstacle information in the current environment.

[0127] It is understood that the present embodiment can fuse features extracted from different LiDAR scans, different time steps and different spatial regions to form a comprehensive description of the environment. This is usually achieved by feature vector concatenation, weighted summation or other fusion techniques.

[0128] This embodiment obtains data to be analyzed in each area of ​​the current environment by preprocessing the collected sensor data, performs feature analysis on the data to be analyzed, obtains obstacle feature information of each area, the obstacle feature information includes time distribution feature information and space distribution feature information, obtains attention weights of each area, performs feature fusion based on the attention weights and the obstacle feature information, and obtains obstacle information in the current environment, thereby realizing feature analysis based on two dimensions of time and space, assigning attention weights to key areas in the current environment through the time-space attention mechanism, and fusion of features extracted from different time steps and different spatial areas, thereby accurately and comprehensively describing the current environment, greatly improving the accuracy of obstacle information, and enhancing the robot's adaptability to dynamic environments.

[0129] refer to Figure 4 , Figure 4 FIG. 4 is a flow chart of a third embodiment of the obstacle avoidance movement method of the present invention.

[0130] Based on the above embodiments, in this embodiment, after step S40, the method further includes:

[0131] Step S50: monitor the current moving path and the distance between the current moving path and obstacles in the current environment.

[0132] It should be noted that the current moving path may be the moving path of the robot from the initial position to the current real-time position. The above-mentioned interval distance may be the real-time interval distance between the robot and the obstacle during the movement.

[0133] Step S60: Based on the monitoring result of the current moving path, determine whether the current moving path deviates from the navigation path of the target moving strategy to obtain a first determination result.

[0134] It can be understood that this embodiment can determine the similarity between the current moving path and the navigation path in the target moving strategy based on the monitoring results of the current moving path. If the similarity is lower than a preset threshold, it is determined that the robot has deviated from the navigation path of the target moving strategy. If the similarity is not lower than the preset threshold, it is determined that the robot has not yet deviated from the navigation path of the target moving strategy.

[0135] Step S70: Based on the monitoring result of the interval distance, determine whether the interval distance between the obstacle in the current environment meets a preset safety distance condition to obtain a second judgment result.

[0136] It can be understood that this embodiment determines whether the robot collides with the obstacle or whether there is a risk of collision with the obstacle by monitoring the distance between the robot and the obstacle. If the distance between the robot and the obstacle is less than the preset distance threshold, it is determined that the preset safety distance condition is not met, and the robot is at risk of collision or has collided.

[0137] Step S80: determining reward and punishment parameters during the moving process according to a preset reward and punishment strategy, the first judgment result and the second judgment result.

[0138] It should be noted that this embodiment can adopt reinforcement learning methods such as reward function and policy optimization for learning. The reward function is used to guide the navigation behavior of the robot, including collision penalty, natural guidance penalty along the global path, and warning penalty for approaching obstacles. By reasonably designing the reward function, the robot is encouraged to achieve collision-free navigation, move along the path and avoid approaching obstacles. Policy optimization is used in the process of interacting with the environment, and the agent collects experience data (state, action, reward, etc.) and stores this data in the experience playback buffer. Then, this data is used to perform batch policy updates to optimize the decision-making strategy of the agent so that it can better adapt to the environment and achieve navigation goals.

[0139] Step S90: updating the target movement strategy based on the reward and punishment parameters, moving in the current environment based on the updated target movement strategy, and returning to execute the step of monitoring the current movement path and the interval distance between obstacles in the current environment.

[0140] It can be understood that this embodiment monitors the movement process of the robot and optimizes the robot's movement strategy through rewards and penalties. By continuously optimizing and updating the movement strategy, it ensures that the robot can better adapt to the dynamic environment, improve the robot's movement efficiency, and avoid collision and yaw risks.

[0141] This embodiment monitors the current moving path and the interval distance between each obstacle in the current environment, determines whether the current movement deviates from the navigation path of the target moving strategy based on the monitoring result of the current moving path, obtains a first judgment result, determines whether the interval distance between the obstacles in the current environment meets the preset safety distance condition based on the monitoring result of the interval distance, obtains a second judgment result, determines the reward and punishment parameters in the moving process according to the preset reward and punishment strategy, the first judgment result and the second judgment result, updates the target moving strategy based on the reward and punishment parameters, moves in the current environment based on the updated target moving strategy, and returns to execute the step of monitoring the current moving path and the interval distance between each obstacle in the current environment, thereby realizing dynamic updating of the moving strategy and continuously optimizing the robot's moving strategy so that it can better adapt to the environment and achieve navigation goals.

[0142] In addition, an embodiment of the present invention further proposes a computer-readable storage medium, on which an obstacle avoidance movement program is stored. When the obstacle avoidance movement program is executed by a processor, the steps of the obstacle avoidance movement method described above are implemented.

[0143] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: 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 embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination of the above.

[0144] The computer-readable storage medium may be included in the obstacle avoidance mobile device; or may exist independently without being assembled into the obstacle avoidance mobile device.

[0145] In addition, an embodiment of the present invention further provides a computer program product, including an obstacle avoidance movement program, which implements the steps of the obstacle avoidance movement method described above when executed by a processor.

[0146] The specific implementation methods of the computer program product of the present invention are basically the same as the above-mentioned embodiments of the obstacle avoidance movement method, and will not be repeated here.

[0147] Reference Figure 5 , Figure 5 It is a structural block diagram of the first embodiment of the obstacle avoidance mobile device of the present invention.

[0148] like Figure 5 As shown, the obstacle avoidance mobile device proposed in the embodiment of the present invention includes:

[0149] The feature analysis module 10 is used to perform feature analysis on the collected sensor data to obtain obstacle information in the current environment, wherein the obstacle information includes obstacle contour information, static obstacle information and dynamic obstacle information;

[0150] A movement planning module 20, configured to generate a plurality of candidate movement strategies according to the obstacle information and the current position of the robot;

[0151] A strategy evaluation module 30 is used to evaluate the value of the multiple candidate mobility strategies and obtain an evaluation result;

[0152] The movement control module 40 is used to select a target movement strategy from the multiple candidate movement strategies based on the evaluation result, and move in the current environment according to the target movement strategy.

[0153] Furthermore, the feature analysis module 10 is also used to pre-process the collected sensor data to obtain the data to be analyzed in each area of ​​the current environment; perform feature analysis on the data to be analyzed to obtain obstacle feature information of each area, wherein the obstacle feature information includes time distribution feature information and space distribution feature information; obtain the attention weight of each area; perform feature fusion based on the attention weight and the obstacle feature information to obtain the obstacle information in the current environment.

[0154] Furthermore, the feature analysis module 10 is also used to determine obstacle coordinate information and obstacle speed information in each area based on the data to be analyzed; determine static obstacles and dynamic obstacles in the current environment according to the obstacle speed information; obtain first relative motion information between the robot and the static obstacle, and second relative motion information between the robot and the dynamic obstacle based on the obstacle coordinate information and the obstacle speed information; predict the trajectory of the dynamic obstacle according to the first relative motion information and the second relative motion information to obtain the obstacle motion trajectory of the dynamic obstacle; perform feature analysis based on the obstacle motion trajectory and the obstacle coordinate information to obtain obstacle feature information of each area.

[0155] Furthermore, the mobile planning module 20 is also used to obtain future path point information; input the obstacle information, the current position of the robot and the future path point information into a pre-built path planning model for path planning to obtain candidate mobile strategies; wherein the path planning model includes an actor network, and the actor network is configured to determine a current state based on the obstacle information, the current position of the robot and the future path point information, determine an action instruction corresponding to the current state based on a preset mapping relationship library, and generate multiple candidate mobile strategies based on the action instruction.

[0156] Furthermore, the strategy evaluation module 30 is also used to input the multiple candidate mobile strategies and the obstacle information into a pre-built path planning model for value evaluation to obtain an evaluation result, which includes the path value parameters of each candidate mobile strategy; wherein the path planning model includes an evaluation network, and the evaluation network is configured to perform action value evaluation on the action instructions in each candidate mobile strategy based on the obstacle information, and determine the path value parameters of each candidate mobile strategy based on the action value evaluation.

[0157] Furthermore, the mobile control module 40 is also used to monitor the current moving path and the interval distance between each obstacle in the current environment; based on the monitoring result of the current moving path, determine whether the current movement deviates from the navigation path of the target moving strategy, and obtain a first judgment result; based on the monitoring result of the interval distance, determine whether the interval distance between the obstacles in the current environment meets the preset safety distance condition, and obtain a second judgment result; determine the reward and punishment parameters during the movement process according to the preset reward and punishment strategy, the first judgment result and the second judgment result; update the target moving strategy based on the reward and punishment parameters, move in the current environment based on the updated target moving strategy, and return to execute the step of monitoring the current moving path and the interval distance between each obstacle in the current environment.

[0158] This embodiment obtains obstacle information in the current environment by performing feature analysis on the collected sensor data, and the obstacle information includes obstacle contour information, static obstacle information and dynamic obstacle information; generates multiple candidate movement strategies according to the obstacle information and the current position of the robot; performs value evaluation on the multiple candidate movement strategies to obtain evaluation results; screens out a target movement strategy from the multiple candidate movement strategies based on the evaluation results, and moves in the current environment according to the target movement strategy; because this embodiment obtains obstacle contour information, dynamic obstacle information and static obstacle information in the current environment by performing feature analysis on the collected sensor data, the static obstacles, dynamic obstacles and the shapes of each obstacle in the environment are accurately identified, multiple candidate movement strategies are generated based on the obstacle information and the current position of the robot, and the target movement strategy is screened out from the candidate movement strategy evaluation results, so as to make the optimal movement decision, effectively avoiding the problem of easy collision with different types of obstacles such as dynamic obstacles and irregularly shaped obstacles during the movement of the robot, and realizing the robot to accurately avoid obstacles and move flexibly in a dynamic and complex environment.

[0159] The obstacle avoidance mobile device provided by the present application adopts the obstacle avoidance mobile method in the above embodiment, which can solve the technical problem of obstacle avoidance mobile. Compared with the prior art, the beneficial effects of the obstacle avoidance mobile device provided by the present application are the same as the beneficial effects of the obstacle avoidance mobile method provided by the above embodiment, and other technical features in the obstacle avoidance mobile device are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0160] It should be understood that the above is only an example and does not constitute any limitation on the technical solution of the present invention. In specific applications, technicians in this field can make settings as needed, and the present invention does not limit this.

[0161] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of the present invention. In practical applications, technicians in this field can select part or all of them according to actual needs to achieve the purpose of the present embodiment, and no limitation is made here.

[0162] In addition, for technical details that are not described in detail in this embodiment, reference can be made to the obstacle avoidance movement method provided in any embodiment of the present invention, and will not be repeated here.

[0163] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0164] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0165] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an 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, which is stored in a storage medium (such as a read-only memory / random access memory, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0166] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. An obstacle avoidance movement method, characterized in that: The obstacle avoidance movement method is applied to a robot, and the obstacle avoidance movement method comprises: Performing feature analysis on the collected sensor data to obtain obstacle information in the current environment, wherein the obstacle information includes obstacle contour information, static obstacle information, and dynamic obstacle information; generating a plurality of candidate movement strategies according to the obstacle information and the current position of the robot; Performing value evaluation on the multiple candidate mobility strategies to obtain evaluation results; A target movement strategy is selected from the plurality of candidate movement strategies based on the evaluation result, and movement is performed in the current environment according to the target movement strategy.

2. The obstacle avoidance movement method according to claim 1, characterized in that: The feature analysis of the collected sensor data to obtain obstacle information in the current environment includes: Preprocess the collected sensor data to obtain the data to be analyzed in each area of ​​the current environment; Performing feature analysis on the data to be analyzed to obtain obstacle feature information of each area, wherein the obstacle feature information includes time distribution feature information and space distribution feature information; Get the attention weight of each region; Feature fusion is performed based on the attention weight and the obstacle feature information to obtain obstacle information in the current environment.

3. The obstacle avoidance movement method according to claim 2, characterized in that: The performing feature analysis on the data to be analyzed to obtain obstacle feature information of each area includes: Determining obstacle coordinate information and obstacle speed information in each area based on the data to be analyzed; Determining static obstacles and dynamic obstacles in the current environment according to the obstacle speed information; Acquire first relative motion information between the robot and the static obstacle, and second relative motion information between the robot and the dynamic obstacle based on the obstacle coordinate information and the obstacle speed information; Predicting a trajectory of the dynamic obstacle according to the first relative motion information and the second relative motion information to obtain an obstacle motion trajectory of the dynamic obstacle; Feature analysis is performed based on the obstacle motion trajectory and the obstacle coordinate information to obtain obstacle feature information of each area.

4. The obstacle avoidance movement method according to any one of claims 1 to 3, characterized in that: The generating a plurality of candidate movement strategies according to the obstacle information and the current position of the robot comprises: Get future path point information; Inputting the obstacle information, the current position of the robot and the future path point information into a pre-built path planning model for path planning to obtain a candidate movement strategy; Among them, the path planning model includes an actor network, which is configured to determine the current state based on the obstacle information, the current position of the robot and the future path point information, determine the action instructions corresponding to the current state based on a preset mapping relationship library, and generate multiple candidate movement strategies based on the action instructions.

5. The obstacle avoidance movement method according to any one of claims 1 to 3, characterized in that: The performing value evaluation on the multiple candidate mobility strategies to obtain evaluation results includes: Inputting the plurality of candidate movement strategies and the obstacle information into a pre-built path planning model for value evaluation to obtain an evaluation result, wherein the evaluation result includes a path value parameter of each candidate movement strategy; Wherein, the path planning model includes an evaluation network, and the evaluation network is configured to perform action value evaluation on the action instructions in each candidate mobile strategy based on the obstacle information, and determine the path value parameters of each candidate mobile strategy based on the action value evaluation.

6. The obstacle avoidance movement method according to any one of claims 1 to 3, characterized in that: After selecting a target mobile strategy from the plurality of candidate mobile strategies based on the evaluation result and moving in the current environment according to the target mobile strategy, the method further includes: Monitor the current moving path and the distance between it and obstacles in the current environment; Determine whether the current movement path deviates from the navigation path of the target movement strategy based on the monitoring result of the current movement path, and obtain a first judgment result; Based on the monitoring result of the interval distance, determine whether the interval distance between the obstacle in the current environment meets the preset safety distance condition, and obtain a second judgment result; Determine reward and punishment parameters during the movement according to the preset reward and punishment strategy, the first judgment result and the second judgment result; The target movement strategy is updated based on the reward and punishment parameters, the target movement strategy is moved in the current environment based on the updated target movement strategy, and the step of monitoring the current movement path and the interval distance between the target movement path and the obstacles in the current environment is returned to be executed.

7. An obstacle avoidance mobile device, characterized in that: The obstacle avoidance mobile device comprises: A feature analysis module is used to perform feature analysis on the collected sensor data to obtain obstacle information in the current environment, wherein the obstacle information includes obstacle profile information, static obstacle information, and dynamic obstacle information; A movement planning module, used to generate a plurality of candidate movement strategies according to the obstacle information and the current position of the robot; A strategy evaluation module, used to evaluate the value of the multiple candidate mobile strategies and obtain an evaluation result; The mobile control module is used to select a target mobile strategy from the multiple candidate mobile strategies based on the evaluation result, and move in the current environment according to the target mobile strategy.

8. An obstacle avoidance mobile device, characterized in that: The obstacle avoidance mobile device comprises: a memory, a processor, and an obstacle avoidance movement program stored in the memory and executable on the processor, wherein the obstacle avoidance movement program is configured to implement the obstacle avoidance movement method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores an obstacle avoidance movement program, and when the obstacle avoidance movement program is executed by the processor, the obstacle avoidance movement method according to any one of claims 1 to 6 is implemented.

10. A computer program product, characterized in that The computer program product comprises an obstacle avoidance movement program, and when the obstacle avoidance movement program is executed by a processor, the steps of the obstacle avoidance movement method according to any one of claims 1 to 6 are implemented.

Images