Robot obstacle avoidance method and device suitable for dynamic obstacle

Abstract

The invention discloses a robot obstacle avoidance method and device suitable for a dynamic obstacle, and the method comprises the steps: obtaining environment perception information in real time in a process of advancing according to a preset advancing route; under the condition of determining that the dynamic obstacle exists in the preset range based on the environment perception information, obtaining the current space position information of the dynamic obstacle and the pre-stored system delay duration; the pre-stored system delay duration is obtained by quantizing and storing the processing performance of the robot in the charging process of the robot; performing spatial position compensation on the current spatial position information according to a pre-stored system delay duration to obtain actual spatial position information of the dynamic obstacle; predicting an obstacle moving track of the dynamic obstacle according to the actual spatial position information; and controlling the robot to avoid the obstacle based on the obstacle moving trajectory. Therefore, by adopting the embodiment of the invention, the motion prediction deviation is overcome, so that the collision risk is reduced, and the safety of the robot is improved.

Description

Technical Field

[0001] The present application relates to the technical field of robots, and particularly to a robot obstacle avoidance method and device applicable to dynamic obstacles. Background Art

[0002] When a robot is applied to a complex dynamic environment (such as a transportation hub or a public event venue), the robot needs to move safely and efficiently among high-speed moving objects (such as pedestrians, vehicles or goods), and at the same time needs to avoid collisions with these dynamic obstacles.

[0003] In the related art, the robot first obtains environmental perception information through sensors, then uses the environmental perception information to detect the spatial position information of dynamic obstacles, and finally predicts the movement trajectory of the dynamic obstacles based on the spatial position information, and performs obstacle avoidance based on the movement trajectory.

[0004] However, the predicted movement trajectory in this method is only the real result at the historical moment when the robot obtains the environmental perception information. As the service time of the robot becomes longer and the processing performance decreases, the detection action and the prediction action need to take a certain amount of time to complete. At this time, for a high-speed moving dynamic obstacle, this result is not the result corresponding to the spatial position information of the robot at the current moment. Therefore, there is a movement prediction deviation, which increases the collision risk and reduces the safety of the robot. Summary of the Invention

[0005] Embodiments of the present application provide a robot obstacle avoidance method and device applicable to dynamic obstacles. To provide a basic understanding of some aspects of the disclosed embodiments, a simple summary is given below. This summary part is not a general review, nor is it intended to identify key / important constituent elements or delineate the protection scope of these embodiments. Its sole purpose is to present some concepts in a simple form as a prelude to the detailed description below.

[0006] In a first aspect, embodiments of the present application provide a robot obstacle avoidance method applicable to dynamic obstacles, which is applied to a robot. The method includes: During the process of traveling according to a preset traveling route, obtaining environmental perception information in real time; When it is determined based on the environmental perception information that there is a dynamic obstacle within a preset range, obtaining the current spatial position information of the dynamic obstacle and the pre-stored system delay duration; the pre-stored system delay duration is obtained by quantitatively storing the processing performance of the robot during the charging process of the robot; According to the pre-stored system delay duration, performing spatial position compensation on the current spatial position information to obtain the actual spatial position information of the dynamic obstacle; According to the actual spatial position information, predicting the obstacle movement trajectory of the dynamic obstacle; Based on the running trajectory of the obstacle, control the robot to avoid obstacles.

[0007] In a second aspect, an embodiment of the present application provides a robot obstacle avoidance device applicable to dynamic obstacles, and the device includes: An environmental perception information acquisition module, configured to acquire environmental perception information in real time during the process of traveling along a preset traveling route; A system delay duration acquisition module, configured to acquire the current spatial position information of the dynamic obstacle and the pre-stored system delay duration when it is determined that there is a dynamic obstacle within a preset range based on the environmental perception information; the pre-stored system delay duration is obtained by quantifying the processing performance of the robot during the charging process of the robot; A spatial position compensation module, configured to perform spatial position compensation on the current spatial position information according to the pre-stored system delay duration to obtain the actual spatial position information of the dynamic obstacle; An obstacle running trajectory prediction module, configured to predict the obstacle running trajectory of the dynamic obstacle according to the actual spatial position information; An obstacle avoidance control module, configured to control the robot to avoid obstacles based on the obstacle running trajectory.

[0008] The technical solutions provided by the embodiments of the present application may include the following beneficial effects: In the embodiment of the present application, on the one hand, by quantifying the processing performance and pre-storing the system delay duration during the charging process of the robot, the pre-stored system delay duration is calculated and stored in advance, avoiding the time-consuming caused by real-time calculation, enabling quick query of the delay duration during the traveling process to compensate the current spatial position information of the dynamic obstacle, thereby obtaining the actual spatial position information closer to the real situation, and enabling more accurate motion trajectory prediction based on this, effectively overcoming the motion prediction deviation. On the other hand, by loading a pre-trained reinforcement learning model, the spatial position calculation network in the model can consider the influence of system delay and perform more accurate calculation on the actual spatial position of the dynamic obstacle, and the reinforcement learning network further optimizes the prediction result through feature reinforcement, thereby outputting the predicted displacement delayed by the dynamic obstacle, enabling the robot to more accurately predict the motion trajectory of the dynamic obstacle, thereby reducing the collision risk.

[0009] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The accompanying drawings here are incorporated into the specification and constitute a part of this specification, showing embodiments consistent with the present application and used together with the specification to explain the principles of the present application.

[0011] Figure 1 It is a schematic flowchart of a method for a robot to avoid obstacles applicable to dynamic obstacles provided by an embodiment of the present application; Figure 2 It is a schematic diagram of an obstacle provided by an embodiment of the present application; Figure 3 It is a schematic diagram of the trajectory of a dynamic obstacle provided by an embodiment of the present application; Figure 4 It is a schematic flowchart of a model training method for a reinforcement learning model provided by an embodiment of the present application; Figure 5 It is a schematic diagram of the architecture of a spatial position calculation network provided by an embodiment of the present application; Figure 6 It is a schematic diagram of the structure of a robot obstacle avoidance device applicable to dynamic obstacles provided by an embodiment of the present application; Figure 7 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Detailed implementation manners

[0012] The following description and drawings fully illustrate the specific implementation manners of the present application, enabling those skilled in the art to practice them.

[0013] It should be clear that the described embodiments are only a part of the embodiments of the present application, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts belong to the scope of protection of the present application.

[0014] When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the present application. On the contrary, they are only examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0015] In the description of the present application, it should be understood that the terms "first", "second", etc. are only used for descriptive purposes and cannot be construed as indicating or implying relative importance. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to specific situations. In addition, in the description of the present application, unless otherwise stated, "a plurality of" means two or more. "And / or" describes the association relationship of associated objects and indicates that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. The character " / " generally represents an "or" relationship between the associated objects before and after.

[0016] Currently, the robot first obtains environmental perception information through sensors, then uses the environmental perception information to detect the spatial position information of dynamic obstacles, and finally predicts the motion trajectory of the dynamic obstacles based on the spatial position information, and performs obstacle avoidance based on this motion trajectory.

[0017] The inventors have realized that the predicted motion trajectory in this method is only the true result at the historical moment when the robot obtains the environmental perception information. As the service time of the robot becomes longer and the processing performance decreases, the detection action and the prediction action need to take a certain amount of time to complete. At this time, for a dynamically moving obstacle, this result is not the result corresponding to the robot's spatial position information at the current moment. Therefore, there is a motion prediction deviation, which increases the collision risk and reduces the safety of the robot.

[0018] To solve the above problems, the present application provides a robot obstacle avoidance method and device applicable to dynamic obstacles to solve the problems existing in the above related technical problems. In an embodiment of the present application, on the one hand, by quantifying the processing performance and pre-storing the system delay duration during the charging process of the robot, the pre-stored system delay duration is calculated and stored in advance, avoiding the time-consuming of real-time calculation, so that the delay duration can be quickly queried during the traveling process to compensate the current spatial position information of the dynamic obstacle, thereby obtaining the actual spatial position information closer to the real situation, and more accurate motion trajectory prediction can be performed based on this, effectively overcoming the motion prediction deviation. On the other hand, by loading a pre-trained reinforcement learning model, the spatial position calculation network in the model can consider the influence of system delay and calculate the actual spatial position of the dynamic obstacle more accurately. The reinforcement learning network further optimizes the prediction result through feature reinforcement, thereby outputting the predicted displacement delayed by the dynamic obstacle, enabling the robot to more accurately predict the motion trajectory of the dynamic obstacle, thereby reducing the collision risk. The following uses exemplary embodiments for detailed description.

[0019] The following will combine with the attached Figure 1 - attached Figure 5 to introduce in detail the robot obstacle avoidance method applicable to dynamic obstacles provided by the embodiments of the present application. This method can be implemented depending on a computer program and can run on a robot obstacle avoidance device applicable to the von Neumann architecture. This computer program can be integrated in an application or run as an independent tool class application.

[0020] Please refer to Figure 1 for a schematic flowchart of a robot obstacle avoidance method applicable to dynamic obstacles provided by an embodiment of the present application, which is applied to a robot. As Figure 1 shown, the method of the embodiment of the present application includes the following steps: S101. While the robot is moving along the preset travel route, it obtains environmental perception information in real time; Among them, the preset travel route is the path pre-planned by the robot before performing the task. The environmental perception information is the information about the surrounding environment collected by the robot through sensors.

[0021] In some embodiments of the present application, before the task starts, the robot generates a path from the starting point to the ending point through a path planning algorithm. This path takes into account the static obstacles in the environment to ensure that the robot can efficiently complete the task. The robot is equipped with a lidar, a camera, and an ultrasonic sensor.

[0022] Specifically, the lidar can scan the surrounding environment at a high frequency and detect the distances of objects within a certain range around the robot. The camera is used to identify the shapes and colors of objects. When the robot approaches an obstacle, the ultrasonic sensor will detect the obstacle at close range and feed the information back to the control system.

[0023] In some embodiments of the present application, after the control system of the robot receives the data fed back by the lidar, the camera, and the ultrasonic sensor, it filters the data collected by the sensors to remove noise and outliers. The data collected by different sensors are fused to obtain more comprehensive environmental perception information. It can be determined whether there are dynamic obstacles within the preset range based on the comprehensive environmental perception information. The final judgment result is, for example Figure 2 as shown, there are dynamic obstacles within the preset range. The identification of this dynamic obstacle can be carried out using existing vision algorithms, which will not be elaborated here.

[0024] S102. When the robot determines that there are dynamic obstacles within the preset range based on the environmental perception information, it obtains the current spatial position information of the dynamic obstacles and the pre-stored system delay duration; the pre-stored system delay duration is obtained by quantitatively storing the processing performance of the robot during the charging process; Among them, a dynamic obstacle is an obstacle whose position and speed will change during the movement of the robot, such as pedestrians, vehicles, etc. The current spatial position information is the precise position information of the dynamic obstacle at the current moment, including its coordinates in three-dimensional space. The pre-stored system delay duration is the system processing delay time pre-calculated and stored by the robot during the charging process, which reflects the time required for the robot to generate the running trajectory of the obstacle from obtaining the environmental perception information.

[0025] In some embodiments of the present application, when it is determined based on the environmental perception information that there are no dynamic obstacles within the preset range, the robot moves along the preset travel route.

[0026] In some other embodiments of the present application, when it is determined based on environmental perception information that there is a dynamic obstacle within a preset range, the robot measures the current spatial position information of the dynamic obstacle through a lidar, and obtains the pre-stored system delay duration corresponding to the historical moment closest to the current moment.

[0027] In some embodiments of the present application, the specific process of generating the pre-stored system delay duration includes: during the charging process of the robot, obtaining the current load and task queue length of the processor as the processor operation parameters; simultaneously extracting a detection function for detecting the spatial position information of the dynamic obstacle and a prediction function for predicting the dynamic obstacle; in the simulation platform, creating the virtual spatial position information of the dynamic virtual obstacle; performing a simulation run of the obstacle movement trajectory according to the virtual spatial position information, the detection function, and the prediction function; counting the function execution time during the simulation run; determining the processor operation time of the processor according to the current load and task queue length; adding the function execution time and the processor operation time, and storing the added result to obtain the pre-stored system delay duration.

[0028] Among them, the current load and task queue length are used to evaluate the current performance and processing ability of the processor. The detection function is an algorithm or function for detecting the spatial position information of the dynamic obstacle. For example, a detection function based on lidar data is used to determine the coordinates of the obstacle. The prediction function is an algorithm or function for predicting the future movement trajectory of the dynamic obstacle. The simulation platform is a virtual test environment for simulating the obstacle detection and prediction process in the real scenario. The function execution time is the actual time consumed by the detection function and the prediction function during the simulation run. The processor operation time is used to evaluate the processing ability of the processor in the current state.

[0029] For example, during the charging process of the robot, run the performance test program to obtain the current load and task queue length of the processor.

[0030] Extract the detection function and the prediction function from the software system of the robot. Create a virtual environment in the simulation platform (such as Gazebo) and set a dynamic virtual obstacle.

[0031] Assume that the virtual obstacle moves in a straight line at a speed of 1 meter per second, and its initial position is (x = 0, y = 0, z = 0). Run the detection function and the prediction function in the simulation platform to simulate the movement trajectory of the obstacle. Obtain the current spatial position information of the virtual obstacle from the simulation platform, calculate the current spatial position using the detection function, predict the future position using the prediction function. During the simulation run, use a high-precision timer to record the execution time of the detection function and the prediction function. Add the function execution time and the processor operation time to obtain the system delay duration. Among them, the system delay duration is shown in Table 1, for example.

[0032] Table 1

[0033] S103. The robot performs spatial position compensation on the current spatial position information according to the pre-stored system delay duration to obtain the actual spatial position information of the dynamic obstacle; Among them, the spatial position compensation is to adjust the current spatial position information of the dynamic obstacle according to the pre-stored system delay duration to compensate for the position deviation caused by the system processing delay.

[0034] In some embodiments of the present application, the specific process of performing spatial position compensation on the current spatial position information according to the pre-stored system delay duration to obtain the actual spatial position information of the dynamic obstacle includes: calculating the instantaneous linear velocity and instantaneous acceleration of the dynamic obstacle according to the current spatial position information and the historical spatial position information of the dynamic obstacle at the previous moment; determining the predicted displacement delayed by the dynamic obstacle according to the pre-stored system delay duration, instantaneous linear velocity and instantaneous acceleration; decomposing the predicted displacement into the horizontal and vertical coordinates of the coordinate system to obtain the horizontal coordinate compensation component and the vertical coordinate compensation component; using the horizontal coordinate compensation component and the vertical coordinate compensation component to perform spatial position compensation on the current spatial position information to obtain the actual spatial position information of the dynamic obstacle.

[0035] Among them, the instantaneous linear velocity is the linear velocity of the dynamic obstacle at a certain moment, indicating the distance it moves along a straight line per unit time. The instantaneous acceleration is the acceleration of the dynamic obstacle at a certain moment, indicating the rate of change of its velocity. Displacement decomposition is to decompose the predicted displacement into components in the horizontal coordinate (x-axis) and vertical coordinate (y-axis) directions. The horizontal coordinate compensation component is the component of the predicted displacement in the horizontal coordinate (x-axis) direction. The vertical coordinate compensation component is the component of the predicted displacement in the vertical coordinate (y-axis) direction. The spatial position compensation is to adjust the current spatial position information of the dynamic obstacle according to the horizontal coordinate compensation component and the vertical coordinate compensation component.

[0036] In the embodiments of the present application, the specific process of determining the predicted displacement delayed by the dynamic obstacle according to the pre-stored system delay duration, instantaneous linear velocity and instantaneous acceleration includes: calling a preset obstacle position compensation model; using the pre-stored system delay duration, instantaneous linear velocity and instantaneous acceleration as the model parameters of the preset obstacle position compensation model to obtain a model expression with parameter supplementation; executing the model expression with parameter supplementation to obtain the predicted displacement delayed by the dynamic obstacle; among them, the model expression of the preset obstacle position compensation model is:

[0037] Among them, is the predicted displacement, is the instantaneous linear velocity of the dynamic obstacle, is the instantaneous acceleration of the dynamic obstacle, is the system processing delay time; Among them, the calculation formula for the actual spatial position information of the dynamic obstacle is: ; Among them, is the actual spatial position information, is the abscissa compensation component, is the ordinate compensation component, is the abscissa of the current spatial position information, is the ordinate of the current spatial position information.

[0038] For example, during the operation of the robot, a dynamic obstacle is detected, and its current spatial position information is (x = 3 meters, y = 2 meters), and the historical spatial position information at the previous moment is (x = 2.8 meters, y = 1.9 meters). The system delay duration is 0.2 seconds. The following is the specific compensation process: Calculate the instantaneous linear velocity and instantaneous acceleration: The instantaneous linear velocity ; Assume that the instantaneous acceleration at the current moment , calculate the predicted displacement: Predicted displacement, .

[0039] Displacement decomposition: The abscissa compensation component (assuming the angle between the movement direction and the x-axis is ), the ordinate compensation component ; Spatial position compensation: The compensated actual spatial position information: , .

[0040] In the embodiment of the present application, according to the pre-stored system delay duration, instantaneous linear velocity, and instantaneous acceleration, the specific process of determining the predicted displacement delayed by the dynamic obstacle includes: loading the pre-trained reinforcement learning model; inputting the pre-stored system delay duration, instantaneous linear velocity, and instantaneous acceleration into the pre-trained reinforcement learning model; the reinforcement learning model includes a spatial position calculation network for calculating the actual spatial position of the dynamic obstacle and a reinforcement learning network for feature reinforcement; outputting the predicted displacement delayed by the dynamic obstacle.

[0041] S104. The robot predicts the obstacle movement trajectory of the dynamic obstacle according to the actual spatial position information; In some embodiments of the present application, the specific process of predicting the obstacle running trajectory of a dynamic obstacle based on the actual spatial position information includes: constructing a second-order kinematic model; wherein, the model expression of the second-order kinematic model is:

[0042] Wherein, and are the predicted position coordinates of the dynamic obstacle at a future moment, is the current moment, is the time interval for future trajectory prediction, and are the actual spatial position information of the dynamic obstacle after system delay compensation, and are the current velocity components of the dynamic obstacle in the axis and axis directions, is the pre-stored system delay duration, and are the acceleration components of the dynamic obstacle in the axis and axis directions; Calculate the motion trend of the dynamic obstacle based on the actual spatial position information and the current velocity of the dynamic obstacle, in combination with the second-order kinematic model; Plot the position sequences at multiple time points under the motion trend to obtain discrete position points; Connect the discrete position points into line segments to obtain the obstacle running trajectory of the dynamic obstacle. The obstacle running trajectory of the dynamic obstacle is shown in, for example, Figure 3 as shown.

[0043] S105. The robot controls the robot to avoid obstacles based on the obstacle running trajectory.

[0044] In some embodiments of the present application, the specific process of controlling the robot to avoid obstacles based on the obstacle running trajectory includes: determining an obstacle avoidance path with reference to the obstacle running trajectory; generating an optimal obstacle avoidance path command according to the obstacle avoidance path; and driving the motor to perform steering / speed regulation operations through the optimal obstacle avoidance path command in combination with a PID controller to avoid obstacles.

[0045] In the embodiments of the present application, on the one hand, by quantifying the processing performance and pre-storing the system delay duration during the charging process of the robot, the pre-stored system delay duration is calculated and stored in advance, avoiding the time-consuming caused by real-time calculation, enabling the rapid query of the delay duration during the movement process to compensate the current spatial position information of the dynamic obstacle, so as to obtain the actual spatial position information closer to the real situation, and more accurate motion trajectory prediction can be performed based on this, effectively overcoming the motion prediction deviation. On the other hand, by loading a pre-trained reinforcement learning model, the spatial position calculation network in the model can consider the influence of system delay and calculate the actual spatial position of the dynamic obstacle more accurately. The reinforcement learning network further optimizes the prediction result through feature reinforcement, thereby outputting the predicted displacement delayed by the dynamic obstacle, enabling the robot to more accurately predict the motion trajectory of the dynamic obstacle, and thus reducing the collision risk.

[0046] Please refer to Figure 4 , which is a schematic flow chart of a model training method for a pre-trained reinforcement learning model provided by the embodiments of the present application. As Figure 4 shown, the method of the embodiments of the present application may include the following steps: S201, build a robot obstacle avoidance simulation environment including dynamic obstacles; S202, in the robot obstacle avoidance simulation environment, control the simulated robot to interact with the simulated obstacle; S203, during the interaction process, collect a preset number of multiple quadruple data; Among them, each quadruple data includes the simulation system delay duration, the current spatial position of the simulated obstacle, and the actual spatial position of the simulated obstacle; the simulation system delay duration is obtained by simulating the processing time of the simulated robot under different loads, the current spatial position of the simulated obstacle is the initial position simulated by the simulated obstacle before trajectory prediction, and the actual spatial position of the simulated obstacle is the final position when the trajectory prediction is completed; S204, use the neural network algorithm to create a spatial position calculation network for calculating the actual spatial position of the dynamic obstacle and a reinforcement learning network for feature reinforcement; S205, integrate the network parameters of the spatial position calculation network and the reinforcement learning network to obtain a reinforcement learning model; S206, input each quadruple data into the reinforcement learning model and output the model loss value; In some embodiments of the present application, the specific process of inputting each quadruple data into the reinforcement learning model and outputting the model loss value includes: inputting the simulation system delay duration and the current spatial position of the simulation obstacle into the spatial position calculation network to output the simulation position result; calculating the position gap between the simulation position result and the actual spatial position of the simulation obstacle; in the case where the position gap is greater than or equal to the preset threshold, adjusting the network parameters of the spatial position calculation network, and associating the simulation system delay duration, the current spatial position of the simulation obstacle with the actual spatial position of the simulation obstacle to obtain an associated sample; inputting the associated sample into the spatial position calculation network and the reinforcement learning network simultaneously for machine learning to output the first network loss value and the second network loss value; summing the first network loss value and the second network loss value as the model loss value.

[0047] Among them, the spatial position calculation network includes a calculation layer, a displacement prediction layer, a displacement decomposition layer, and a spatial position compensation layer, for example Figure 5 as shown.

[0048] Specifically, the calculation layer calculates the instantaneous linear velocity and instantaneous acceleration of the simulation obstacle according to the current spatial position of the simulation obstacle; the displacement prediction layer determines the predicted displacement delayed by the simulation obstacle according to the instantaneous linear velocity and instantaneous acceleration of the simulation obstacle; the displacement decomposition layer decomposes the predicted displacement delayed by the simulation obstacle in the abscissa and ordinate of the coordinate system to obtain the horizontal simulation coordinate compensation component and the simulation vertical coordinate compensation component; the spatial position compensation layer uses the horizontal simulation coordinate compensation component and the simulation vertical coordinate compensation component to perform spatial position compensation on the current spatial position of the simulation obstacle to obtain the actual spatial position of the simulation obstacle as the simulation position result.

[0049] S207. When the model loss value reaches the minimum, a pre-trained reinforcement learning model is obtained.

[0050] In the embodiments of the present application, on the one hand, by quantifying the processing performance and pre-storing the system delay duration during the robot charging process, the pre-stored system delay duration is calculated and stored in advance, avoiding the time-consuming of real-time calculation, enabling quick query of the delay duration during the traveling process to compensate the current spatial position information of the dynamic obstacle, so as to obtain the actual spatial position information closer to the real situation, and more accurate motion trajectory prediction can be performed based on this, effectively overcoming the motion prediction deviation. On the other hand, by loading the pre-trained reinforcement learning model, the spatial position calculation network in the model can consider the influence of system delay and calculate the actual spatial position of the dynamic obstacle more accurately. The reinforcement learning network further optimizes the prediction result through feature reinforcement, thereby outputting the predicted displacement delayed by the dynamic obstacle, enabling the robot to more accurately predict the motion trajectory of the dynamic obstacle, and thus reducing the collision risk.

[0051] The following is an embodiment of the device of the present application, which can be used to execute the method embodiment of the present application. For details not disclosed in the device embodiment of the present application, please refer to the method embodiment of the present application.

[0052] Please refer to Figure 6 , which shows a schematic structural diagram of a robot obstacle avoidance device applicable to dynamic obstacles provided by an exemplary embodiment of the present application. The robot obstacle avoidance device applicable to dynamic obstacles can be implemented as all or part of an electronic device through software, hardware, or a combination of both. The device 1 includes an environmental perception information acquisition module 10, a system delay duration acquisition module 20, a spatial position compensation module 30, an obstacle movement trajectory prediction module 40, and an obstacle avoidance control module 50.

[0053] The environmental perception information acquisition module 10 is configured to acquire environmental perception information in real time during the process of moving along a preset travel route; The system delay duration acquisition module 20 is configured to acquire the current spatial position information of the dynamic obstacle and the pre-stored system delay duration when it is determined that there is a dynamic obstacle within a preset range based on the environmental perception information; the pre-stored system delay duration is obtained by quantitatively storing the processing performance of the robot during the charging process of the robot; The spatial position compensation module 30 is configured to perform spatial position compensation on the current spatial position information according to the pre-stored system delay duration to obtain the actual spatial position information of the dynamic obstacle; The obstacle movement trajectory prediction module 40 is configured to predict the obstacle movement trajectory of the dynamic obstacle according to the actual spatial position information; The obstacle avoidance control module 50 is configured to control the robot to avoid obstacles based on the obstacle movement trajectory.

[0054] It should be noted that when the above-mentioned robot obstacle avoidance device applicable to dynamic obstacles executes the robot obstacle avoidance method applicable to dynamic obstacles, only the above-mentioned division of each functional module is used for illustration. In actual application, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the above-mentioned robot obstacle avoidance device applicable to dynamic obstacles and the method embodiment of the robot obstacle avoidance method applicable to dynamic obstacles belong to the same concept. The implementation process is detailed in the method embodiment and will not be repeated here.

[0055] The serial numbers of the above embodiments of the present application are only for description and do not represent the advantages and disadvantages of the embodiments.

[0056] In an embodiment of the present application, on the one hand, by quantifying the processing performance during the charging process of the robot and pre-storing the system delay duration, which is calculated and stored in advance, the time-consuming caused by real-time calculation is avoided, so that the delay duration can be quickly queried during the movement process to compensate the current spatial position information of the dynamic obstacle, thereby obtaining the actual spatial position information closer to the real situation, and more accurate motion trajectory prediction can be performed based on this, effectively overcoming the motion prediction deviation. On the other hand, by loading a pre-trained reinforcement learning model, the spatial position calculation network in the model can consider the influence of system delay and calculate the actual spatial position of the dynamic obstacle more accurately. The reinforcement learning network further optimizes the prediction result through feature reinforcement, thereby outputting the predicted displacement delayed by the dynamic obstacle, enabling the robot to more accurately predict the motion trajectory of the dynamic obstacle, and thus reducing the collision risk.

[0057] The present application also provides a computer-readable medium, on which program instructions are stored. When the program instructions are executed by a processor, the robot obstacle avoidance method applicable to dynamic obstacles provided by each of the above method embodiments is implemented.

[0058] The present application also provides a computer program product containing instructions. When it runs on a computer, the computer is enabled to execute the robot obstacle avoidance method applicable to dynamic obstacles provided by each of the above method embodiments.

[0059] Please refer to Figure 7 , which is a schematic structural diagram of an electronic device provided by an embodiment of the present application. As Figure 7 shown, the electronic device 1000 may include: at least one processor 1001, at least one network interface 1004, a user interface 1003, a memory 1005, and at least one communication bus 1002.

[0060] Among them, the communication bus 1002 is used to realize the connection and communication between these components.

[0061] Among them, the user interface 1003 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 1003 may further include a standard wired interface and a wireless interface.

[0062] Among them, the network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface).

[0063] Among them, the processor 1001 may include one or more processing cores. The processor 1001 connects various parts within the entire electronic device 1000 through various interfaces and circuits. By running or executing instructions, programs, code sets, or instruction sets stored in the memory 1005, and by calling the data stored in the memory 1005, it performs various functions of the electronic device 1000 and processes data. Optionally, the processor 1001 may be implemented in at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), or programmable logic array (PLA). The processor 1001 may integrate one or a combination of several of a central processing unit (CPU), a graphics processing unit (GPU), and a modem, etc. Among them, the CPU mainly processes the operating system, user interface, application programs, etc.; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 1001 and may be implemented separately by a single chip.

[0064] Among them, the memory 1005 may include a random access memory (RAM) and may also include a read-only memory. Optionally, the memory 1005 includes a non-transitory computer-readable storage medium. The memory 1005 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 1005 may include a program storage area and a data storage area. Among them, the program storage area can store instructions for implementing the operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area can store the data involved in the above-mentioned various method embodiments. Optionally, the memory 1005 may also be at least one storage system located far from the aforementioned processor 1001. As Figure 7 shown, the memory 1005, as a computer storage medium, may include an operating system, a network communication module, a user interface module, and a robot obstacle avoidance application program suitable for dynamic obstacles.

[0065] In Figure 7In the illustrated electronic device 1000, the user interface 1003 is mainly used to provide an interface for the user to input and obtain the data input by the user. The processor 1001 can be used to call the robot obstacle avoidance application program applicable to dynamic obstacles stored in the memory 1005 and specifically perform the following operations: During the process of traveling along the preset travel route, obtain the environmental perception information in real time; When it is determined based on the environmental perception information that there are dynamic obstacles within the preset range, obtain the current spatial position information of the dynamic obstacles and the pre-stored system delay duration. The pre-stored system delay duration is quantitatively stored for the processing performance of the robot during the charging process of the robot; According to the pre-stored system delay duration, perform spatial position compensation on the current spatial position information to obtain the actual spatial position information of the dynamic obstacles; According to the actual spatial position information, predict the obstacle running trajectory of the dynamic obstacles; Based on the obstacle running trajectory, control the robot to avoid obstacles.

[0066] In one embodiment, when the processor 1001 performs spatial position compensation on the current spatial position information according to the pre-stored system delay duration to obtain the actual spatial position information of the dynamic obstacles, it specifically performs the following operations: According to the current spatial position information and the historical spatial position information of the dynamic obstacles at the previous moment, calculate the instantaneous linear velocity and instantaneous acceleration of the dynamic obstacles; According to the pre-stored system delay duration, instantaneous linear velocity and instantaneous acceleration, determine the predicted displacement delayed by the dynamic obstacles; Decompose the predicted displacement into the horizontal and vertical coordinates of the coordinate system to obtain the horizontal coordinate compensation component and the vertical coordinate compensation component; Use the horizontal coordinate compensation component and the vertical coordinate compensation component to perform spatial position compensation on the current spatial position information to obtain the actual spatial position information of the dynamic obstacles.

[0067] In one embodiment, when the processor 1001 determines the predicted displacement delayed by the dynamic obstacles according to the pre-stored system delay duration, instantaneous linear velocity and instantaneous acceleration, it specifically performs the following operations: Call the preset obstacle position compensation model; Use the pre-stored system delay duration, instantaneous linear velocity and instantaneous acceleration as the model parameters of the preset obstacle position compensation model to obtain the model expression with parameter supplementation; Execute the model expression with parameter supplementation to obtain the predicted displacement delayed by the dynamic obstacles; Among them, the model expression of the preset obstacle position compensation model is:

[0068] Among them, is the predicted displacement, is the instantaneous linear velocity of the dynamic obstacle, is the instantaneous acceleration of the dynamic obstacle, is the system processing delay time; Among them, the calculation formula for the actual spatial position information of the dynamic obstacle is: ; Among them, is the actual spatial position information, is the abscissa compensation component, is the ordinate compensation component, is the abscissa of the current spatial position information, is the ordinate of the current spatial position information.

[0069] In one embodiment, when the processor 1001 executes to generate the pre-stored system delay duration, the following operations are specifically performed: During the charging process of the robot, obtain the current load and task queue length of the processor as the processor operation parameters; At the same time, extract the detection function for detecting the spatial position information of the dynamic obstacle and the prediction function for predicting the dynamic obstacle; In the simulation platform, create the virtual spatial position information of the dynamic virtual obstacle; Perform the simulation operation of the obstacle running trajectory according to the virtual spatial position information, detection function and prediction function; Statistical function execution time during the simulation operation process; Determine the processor operation time of the processor according to the current load and task queue length; Accumulate the function execution time and the processor operation time, and store the accumulated result to obtain the pre-stored system delay duration.

[0070] In one embodiment, when the processor 1001 executes to determine the predicted displacement delayed by the dynamic obstacle according to the pre-stored system delay duration, instantaneous linear velocity and instantaneous acceleration, the following operations are specifically performed: Load the pre-trained reinforcement learning model; Input the pre-stored system delay duration, instantaneous linear velocity and instantaneous acceleration into the pre-trained reinforcement learning model; the reinforcement learning model includes a spatial position calculation network for calculating the actual spatial position of the dynamic obstacle and a reinforcement learning network for feature reinforcement; Output the predicted displacement delayed by the dynamic obstacle.

[0071] In one embodiment, when the processor 1001 executes to generate a pre-trained reinforcement learning model, it specifically performs the following operations: Build a robot obstacle avoidance simulation environment including dynamic obstacles; In the robot obstacle avoidance simulation environment, control the simulated robot to interact with the simulated obstacles; During the interaction process, collect a preset number of multiple quadruple data; where Each quadruple data includes the simulation system delay duration, the current spatial position of the simulated obstacle, and the actual spatial position of the simulated obstacle; the simulation system delay duration is obtained by simulating the processing time of the simulated robot under different loads, the current spatial position of the simulated obstacle is the initial position simulated before trajectory prediction, and the actual spatial position of the simulated obstacle is the final position when the trajectory prediction is completed; Use a neural network algorithm to create a spatial position calculation network for calculating the actual spatial position of the dynamic obstacle and a reinforcement learning network for feature reinforcement; Integrate the network parameters of the spatial position calculation network and the reinforcement learning network to obtain a reinforcement learning model; Input each quadruple data into the reinforcement learning model and output the model loss value; When the model loss value reaches the minimum, obtain the pre-trained reinforcement learning model.

[0072] In one embodiment, when the processor 1001 executes to input each quadruple data into the reinforcement learning model and output the model loss value, it specifically performs the following operations: Input the simulation system delay duration and the current spatial position of the simulated obstacle into the spatial position calculation network and output the simulation position result; Calculate the position gap between the simulation position result and the actual spatial position of the simulated obstacle; In the case where the position gap is greater than or equal to the preset threshold, adjust the network parameters of the spatial position calculation network and associate the simulation system delay duration, the current spatial position of the simulated obstacle with the actual spatial position of the simulated obstacle to obtain an associated sample; Input the associated sample into the spatial position calculation network and the reinforcement learning network simultaneously for machine learning and output the first network loss value and the second network loss value; Sum the first network loss value and the second network loss value as the model loss value.

[0073] In one embodiment, when the processor 1001 executes the process of inputting the simulation system delay duration and the current spatial position of the simulated obstacle into the spatial position calculation network and outputting the simulation position result, it specifically performs the following operations: The calculation layer calculates the instantaneous linear velocity and instantaneous acceleration of the simulated obstacle based on the current spatial position of the simulated obstacle; The displacement prediction layer determines the predicted displacement delayed by the simulated obstacle based on the instantaneous linear velocity and instantaneous acceleration of the simulated obstacle; The displacement decomposition layer decomposes the predicted displacement delayed by the simulated obstacle in the abscissa and ordinate of the coordinate system to obtain the horizontal simulated coordinate compensation component and the simulated ordinate compensation component; The spatial position compensation layer uses the horizontal simulated coordinate compensation component and the simulated ordinate compensation component to perform spatial position compensation on the current spatial position of the simulated obstacle to obtain the actual spatial position of the simulated obstacle as the simulation position result.

[0074] In one embodiment, when the processor 1001 executes to predict the obstacle movement trajectory of the dynamic obstacle according to the actual spatial position information, the following operations are specifically performed: Construct a second-order kinematic model; wherein, the model expression of the second-order kinematic model is:

[0075] Wherein, and are the predicted position coordinates of the dynamic obstacle at a future moment, is the current moment, is the time interval for future trajectory prediction, and are the actual spatial position information of the dynamic obstacle after system delay compensation, and are the current velocity components of the dynamic obstacle in the axis and axis directions, is the pre-stored system delay duration, and are the acceleration components of the dynamic obstacle in the axis and axis directions; Calculate the movement trend of the dynamic obstacle according to the actual spatial position information, the current velocity of the dynamic obstacle, and the second-order kinematic model; Draw the position sequence of multiple time points under the movement trend to obtain discrete position points; Connect the discrete position points into line segments to obtain the obstacle movement trajectory of the dynamic obstacle.

[0076] In the embodiments of the present application, on the one hand, by quantifying the processing performance and pre-storing the system delay duration during the charging process of the robot, the pre-stored system delay duration is calculated and stored in advance, avoiding the time-consuming caused by real-time calculation, enabling the quick query of the delay duration during the movement process to compensate the current spatial position information of the dynamic obstacle, so as to obtain the actual spatial position information closer to the real situation, and based on this, a more accurate motion trajectory prediction can be made, effectively overcoming the motion prediction deviation. On the other hand, by loading a pre-trained reinforcement learning model, the spatial position calculation network in the model can consider the influence of the system delay and calculate the actual spatial position of the dynamic obstacle more accurately. The reinforcement learning network further optimizes the prediction result through feature reinforcement, thereby outputting the predicted displacement delayed by the dynamic obstacle, enabling the robot to more accurately predict the motion trajectory of the dynamic obstacle, thus reducing the collision risk.

[0077] Those of ordinary skill in the art can understand that all or part of the processes in the above-described method embodiments can be completed by instructing relevant hardware through a computer program. The program applicable to the robot obstacle avoidance for dynamic obstacles can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above-described method embodiments. Among them, the storage medium of the program applicable to the robot obstacle avoidance for dynamic obstacles can be a magnetic disk, an optical disk, a read-only memory, or a random access memory, etc.

[0078] The above-disclosed are only the preferred embodiments of the present application. Of course, the scope of rights of the present application cannot be limited thereby. Therefore, equivalent changes made according to the claims of the present application still fall within the scope covered by the present application.

Claims

1. A robot obstacle avoidance method applicable to dynamic obstacles, characterized in that, Applied to a robot, the method includes: During the process of traveling along a preset travel route, obtaining environmental perception information in real time; When it is determined based on the environmental perception information that there is a dynamic obstacle within a preset range, obtaining the current spatial position information of the dynamic obstacle and the pre-stored system delay duration; the pre-stored system delay duration is quantitatively stored for the processing performance of the robot during the charging process of the robot; According to the pre-stored system delay duration, performing spatial position compensation on the current spatial position information to obtain the actual spatial position information of the dynamic obstacle; According to the actual spatial position information, predicting the obstacle movement trajectory of the dynamic obstacle; Based on the obstacle movement trajectory, controlling the robot to avoid obstacles.

2. The method according to claim 1, wherein The performing spatial position compensation on the current spatial position information according to the pre-stored system delay duration to obtain the actual spatial position information of the dynamic obstacle includes: According to the current spatial position information and the historical spatial position information of the dynamic obstacle at the previous moment, calculating the instantaneous linear velocity and instantaneous acceleration of the dynamic obstacle; According to the pre-stored system delay duration, the instantaneous linear velocity and the instantaneous acceleration, determining the predicted displacement delayed by the dynamic obstacle; Decomposing the predicted displacement into displacement components in the abscissa and ordinate of the coordinate system to obtain an abscissa compensation component and an ordinate compensation component; Using the abscissa compensation component and the ordinate compensation component to perform spatial position compensation on the current spatial position information to obtain the actual spatial position information of the dynamic obstacle.

3. The method according to claim 2, characterized in that, The determining the predicted displacement delayed by the dynamic obstacle according to the pre-stored system delay duration, the instantaneous linear velocity and the instantaneous acceleration includes: Invoking a preset obstacle position compensation model; Taking the pre-stored system delay duration, the instantaneous linear velocity and the instantaneous acceleration as the model parameters of the preset obstacle position compensation model to obtain a model expression with parameter supplementation; Executing the model expression with parameter supplementation to obtain the predicted displacement delayed by the dynamic obstacle; Wherein, the model expression of the preset obstacle position compensation model is: ； wherein, is the predicted displacement, is the instantaneous linear velocity of the dynamic obstacle, is the instantaneous acceleration of the dynamic obstacle, is the system processing delay time; Wherein, the calculation formula for the actual spatial position information of the dynamic obstacle is: ； Among them, is the actual spatial position information, is the abscissa compensation component, is the ordinate compensation component, is the abscissa of the current spatial position information, is the ordinate of the current spatial position information.

4. The method according to any one of claims 1 to 3, characterized in that, Generating the pre-stored system delay duration according to the following steps, including: During the charging process of the robot, obtaining the current load and task queue length of the processor as the processor operation parameters; Simultaneously extracting a detection function for detecting the spatial position information of the dynamic obstacle and a prediction function for predicting the dynamic obstacle; In a simulation platform, creating virtual spatial position information of a dynamic virtual obstacle; Performing simulation operation of the obstacle movement trajectory according to the virtual spatial position information, the detection function and the prediction function; Counting the function execution time during the simulation operation process; According to the current load and task queue length, determining the processor operation time of the processor; Accumulating the function execution time and the processor operation time and storing the accumulated result to obtain the pre-stored system delay duration.

5. The method according to claim 2, wherein Determining the predicted displacement delayed by the dynamic obstacle according to the pre-stored system delay duration, the instantaneous linear velocity, and the instantaneous acceleration includes: Loading a pre-trained reinforcement learning model; Inputting the pre-stored system delay duration, the instantaneous linear velocity, and the instantaneous acceleration into the pre-trained reinforcement learning model; the reinforcement learning model includes a spatial position calculation network for calculating the actual spatial position of the dynamic obstacle and a reinforcement learning network for feature reinforcement; Outputting the predicted displacement delayed by the dynamic obstacle.

6. The method according to claim 5, wherein Generating the pre-trained reinforcement learning model in the following manner, including: Building a robot obstacle avoidance simulation environment containing dynamic obstacles; Controlling the simulated robot to interact with the simulated obstacles in the robot obstacle avoidance simulation environment; During the interaction, collecting a preset number of multiple quadruple data; where Each quadruple data includes the simulation system delay duration, the current spatial position of the simulated obstacle, and the actual spatial position of the simulated obstacle; the simulation system delay duration is obtained by simulating the processing time of the simulated robot under different loads, the current spatial position of the simulated obstacle is the initial position simulated by the simulated obstacle before trajectory prediction, and the actual spatial position of the simulated obstacle is the final position of the simulated obstacle when the trajectory prediction is completed; Using a neural network algorithm to create a spatial position calculation network for calculating the actual spatial position of the dynamic obstacle and a reinforcement learning network for feature reinforcement; Integrating the network parameters of the spatial position calculation network and the reinforcement learning network to obtain a reinforcement learning model; Inputting each quadruple data into the reinforcement learning model and outputting a model loss value; When the model loss value reaches the minimum, obtaining the pre-trained reinforcement learning model.

7. The method according to claim 6, characterized in that, The step of inputting each quadruple data into the reinforcement learning model and outputting a model loss value includes: Inputting the simulation system delay duration and the current spatial position of the simulated obstacle into the spatial position calculation network and outputting a simulation position result; Calculating the position difference between the simulation position result and the actual spatial position of the simulated obstacle; In the case where the position difference is greater than or equal to a preset threshold, adjusting the network parameters of the spatial position calculation network and associating the simulation system delay duration, the current spatial position of the simulated obstacle, and the actual spatial position of the simulated obstacle to obtain an associated sample; Inputting the associated sample into both the spatial position calculation network and the reinforcement learning network for machine learning and outputting a first network loss value and a second network loss value; Summing the first network loss value and the second network loss value as the model loss value.

8. The method according to claim 6, wherein The spatial position calculation network includes a calculation layer, a displacement prediction layer, a displacement decomposition layer, and a spatial position compensation layer; The step of inputting the simulation system delay duration and the current spatial position of the simulated obstacle into the spatial position calculation network and outputting a simulation position result includes: The calculation layer calculates the instantaneous linear velocity and the instantaneous acceleration of the simulated obstacle according to the current spatial position of the simulated obstacle; The displacement prediction layer determines the predicted displacement delayed by the simulated obstacle according to the instantaneous linear velocity and instantaneous acceleration of the simulated obstacle; The displacement decomposition layer decomposes the predicted displacement delayed by the simulated obstacle in the abscissa and ordinate of the coordinate system to obtain the horizontal simulation coordinate compensation component and the simulation vertical coordinate compensation component; The spatial position compensation layer uses the horizontal simulation coordinate compensation component and the simulation vertical coordinate compensation component to perform spatial position compensation on the current spatial position of the simulated obstacle to obtain the actual spatial position of the simulated obstacle as the simulation position result.

9. The method according to claim 1, wherein Predicting the obstacle movement trajectory of the dynamic obstacle according to the actual spatial position information includes: Constructing a second-order kinematic model; where the model expression of the second-order kinematic model is: Wherein, and are the predicted position coordinates of the dynamic obstacle at a future moment, is the current moment, is the time interval for future trajectory prediction, and are the actual spatial position information of the dynamic obstacle after system delay compensation, and are the current velocity components of the dynamic obstacle in the axis and axis directions, is the pre-stored system delay duration, and are the acceleration components of the dynamic obstacle in the axis and axis directions; Calculating the movement trend of the dynamic obstacle according to the actual spatial position information, the current speed of the dynamic obstacle, and in combination with the second-order kinematic model; Drawing the position sequence of multiple time points under the movement trend to obtain discrete position points; Connecting the discrete position points into line segments to obtain the obstacle movement trajectory of the dynamic obstacle.

10. A robot obstacle avoidance device applicable to dynamic obstacles, characterized in that, The device includes: An environmental perception information acquisition module, configured to acquire environmental perception information in real time during the process of traveling according to a preset traveling route; A system delay duration acquisition module, configured to acquire the current spatial position information and the pre-stored system delay duration of the dynamic obstacle when it is determined that there is a dynamic obstacle within a preset range based on the environmental perception information; the pre-stored system delay duration is quantitatively stored for the processing performance of the robot during the charging process of the robot; A spatial position compensation module, configured to perform spatial position compensation on the current spatial position information according to the pre-stored system delay duration to obtain the actual spatial position information of the dynamic obstacle; An obstacle movement trajectory prediction module, configured to predict the obstacle movement trajectory of the dynamic obstacle according to the actual spatial position information; An obstacle avoidance control module, configured to control the robot to avoid obstacles based on the obstacle movement trajectory.

Images