Robot repositioning method and device, electronic equipment and readable storage medium

Abstract

The embodiment of the invention provides a robot repositioning method and device, electronic equipment and a readable storage medium, and relates to the technical field of robots. The method comprises the steps that under the condition that positioning of the robot is lost, movement information is obtained according to first environment information currently obtained by the robot, and the robot is controlled to move according to the movement information; whether the robot moves to the target local area or not is judged, and under the condition that it is determined that the robot moves to the target local area, first pose prediction information obtained based on first environment information obtained when the robot moves to the target local area through a pre-trained positioning network serves as a repositioning result of the robot; the positioning network is obtained through training based on robot pose information in a target local area in the target scene and corresponding environment information, and the target local area is a part of area of the target scene. In this way, rapid repositioning can be achieved through a small amount of calculation in a large scene.

Description

Technical Field

[0001] This application relates to the field of robot technology. Specifically, it relates to a robot relocalization method, device, electronic device, and readable storage medium. Background Art

[0002] Traditional robot relocalization methods usually use filter technology and perform iterative searches in the pose space. When dealing with large task scenarios, this method consumes a large amount of time for convergence calculation, resulting in low relocalization efficiency. Summary of the Invention

[0003] Embodiments of this application provide a robot relocalization method, device, electronic device, and readable storage medium, which can achieve fast relocalization with less computational effort in large scenarios.

[0004] The embodiments of this application can be implemented as follows:

[0005] In a first aspect, an embodiment of this application provides a robot relocalization method, and the method includes:

[0006] In the case where the robot's positioning is lost, obtain movement information based on the first environmental information currently obtained by the robot, and control the movement of the robot according to the movement information;

[0007] Determine whether the robot has moved to a target local area, and in the case where it is determined that the robot has moved to the target local area, use the first pose prediction information obtained by the pre-trained positioning network based on the first environmental information obtained when the robot moves to the target local area as the relocalization result of the robot. Herein, the positioning network is trained based on the robot pose information and the corresponding environmental information in the target local area in the target scene, and the target local area is a part of the target scene.

[0008] In a second aspect, an embodiment of this application provides a robot relocalization device, and the device includes:

[0009] An autonomous movement module, configured to obtain movement information based on the first environmental information currently obtained by the robot in the case where the robot's positioning is lost, and control the movement of the robot according to the movement information;

[0010] A processing module is configured to determine whether the robot has moved to a target local area, and in the case where it is determined that the robot has moved to the target local area, use a pre-trained positioning network to obtain a first pose prediction information based on the first environmental information obtained when the robot moves to the target local area as the repositioning result of the robot. Wherein, the positioning network is trained based on the robot pose information and the corresponding environmental information in the target local area of the target scene, and the target local area is a part of the target scene.

[0011] In a third aspect, an embodiment of the present application provides an electronic device, including a processor and a memory. The memory stores machine-executable instructions that can be executed by the processor, and the processor can execute the machine-executable instructions to implement the robot repositioning method described in the foregoing embodiments.

[0012] In a fourth aspect, an embodiment of the present application provides a readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the robot repositioning method described in the foregoing embodiments.

[0013] The robot repositioning method, device, electronic device, and readable storage medium provided by the embodiments of the present application control the movement of the robot according to the movement information obtained from the first environmental information currently obtained by the robot in the case of lost robot positioning. During the movement, it is determined whether the robot has moved to the target local area, and in the case where it is determined that the robot has moved to the target local area, the first pose prediction information obtained by using the pre-trained positioning network based on the first environmental information obtained when the robot moves to the target local area is used as the repositioning result of the robot. Wherein, the positioning network is trained based on the robot pose information and the corresponding environmental information in the target local area of the target scene, and the target local area is a part of the target scene. In this way, fast repositioning can be achieved with less computational effort in a large scene. Description of the Drawings

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present application and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0015] Figure 1 It is a block diagram of the electronic device provided by the embodiment of the present application;

[0016] Figure 2Schematic flowchart of the robot relocalization method provided by an embodiment of this application;

[0017] Figure 3 Schematic diagram of steps for obtaining a localization network;

[0018] Figure 4 Schematic diagram of steps for obtaining a policy network;

[0019] Figure 5 For Figure 2 Schematic flowchart of sub-steps included in step S210 in

[0020] Figure 6 For Figure 2 Schematic flowchart of sub-steps included in step S220 in

[0021] Figure 7 For Figure 6 Schematic flowchart of sub-steps included in sub-step S223 in

[0022] Figure 8 Block diagram of the robot relocalization device provided by an embodiment of this application.

[0023] Icons: 100 - electronic device; 110 - memory; 120 - processor; 130 - communication unit; 200 - robot relocalization device; 210 - autonomous movement module; 220 - processing module. Detailed implementation manners

[0024] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the following will clearly and completely describe the technical solutions in the embodiments of this application with reference to the accompanying drawings in the embodiments of this application. Apparently, the described embodiments are some but not all of the embodiments of this application. Usually, the components of the embodiments of this application described and illustrated herein can be arranged and designed in various different configurations.

[0025] Therefore, the following detailed description of the embodiments of this application provided in the drawings is not intended to limit the scope of this application that is required to be protected, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative efforts fall within the scope of protection of this application.

[0026] It should be noted that relational terms such as "first" and "second" are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the phrase "comprising a..." does not exclude the presence of additional identical elements in the process, method, article or device comprising said element.

[0027] In addition to using filter technology for robot relocalization, in recent years, with the rapid development of deep learning technology, researchers have also studied many neural network-based robot relocalization methods. The computational complexity of these neural network-based relocalization methods only depends on the structure of the neural network itself and is independent of the area size of the task scenario. However, the effectiveness of this end-to-end relocalization method mainly depends on the number of training samples. However, in large scenarios, the cost of robots collecting enough training samples is usually very high, and the relocalization problem in large scenarios has not been completely solved yet.

[0028] In view of the above situation, the embodiments of the present application provide a robot relocalization method, device, electronic device and readable storage medium, which can achieve fast relocalization with less computational effort in large scenarios.

[0029] The following will describe in detail some embodiments of the present application with reference to the accompanying drawings. Without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0030] Please refer to Figure 1 , Figure 1 , which is a block diagram of the electronic device 100 provided by the embodiments of the present application. The electronic device 100 can be, but is not limited to, a computer, a server, a robot or a control unit of a robot, etc. The electronic device 100 may include a memory 110, a processor 120 and a communication unit 130. The elements of the memory 110, the processor 120 and the communication unit 130 are directly or indirectly electrically connected to each other to achieve data transmission or interaction. For example, these elements may be electrically connected to each other through one or more communication buses or signal lines.

[0031] Among them, the memory 110 is used to store programs or data. The memory 110 can be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), etc.

[0032] The processor 120 is used to read / write the data or programs stored in the memory 110 and perform corresponding functions. For example, a robot repositioning device 200 is stored in the memory 110. The robot repositioning device 200 includes at least one software function module that can be stored in the memory 110 in the form of software or firmware. The processor 120 executes various functional applications and data processing by running the software programs and modules stored in the memory 110, such as the robot repositioning device 200 in the embodiments of the present application, thereby implementing the robot repositioning method in the embodiments of the present application.

[0033] The communication unit 130 is used to establish a communication connection between the electronic device 100 and other communication terminals through a network, and is used to transmit and receive data through the network.

[0034] It should be understood that Figure 1 The structure shown is only a schematic diagram of the structure of the electronic device 100. The electronic device 100 may further include more or fewer components than those shown Figure 1 in the figure, or have a different configuration from that shown Figure 1 in the figure. Figure 1 Each component shown in the figure can be implemented by hardware, software, or a combination thereof.

[0035] Please refer to Figure 2 , Figure 2 which is a schematic flowchart of the robot repositioning method provided by the embodiments of the present application. The method can be applied to the above-mentioned electronic device. The specific process of the robot repositioning method will be elaborated in detail below. In this embodiment, the method may include step S210 to step S220.

[0036] Step S210, in the case where the robot positioning is lost, obtain movement information according to the first environmental information currently obtained by the robot, and control the movement of the robot according to the movement information.

[0037] In step S220, it is determined whether the robot has moved to the target local area. When it is determined that the robot has moved to the target local area, the first pose prediction information obtained by using the pre-trained positioning network based on the first environmental information obtained when the robot moves to the target local area is used as the relocalization result of the robot.

[0038] In this embodiment, when the robot moves in the target scenario, it can be determined in any way whether the robot's positioning is lost. When it is determined that the robot's positioning is lost, the information of the environment where the robot is currently located can be obtained as the first environmental information, and then the movement information can be analyzed and determined based on this first environmental information, and then the robot's movement can be controlled based on this movement information. Among them, the first environmental information is the environmental observation information of the robot, and can be specifically determined in combination with the sensors provided on the robot. For example, if a lidar is provided on the robot, the environmental information can be lidar point cloud data; for another example, if a camera is provided on the robot, the environmental information can be the image collected by the camera. It can be understood that different sensors provide environmental information in different formats. The movement information is used to indicate movement actions, and the movement actions can be defined according to the modeled scenario. For example, in a task scenario on a two-dimensional plane, the movement actions can be moving up, moving down, moving left, moving right, etc.

[0039] When it is continuously determined that the robot's positioning is lost (that is, the relocalization result cannot be obtained), the corresponding movement information can be continuously obtained according to the environmental information of the current environment, and the robot's movement can be performed based on this movement information. That is, before the robot completes relocalization, the robot continuously moves autonomously in the above manner.

[0040] Among them, optionally, it can be determined how to determine the movement information based on the first environmental information according to actual needs. For example, the movement information can be determined based on the first environmental information based on the requirement of collision-free.

[0041] When the robot's positioning is lost and during the autonomous movement of the robot, it can be determined whether the robot has moved to the target local area. Among them, the specific method for determining whether the robot has moved to the target local area can be determined according to actual needs; for example, sensors are arranged at the edge of the target local area, and when the robot senses the sensor, it can be determined that it has moved to the target local area. It can be understood that the above method is only an example, and other methods can also be used to determine whether the robot has moved to the target local area. The target local area is a part of the target scene, that is, the area of the target local area is smaller than the area of the target scene. The specific size of the target local area can be determined according to actual needs. For example, when the target scene is a large task scene, the target local area can be an area covering 20%-30% of the area of the target scene.

[0042] The electronic device may store a pre-trained positioning network. The positioning network is trained based on the robot pose information and the corresponding environmental information in the target local area of the target scene. The specific training method and training device can be determined according to actual needs and will not be specifically limited here. The positioning network can predict the corresponding robot pose information when obtaining the environmental information based on the environmental information within the target local area.

[0043] When it is determined that the robot has moved to the target local area, the first pose prediction information of the positioning network based on the first environmental information obtained when the robot moves to the target local area can be determined. The first pose prediction information can be used as the relocalization result of the robot, that is, the first pose prediction information is used as the pose information of the robot at this time.

[0044] In this embodiment, when the robot's positioning is lost, the robot moves autonomously based on the currently obtained environmental information. When it is determined that it has autonomously moved to the target local area, the pose information predicted by the positioning network based on the environmental information obtained when moving to the target local area is used as the relocalization result of the robot. The positioning network is a network pre-trained using the pose information of the robot in the target local area and the obtained corresponding environmental information. In this way, fast active relocalization can be achieved with less computational effort in a large target scene.

[0045] As a possible implementation, the positioning network can be obtained through Figure 3 the steps S110 to S120 shown. The positioning network can be trained by the electronic device or can be trained by other devices and sent to the electronic device.

[0046] Step S110, control the robot to perform autonomous movement exploration, and obtain the pose information and corresponding environment information of the robot during the autonomous movement exploration.

[0047] Step S120, according to the pose information and the corresponding environment information, obtain the positioning network through deep learning.

[0048] In this embodiment, the robot can be made to perform autonomous movement in the target scene based on the exploration environment strategy. When the robot moves autonomously in the target scene, it can record its own pose information and the environment information of the current pose at each moment. Among them, the pose information can be provided by the robot's positioning algorithm, and the environment information can be provided by the robot's sensor. For example, the positioning algorithm usually uses a filter-based global positioning algorithm, and the sensor can use a lidar or a camera. Different sensors provide different formats of environment information.

[0049] The pose information and environment information obtained during the exploration of the environment are aligned according to the time stamp to form training samples. A finite number of training samples are collected to form a training sample set. It can be understood that the exploration area corresponding to the finite number of training samples is the area of the autonomous movement exploration, that is, the target local area. When the target scene is a large scene, the training sample set can be considered to cover an area of 20%-30% of the target scene.

[0050] After obtaining the training sample set, a end-to-end positioning neural network can be learned based on the training sample set using deep learning technology. The input of the positioning neural network is the environment information, and the output is the pose information of the robot. For example, the positioning neural network can be composed of a series of convolutional neural network layers and fully connected layers. The specific training algorithm can use the stochastic gradient descent method, and the optimization method can use the dropout method. The learned positioning neural network is the pre-trained positioning network.

[0051] As a possible implementation, an active path planner can also be obtained in advance. The active path planner can be a robot autonomous exploration model or a policy network. Subsequently, when the robot loses its positioning, it can perform autonomous movement based on the robot active path planner and then move to the target local area. Optionally, the policy network can be obtained through Figure 4 Steps S130 to S140 therein.

[0052] Step S130, control the robot to perform a relocalization experiment in the target scene and obtain the experiment information.

[0053] Step S140, according to the experiment information of each experiment, obtain the policy network through reinforcement learning.

[0054] In this embodiment, the robot can perform multiple relocalization experiments in the target scenario and obtain the experimental information of each relocalization experiment. The experimental information of each relocalization experiment includes the path trajectory during the relocalization process and the path planning process of the path trajectory. The path planning process includes a series of environmental information, the specific actions of the robot, and environmental rewards (i.e., including the reward feedback corresponding to each action in the path). Different environmental rewards are given according to the different environmental impacts generated by the robot's path trajectory each time. Then, combined with the collected path trajectories, the policy network of the robot is optimized through policy gradient learning. In this way, a policy network can be learned using reinforcement learning technology, enabling the robot to select actions at each moment through the policy network and actively move in the task scenario. The input of the policy network is environmental observation information, and based on the output of the policy network, the next action of the robot can be determined (this action is used to represent the movement direction, etc.).

[0055] Optionally, during the relocalization experiment, the input of the policy network is environmental information, and the output is the action probability distribution of the robot. The next action of the robot can be determined based on the probabilities corresponding to each action in the output. For example, random sampling can be performed according to the probabilities corresponding to each action to determine the next action of the robot.

[0056] During the process of training the policy network, the policy with a small reinforcement cost can be trained through reinforcement learning rewards. In this way, when the robot moves autonomously based on this policy network later, the robot can move to the target local area with as small a cost as possible.

[0057] When the robot moves in the target scenario but loses its position, the active path planner can obtain the movement information based on the first environmental information obtained currently. The active path planner is a model used to determine the specific movement actions of the robot at each moment, enabling the robot to move actively in the target scenario (i.e., the task scenario). That is to say, the active path planner is used to determine the movement information for the next moment based on the first environmental information at one moment, and the robot moves based on this movement information for the next moment. Among them, the active path planner can be, but is not limited to, an autonomous exploration model or a policy network obtained through training. The autonomous exploration model is used to determine the movement information used by the robot when moving autonomously in the target scenario using SLAM technology and a preset exploration strategy. In this way, the robot uses SLAM technology and the exploration strategy to move autonomously in the target scenario, such as exploring according to environmental feature information or a certain specific motion planning strategy.

[0058] As a possible implementation, the active path planner is a policy network. In this case, it can be achieved through Figure 5Obtain the corresponding movement information in the manner shown. Please refer to Figure 5 , Figure 5 is Figure 2 a schematic flowchart of the sub-steps included in step S210 in

[0059] Sub-step S211: Obtain different candidate actions and the probabilities corresponding to the candidate actions based on the policy network and the first environmental information.

[0060] Sub-step S212: Determine the target action from different candidate actions by random sampling according to the obtained probabilities.

[0061] In an embodiment, the first environmental information at a moment can be input into the policy network to obtain different candidate actions and the probabilities corresponding to each candidate action. Then, based on each candidate action and the probability corresponding to each candidate action, by random sampling, a candidate action is determined from the obtained different candidate actions as the target action. The action indicated by the movement information obtained based on the first environmental information at this moment is the target action.

[0062] After determining the movement information based on the first environmental information at a moment, the robot moves based on this movement information; repeating the above process can enable continuous movement. The environmental information obtained in advance within the target local area can be stored in the electronic device. During the process of the robot losing its position and moving autonomously, the first environmental information currently obtained by the robot can be compared with the environmental information of the saved target local area. If the similarity obtained by the comparison is higher than the threshold, it can be determined that the robot has moved to the target local area at this time. It is also possible to determine whether the robot has moved to the target local area at this time through other means.

[0063] As a possible implementation manner, it can be determined whether the robot has moved to the target local area through the Figure 6 method shown. Please refer to Figure 6 , Figure 6 is Figure 2 a schematic flowchart of the sub-steps included in step S220 in

[0064] Sub-step S221: During the process of the robot moving based on the movement information, use the positioning network to obtain the second pose prediction information corresponding to the first environmental information based on the currently obtained first environmental information.

[0065] Sub-step S222: Obtain the corresponding second environmental information according to the target local environmental map corresponding to the target local area and the second pose prediction information.

[0066] Sub-step S223: Determine whether the robot has moved to the target local area according to the first environmental information and the second environmental information.

[0067] In this embodiment, after obtaining the first environmental information at a certain moment, not only the corresponding movement information is obtained based on this first environment, but also the positioning network can be used to perform pose estimation based on this first environmental information to obtain the second pose prediction information corresponding to this first environmental information. Then, according to the target local environmental map corresponding to the saved target local area and this second pose prediction information, the second environmental information can be obtained. This second environmental information is the environmental information obtained when the pose information of the robot is the second pose prediction information, that is, this second environmental information is simulated environmental information. Then, based on this second environmental information and this first environmental information, it is determined whether the robot has moved to the target local area.

[0068] Among them, the specific method of obtaining the second environmental information according to the target local environmental map and this second pose prediction information can be determined according to actual needs and will not be specifically limited here. For example, the ray casting method can be used to project the part corresponding to the second pose prediction information in the existing environmental map (i.e., the target local environmental map), and simulate the environmental observation information of the robot at the pose corresponding to the second pose prediction information.

[0069] As a possible implementation manner, it can be determined whether the robot has moved to the target local area based on the obtained environmental information through Figure 7 the manner shown. Please refer to Figure 7 , Figure 7 which is Figure 6 a schematic flowchart of the sub-steps included in sub-step S223 in

[0070] Sub-step S2231: Calculate the information matching degree between the first environmental information and the second environmental information.

[0071] Sub-step S2232: When the information matching degree is not less than the preset matching degree, determine that the robot has moved to the target local area.

[0072] Sub-step S2233: When the information matching degree is less than the preset matching degree, determine that the robot has not moved to the target local area.

[0073] In this embodiment, the similarity degree between the first environmental information and the second environmental information can be calculated as the information matching degree of the two pieces of information. Then, it is determined whether the information matching degree is greater than a preset matching degree, and the specific value of the preset matching degree can be determined in combination with actual requirements. For example, it is set to 80%. If the information matching degree is greater than or equal to the preset matching degree, it is determined that the robot has moved to the target local area. If the information matching degree is less than the preset matching degree, it is determined that the robot has not moved to the target local area.

[0074] When it is determined that the robot has moved to the target local area, the first pose prediction information obtained by the positioning network based on the first environmental information obtained when the robot moves to the target local area can be used as the relocalization result of the robot. Optionally, when determining that the robot moves to the target local area through sub-steps S221 to S223, the second pose prediction information obtained in sub-step S221 can be used as the relocalization result of the robot. That is, assuming that the second pose prediction information is first obtained based on the first environmental information at time t, and then it is determined that the robot has moved to the target local area based on the second environmental information obtained from the second pose prediction information and the first environmental information at time t, finally, the second pose prediction information first obtained based on the first environmental information at time t can be used as the relocalization result of the robot.

[0075] If it is determined that the moving person has not moved to the target local area, the robot is controlled to move autonomously based on the first environmental information immediately obtained by the robot, and it is determined whether it has moved to the target local area until a relocalization result is obtained.

[0076] To execute the corresponding steps in the above embodiments and various possible ways, an implementation manner of a robot relocalization device 200 is given below. Optionally, the robot relocalization device 200 may adopt the Figure 1 device structure of the electronic device shown above. Further, please refer to Figure 8 , Figure 8 which is a block diagram of the robot relocalization device 200 provided in an embodiment of the present application. It should be noted that for the robot relocalization device 200 provided in this embodiment, its basic principle and the technical effects generated are the same as those in the above embodiments. For the sake of brief description, for the parts not mentioned in this embodiment, reference can be made to the corresponding content in the above embodiments. In this embodiment, the robot relocalization device 200 may include: an autonomous movement module 210 and a processing module 220.

[0077] The autonomous movement module 210 is configured to obtain movement information according to the first environmental information currently obtained by the robot in the case of robot positioning loss, and control the movement of the robot according to the movement information.

[0078] The processing module 220 is used to determine whether the robot has moved to the target local area, and when it is determined that the robot has moved to the target local area, use a pre-trained positioning network to obtain first posture prediction information based on first environmental information obtained when the robot moves to the target local area as the relocalization result of the robot, wherein the positioning network is trained based on the robot posture information and corresponding environmental information in the target local area in the target scene, and the target local area is a part of the target scene.

[0079] Optionally, the above modules can be stored in the form of software or firmware. Figure 1 The memory 110 shown in the figure may be fixed in the operating system (OS) of the electronic device 100 and may be Figure 1 Meanwhile, the data and program codes required for executing the above modules may be stored in the memory 110.

[0080] An embodiment of the present application further provides a readable storage medium having a computer program stored thereon, and the computer program implements the robot repositioning method when executed by a processor.

[0081] In summary, the embodiments of the present application provide a robot repositioning method, device, electronic device and readable storage medium. In the case of robot positioning loss, the robot movement is controlled according to the movement information obtained from the first environmental information currently obtained by the robot, and during the movement process, it is determined whether the robot has moved to the target local area. In the case of determining that the robot has moved to the target local area, the first position prediction information obtained by using the pre-trained positioning network based on the first environmental information obtained when the robot moves to the target local area is used as the repositioning result of the robot. The positioning network is trained based on the robot position information and corresponding environmental information in the target local area in the target scene, and the target local area is a part of the target scene. In this way, fast repositioning can be achieved in a large scene with less calculation.

[0082] In several embodiments provided in this application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions, and operations of devices, methods, and computer program products according to multiple embodiments of this application. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than that marked in the accompanying drawings. For example, two consecutive blocks can actually be executed substantially in parallel, and they can sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, as well as the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0083] In addition, in each embodiment of this application, the various functional modules can be integrated together to form an independent part, or each module can exist alone, or two or more modules can be integrated to form an independent part.

[0084] If the described functions are implemented in the form of software functional modules and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of this application. The aforementioned storage medium includes various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs.

[0085] The above are only optional embodiments of this application and are not used to limit this application. For those skilled in the art, this application can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this application shall be included in the protection scope of this application.

Claims

1. A robot relocalization method, characterized in that, The method includes: In the case of the robot losing its position, obtaining movement information based on the first environmental information currently obtained by the robot, and controlling the movement of the robot according to the movement information; Determining whether the robot has moved to the target local area, and in the case of determining that the robot has moved to the target local area, using the first pose prediction information obtained by the pre-trained positioning network based on the first environmental information obtained when the robot moves to the target local area as the repositioning result of the robot, where the positioning network is trained based on the robot pose information and the corresponding environmental information in the target local area of the target scene, and the target local area is a part of the target scene.

2. The method according to claim 1, wherein The determining whether the robot has moved to the target local area includes: During the movement of the robot based on the movement information, using the positioning network to obtain the second pose prediction information corresponding to the first environmental information based on the first environmental information currently obtained; Obtaining the corresponding second environmental information according to the target local environment map corresponding to the target local area and the second pose prediction information; Determining whether the robot has moved to the target local area according to the first environmental information and the second environmental information.

3. The method according to claim 2, characterized in that, The determining whether the robot has moved to the target local area according to the first environmental information and the second environmental information includes: Calculating the information matching degree between the first environmental information and the second environmental information; In the case where the information matching degree is not less than the preset matching degree, determining that the robot has moved to the target local area; In the case where the information matching degree is less than the preset matching degree, determining that the robot has not moved to the target local area.

4. The method according to any one of claims 1 to 3, characterized in that, The positioning network is obtained in the following manner: Controlling the robot to perform autonomous movement exploration, and obtaining the pose information and the corresponding environmental information of the robot during the autonomous movement exploration, where the area of the autonomous movement exploration is the target local area; Obtaining the positioning network through deep learning according to the pose information and the corresponding environmental information.

5. The method according to any one of claims 1 to 3, characterized in that, The obtaining movement information based on the first environmental information currently obtained by the robot in the case of the robot losing its position includes: Using an active path planner to obtain the movement information based on the first environmental information currently obtained, where the active path planner is an autonomous exploration model or a policy network obtained through training, and the autonomous exploration model is used to determine the movement information used by the robot during autonomous movement in the target scene by using SLAM technology and a preset exploration strategy.

6. The method according to claim 5, wherein In the case where the active path planner is a policy network, the obtaining movement information based on the first environmental information currently obtained by the active path planner includes: Obtaining different candidate actions and the probabilities corresponding to the candidate actions based on the policy network and the first environmental information; Determining the target action from different candidate actions by random sampling according to the obtained probabilities, where the action corresponding to the movement information is the target action.

7. The method according to claim 5, characterized in that, When the active path planner is a policy network, the policy network is obtained in the following manner: Control the robot to perform a relocalization experiment in the target scenario and obtain experiment information, where the experiment information includes a path and the reward feedback corresponding to each action in the path; Obtain the policy network through reinforcement learning according to the experiment information of each experiment.

8. A robot relocating device, characterized in that, The device includes: An autonomous movement module, configured to obtain movement information according to the first environment information currently obtained by the robot when the robot's positioning is lost, and control the robot to move according to the movement information; A processing module, configured to determine whether the robot has moved to a target local area, and when it is determined that the robot has moved to the target local area, use the first pose prediction information obtained by the pre-trained positioning network based on the first environment information obtained when the robot moves to the target local area as the relocalization result of the robot, where the positioning network is trained based on the robot pose information and the corresponding environment information in the target local area in the target scenario, and the target local area is a part of the target scenario.

9. An electronic device, characterized in that, It includes a processor and a memory, the memory stores machine-executable instructions that can be executed by the processor, and the processor can execute the machine-executable instructions to implement the robot relocalization method according to any one of claims 1-7.

10. A readable storage medium storing a computer program thereon, characterized in that, When the computer program is executed by the processor, it implements the robot relocalization method according to any one of claims 1-7.

Images