Training method and device for autonomous movement of lunar miniature robot in non-structural environment

Abstract

The invention discloses a method and a device for training autonomous movement of a lunar miniature robot in a non-structural environment, and belongs to the field of extraterrestrial exploration. The method comprises the following steps: constructing a three-dimensional simulation scene for autonomous movement training of the lunar miniature robot; performing domain randomization processing on key parameters of the robot in the three-dimensional simulation scene according to a preset threshold range, and training according to a processing result to generate an autonomous mobile intelligent model; constructing a non-structural physical scene for the autonomous movement test of the lunar miniature robot; and testing the autonomous mobile intelligent model in the non-structural physical environment, and adjusting key environment parameters to obtain a test result of the autonomous mobile intelligent model. According to the invention, the adaptability of the autonomous mobile intelligent model to the key unknown environment can be improved.

Description

Technical Field

[0001] The present invention relates to the field of extraterrestrial exploration technology, and in particular to a training method and device for autonomous movement of a lunar micro-robot in a non-structured environment. Background Art

[0002] Extraterrestrial exploration scenarios are unstructured environments with notable characteristics such as harsh and unknown environments, lack of prior knowledge, and poor communication conditions. In such environments, the demand for robot autonomy and intelligence is extremely urgent. Artificial intelligence algorithms are an effective way to improve the level of intelligent autonomy of extraterrestrial robots. Robot intelligence algorithms are usually trained in simulation scenarios and then migrated to physical scenarios for testing.

[0003] In the related technology, since the actual lunar working environment is unknown and difficult to accurately model, the intelligent model trained by the lunar micro-robot in the existing simulation scenario is difficult to be successfully applied in the unknown lunar environment.

[0004] Based on this, there is an urgent need for a training method and device for autonomous movement of lunar micro-robots in unstructured environments to solve the above technical problems. Summary of the invention

[0005] The present invention provides a method and device for training a lunar micro robot to move autonomously in an unstructured environment, which can enable the intelligent model trained in a simulation scene to better adapt to the lunar environment. The technical solution is as follows:

[0006] On the one hand, a training method for autonomous movement of a lunar micro robot in an unstructured environment is provided, the method comprising:

[0007] Construct a three-dimensional simulation scene for autonomous movement training of small lunar robots;

[0008] Performing domain randomization processing on key parameters of the robot in the three-dimensional simulation scene according to a preset threshold range, and training and generating an autonomous mobile intelligent model according to the processing results;

[0009] Construct a non-structural physical scene for testing the autonomous mobility of small lunar robots;

[0010] The autonomous mobile intelligent model is tested in the non-structural physical scene, and the key parameters are adjusted according to the test results to obtain the test results of the autonomous mobile intelligent model.

[0011] On the other hand, a training device for autonomous movement of a lunar micro-robot in an unstructured environment is provided, the device comprising:

[0012] The first modeling module is used to construct a three-dimensional simulation scene for autonomous movement training of lunar micro-robots;

[0013] A training module, used to perform domain randomization processing on key parameters of the robot in the three-dimensional simulation scene according to a preset threshold range, and train and generate an autonomous mobile intelligent model according to the processing results;

[0014] The second modeling module builds a non-structural physical scene for the autonomous mobility test of the lunar micro-robot;

[0015] The testing module is used to test the autonomous mobile intelligent model in the non-structural physical scene, and adjust the key parameters according to the test results to obtain the test results of the autonomous mobile intelligent model.

[0016] On the other hand, a computer device is provided, which includes a memory and a processor, wherein the memory is used to store computer programs, and the processor is used to execute the computer programs stored in the memory to implement the steps of the above-mentioned training method for autonomous movement of lunar micro-robots in an unstructured environment.

[0017] On the other hand, a computer-readable storage medium is provided, in which a computer program is stored. When the computer program is executed by a processor, the steps of the training method for autonomous movement of a lunar micro-robot in an unstructured environment are implemented.

[0018] On the other hand, a computer program product is provided, including a computer program, which, when executed by a processor, implements the steps of the above-mentioned training method for autonomous movement of a lunar micro-robot in an unstructured environment.

[0019] The technical solution provided by the present invention can at least bring the following beneficial effects: the method first constructs a three-dimensional simulation environment for autonomous mobile training of robots in a simulator, then performs domain randomization on the key parameters of the model according to the threshold range during the training process to obtain an autonomous mobile intelligent model, and trains the autonomous mobile intelligent model in the constructed physical test scene to adjust the key parameters to obtain the test results of the autonomous mobile intelligent model. This method constructs an unstructured environment for autonomous mobile training of lunar micro-robots, improves the adaptability of the intelligent model to key unknown environments, and constructs a ground physical test scene to test the intelligent model, verifying the feasibility of digital physical migration of the intelligent model, which is very important for applying the intelligent model in an unknown lunar environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 It is a flow chart of a training method for autonomous movement of a lunar micro robot in an unstructured environment provided by one embodiment of the present invention;

[0022] Figure 2 It is a structural diagram of a training device for autonomous movement of a lunar micro-robot in an unstructured environment provided by one embodiment of the present invention;

[0023] Figure 3 It is a hardware architecture diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0025] As mentioned earlier, for tiny robots on the lunar surface, their actual working environment is unknown and difficult to model accurately. Moreover, the environmental characteristics and physical properties of the simulated scene are usually difficult to be completely consistent with the physical scene. The application effect of the autonomous mobile intelligent model trained on this basis is often poor.

[0026] Based on this, the concept of the present invention is to adopt a parameter domain randomization method when training the robot's mobile intelligent model using a simulation scenario, which greatly improves the randomness of the training and ensures that the mobile intelligent model can effectively cope with the unknown lunar environment.

[0027] The specific implementation of the above concept is described below.

[0028] Please refer to Figure 1 , an embodiment of the present invention provides a method for training a lunar micro robot for autonomous movement in an unstructured environment, the method comprising:

[0029] Step 100, constructing a three-dimensional simulation scene for autonomous movement training of lunar micro-robots;

[0030] Step 102, performing domain randomization processing on key parameters of the robot in the three-dimensional simulation scene according to a preset threshold range, and training and generating an autonomous mobile intelligent model according to the processing results;

[0031] Step 104, constructing a non-structural physical scene for testing the autonomous movement of the lunar micro-robot;

[0032] Step 106, testing the autonomous mobile intelligent model in the non-structural physical scene, and adjusting the key parameters according to the test results to obtain the test results of the autonomous mobile intelligent model.

[0033] In the embodiment of the present invention, the method first constructs a three-dimensional simulation environment for autonomous mobile training of robots in a simulator, then performs domain randomization on key parameters of the model according to a threshold range during the training process to obtain an autonomous mobile intelligent model, and trains the autonomous mobile intelligent model in a constructed physical test scene to adjust key parameters to obtain test results of the autonomous mobile intelligent model. This method constructs an unstructured environment for autonomous mobile training of lunar micro-robots, improves the adaptability of the intelligent model to key unknown environmental parameters, and constructs a ground physical test scene to test the intelligent model, verifying the feasibility of digital-physical migration of the intelligent model, which is very important for applying the intelligent model in an unknown lunar environment.

[0034] Described below Figure 1 How the various steps are performed.

[0035] First, with respect to step 100, a three-dimensional simulation scene is constructed for autonomous movement training of a lunar micro-robot.

[0036] Due to the high training costs in physical scenarios, robot intelligence algorithms are usually trained in simulation scenarios and then migrated to physical scenarios for testing.

[0037] Therefore, this embodiment constructs a non-structural simulation scene for autonomous movement training of the lunar surface micro-robot in the simulator, and establishes a three-dimensional simulation scene including the lunar surface, deep space background, the landing assembly, and the lunar surface micro-robot, wherein the landing assembly and the lunar surface micro-robot are consistent with the real size, the landing assembly is fixed at a specified position on the lunar surface, the lunar surface micro-robot is a certain distance away from the center of the landing assembly, and the closest distance during the movement is not less than 2m, and the lunar surface micro-robot can simulate the movement of the real micro-robot and move according to the control instructions output by the intelligent algorithm. In this embodiment:

[0038] (1) Micro-robot camera simulation: The camera is fixed on the back of the small robot, with a field of view of 90°×65°, a height of 10 cm from the bottom of the small robot, and an angle of 20° between the line of sight and the horizontal direction;

[0039] (2) Micro-robot motion simulation: The motion of the small robot is simulated, and different motion actions such as forward, backward, left turn, and right turn are achieved through differential rotation control of the left and right drive motors;

[0040] (3) Geometric simulation of the landing assembly: A 1:1 scale model of the landing assembly is used to simulate the geometric shapes and relative positional relationships of the lander, ascender, flag, and sailboard. The positions and specific morphologies of other components are not of concern to this simulation environment. Texture materials are realized through mapping, and the color and reflectivity of the mapping are not of concern to this simulation environment.

[0041] (4) Simulation environment interface: The simulation environment includes external program interfaces such as C, C++, and Python. The simulator can be called through external programs to control various components in the simulation environment, including robot motion control, camera image acquisition, and simulation environment reset.

[0042] Then, for step 102, domain randomization processing is performed on key parameters of the robot in the three-dimensional simulation scene according to a preset threshold range, and an autonomous mobile intelligent model is trained and generated according to the processing results.

[0043] Since there are certain domain differences between the actual state of the lunar micro-robot in the unstructured environment of the lunar surface and the ground training environment, in order to improve the generalization ability of the intelligent model trained in this simulation environment, the key parameters are domain randomized in the simulation environment constructed in the above steps, so that the intelligent model trained in this simulation environment can autonomously adapt to the real unstructured environment of the lunar surface.

[0044] In this embodiment, the training process includes: determining the domain randomization threshold range of the parameters in the initial simulation scene according to the ideal working state of the micro-robot on the lunar surface and the movement ability of the micro-robot; initializing the key parameters of each training round within the threshold range, and performing multiple rounds of training on the robot according to the initialization results to obtain the autonomous mobile intelligent model.

[0045] Specifically, according to the ideal working state of the micro-robot on the moon and the movement ability of the micro-robot, it is assumed that the ideal working position of the micro-robot in the world coordinate system is (x d ,y d ,z d ), set the domain randomization range for the following parameters in the simulation environment:

[0046] Determine the value of the robot's target position (x goal ,y goal ,z goal ), where x goal The value range of x is: d +(x d_err_min ,xd_err_max ), y goal The value range of y is: d +(y d_err_min ,y d_err_max ), z goal The value range of z is: d +(z d_err_min ,z d_err_max ); err_min is the minimum error; err_max is the maximum error; that is, x d_err_min For x d The minimum error, x d_err_max For x d The maximum error of

[0047] Determine the robot's target posture Euler angle as Where, the roll angle φ goal is 0°, the pitch angle θ goal is 0°, yaw angle The random range is

[0048] According to the robot target position, determine the value of the robot's initial position (x init ,y init ,z init ), where x init The value range of x is: goal ±(x init_err_min ,x init_err_max );y init The value range of y is: goal ±(y init_err_min ,y init_err_max );z init The value range of z is: goal ±(z init_err_min ,z init_err_max );

[0049] Determine the robot's initial posture Euler angle as Where, the roll angle φ init The random range is (φ init_min ,φ init_max )°, pitch angle θ init The random range is (θ init_min ,θ init_max )°, yaw angle The random range is According to the normal working ability of the micro robot, the random range of the roll angle and pitch angle does not exceed ±15°;

[0050] The ideal arrival position (x desired ,y desired ,zdesired ), determine the value of the robot's actual arrival position (x real ,y real ,z real ), where x real The value range of x is: desired ±(0,x desired_err_max );y real The value range of y is: desired ±(0,y desired_err_max ) real The value range is: desired ±(0,z desired_err_max );

[0051] Euler angles of the ideal arrival posture according to the robot's execution of control instructions Determine the value of the Euler angle of the actual arrival posture Among them, φ real The value range of φ is: desired ±(0,φ desired_err_max );θ real The value range of θ is: desired ±(0,θ desired_err_max ); The value range of is:

[0052] For example, the target position of the lunar microrobot in the world coordinate system is: the x-axis position x goal The random range is 0+(-3,3)m, and the z-axis position is z goal The random range is 8+(-4,4)m, and the y-axis position is y goal 0+(-0.05, 0.05)m;

[0053] The target posture Euler angle of the lunar micro-robot in the world coordinate system is expressed as: in Randomize to (-30,30)°;

[0054] The initial position of the lunar microrobot in the world coordinate system: the x-axis position x init The random range is x goal ±(1,3)m, z-axis position z init The random range is z goal ±(1,3)m, y-axis position is y goal ±(-0.05, 0.05)m;

[0055] The target posture of the lunar microrobot in the world coordinate system is expressed in Euler angles as follows: init The random range is (-15,15)°, the pitch angle θ initThe random range is (-15,15)°, yaw angle The random range is (0,360)°;

[0056] The actual position reached by the lunar microrobot after executing the control command is: real =x desired ±(0,0.025)m,y real =y desired ±(0,0.025)m,z real =z desired ±(0,0.025)m;

[0057] The actual position reached by the lunar microrobot after executing the control command is: real =x desired ±(0,0.025)m,y real =y desired ±(0,0.025)m,z real =z desired ±(0,0.025)m.

[0058] With respect to step 104, a non-structural physical scene is constructed for testing the autonomous movement of the lunar micro-robot.

[0059] In the embodiment of the present invention, the ground non-structured physical scene includes the lunar surface, the black background of deep space, the landing assembly model, the lunar surface micro-robot, and the solar simulator.

[0060] Specifically, the scene simulation includes: the size of the constructed ground physical scene is 12m×4m×4m, and the distance between the tiny robot and the upper assembly is 2 to 10m;

[0061] Lunar surface terrain simulation: Use sand and rocks to simulate lunar soil and terrain. The thickness of sand at different locations is uneven, and the undulation of the sand does not exceed the range of motion of the lunar surface micro-robots; Lunar surface deep space background simulation: Use black light-absorbing cloth to simulate the black background of deep space, and arrange black light-absorbing cloth around the physical site; Upper assembly simulation: Make a 1:1 physical model of the upper assembly and place it fixedly in the site; Micro-robot simulation: The lunar surface micro-robot is placed on the sand, and its motion characteristics are consistent with the real micro-robots working on the lunar surface; Sun simulation: Use a solar simulator to simulate the sunlight on the lunar surface, and its illumination intensity and angle can be adjusted;

[0062] The solar simulator includes 2 movable LED lights, a fixed top lighting (ceiling light) and a large solar simulator.

[0063] With respect to step 106, the autonomous mobile intelligent model is tested in the non-structural physical scene, and the key parameters are adjusted according to the test results to obtain the test results of the autonomous mobile intelligent model.

[0064] In the embodiment of the present invention, the test process includes: initializing the environmental parameters of the non-structural physical scene, and testing the environmental adaptability of the autonomous mobile intelligent model according to the initialization result; wherein the environmental parameters include the thickness of the simulated lunar surface, the placement position of the robot, the simulated lighting conditions and the installation position of the components of the upper assembly; the environmental adaptability includes lighting adaptability, terrain adaptability, target image adaptability and initial image adaptability; determining whether the test result meets the preset end condition; if so, ending the test to obtain the test result of the autonomous mobile model, otherwise continuing the test under this working condition until the preset end condition is met; resetting the key parameters and testing again after the environmental parameters are updated until the test of all working conditions is completed. .

[0065] Specifically, autonomous movement tests of the lunar micro-robot were carried out in the physical scene constructed in the above steps. Before each test, the thickness of the sand was readjusted, and the lunar micro-robot was replaced at different locations on the sand in the test site. The position, posture and light intensity of the solar simulator were readjusted, and the installation positions of small components on the upper assembly were readjusted.

[0066] In this embodiment, the following test conditions are considered in the test scenario:

[0067] (1) Light adaptability:

[0068] The light source brightness includes the following working conditions: no light, only ceiling light, only one movable light, two movable lights, ceiling light + one movable light, ceiling light + two movable lights, and solar simulator.

[0069] The distance between the light source and the assembly includes the following working conditions: 5m from the assembly, 10m from the assembly;

[0070] The irradiation angle of the light source includes the following working conditions: facing the combination, irradiating on the left side of the combination, and irradiating on the right side of the combination.

[0071] (2) Terrain adaptability:

[0072] The thickness of the sand layer includes the following working conditions: thin sand (<5 mm), medium-thick sand (5 mm to 2 cm), and thick sand (>2 cm);

[0073] The terrain undulation includes the following working conditions: flat terrain (≤5°), gentle slope in front (≤15°), gentle slope in the back (≤15°), gentle slope on the left (≤15°), and gentle slope on the right (≤15°).

[0074] (3) Target image adaptability:

[0075] This includes the cases where the assembly is in the middle of the image and the flag is in the middle of the image.

[0076] (4) Initial image adaptability:

[0077] The distance from the target image shooting position includes the following working conditions: <1m, 1~2m, 2~4m;

[0078] The position of the assembly in the initial image includes the following conditions: the assembly is completely visible on the left side of the initial image, the assembly is completely visible in the middle of the initial image, the assembly is completely visible on the right side of the initial image, the assembly is partially visible in the initial image, and the assembly is not visible in the initial image.

[0079] In summary, when the real application scenarios on the lunar surface are unknown, we constructed diverse scenarios that are close to real lunar surface scenarios in simulation and ground environments for autonomous mobility training and testing of lunar micro-robots, thereby achieving efficient digital-physical migration of intelligent models. At the same time, domain randomization of parameters also improves the generalization ability of autonomous mobility intelligent models for unknown scenario environments.

[0080] Please refer to Figure 2 The embodiment of the present invention provides a training device for autonomous movement of a lunar micro-robot in an unstructured environment, the device comprising:

[0081] The first modeling module 200 is used to construct a three-dimensional simulation scene for autonomous movement training of a lunar micro-robot;

[0082] A training module 202 is used to perform domain randomization processing on key parameters of the robot in the three-dimensional simulation scene according to a preset threshold range, and train and generate an autonomous mobile intelligent model according to the processing results;

[0083] The second modeling module 204 constructs a non-structural physical scene for the autonomous movement test of the lunar micro robot;

[0084] The testing module 206 is used to test the autonomous mobile intelligent model in the non-structural physical scene, and adjust the key parameters according to the test results to obtain the test results of the autonomous mobile intelligent model.

[0085] In the embodiment of the present invention, the three-dimensional simulation scene and the non-structural physical scene constructed by the first modeling module 200 and the second modeling module 204 both include a simulated lunar surface, a deep space background, a landing assembly and a lunar surface micro-robot.

[0086] In an embodiment of the present invention, when the training module 202 performs domain randomization processing on the key parameters of the robot in the three-dimensional simulation scene according to a preset threshold range, and generates an autonomous mobile intelligent model through training according to the processing results, it is specifically used to perform the following operations: determine the domain randomization threshold range of the parameters in the initial simulation scene according to the ideal working state of the micro-robot on the lunar surface and the movement ability of the micro-robot; initialize and set the key parameters of each training within the threshold range, and perform multiple rounds of training on the robot according to the initialization results to obtain the autonomous mobile intelligent model.

[0087] In the embodiment of the present invention, the domain randomization threshold range of the parameters in the initial simulation scene is determined according to the ideal working state of the micro-robot on the lunar surface and the movement ability of the micro-robot, including:

[0088] According to the ideal working position of the robot (x d ,y d ,z d ), determine the value of the robot's target position (x goal ,y goal ,z goal ), where x goal The value range of x is: d +(x d_err_min ,x d_err_max ), y goal The value range of y is: d +(y d_err_min ,y d_err_max ), z goal The value range of z is: d +(z d_err_min ,z d_err_max ); err_min is the minimum error; err_max is the maximum error; determine the robot's target posture Euler angle as Where, the roll angle φ goal is 0°, the pitch angle θ goal is 0°, yaw angle The random range is According to the robot target position, determine the value of the robot's initial position (x init ,y init ,z init ), where x init The value range of x is: goal ±(x init_err_min ,x init_err_max );y init The value range of y is: goal ±(y init_err_min ,y init_err_max );z init The value range of z is:goal ±(z init_err_min ,z init_err_max ); Determine the robot's initial posture Euler angle as Where, the roll angle φ init The random range is (φ init_min ,φ init_max )°, pitch angle θ init The random range is (θ init_min ,θ init_max )°, yaw angle The random range is The ideal arrival position (x desired ,y desired ,z desired ), determine the value of the robot's actual arrival position (x real ,y real ,z real ), where x real The value range of x is: desired ±(0,x desired_err_max );y real The value range of y is: desired ±(0,y desired_err_max ) real The value range is: desired ±(0,z desired_err_max );Euler angle of the ideal arrival posture according to the robot's execution of control instructions Determine the value of the Euler angle of the actual arrival posture Among them, φ real The value range of φ is: desired ±(0,φ desired_err_max );θ real The value range of θ is: desired ±(0,θ desired_err_max ); The value range of is:

[0089] In the embodiment of the present invention, when the test module performs the test on the autonomous mobile intelligent model in the non-structural physical scene and adjusts the key parameters according to the test results to obtain the test results of the autonomous mobile intelligent model, it is specifically used to perform the following operations: initialize the environmental parameters of the non-structural physical scene, and test the environmental adaptability of the autonomous mobile intelligent model according to the initialization results; wherein the environmental parameters include the thickness of the simulated lunar surface, the placement position of the robot, the simulated lighting conditions and the installation position of the components of the assembly; determine whether the test results meet the preset end conditions;

[0090] If yes, the test is terminated to obtain the test result of the autonomous mobility model; otherwise, the test is continued under the working condition until the preset end condition is met; the key parameters are reset and the test is performed again after the environmental parameters are updated until the test of all working conditions is completed.

[0091] In the embodiment of the present invention, the environmental adaptability includes illumination adaptability, terrain adaptability, target image adaptability and initial image adaptability.

[0092] It should be noted that the training device for autonomous movement of a tiny robot on the lunar surface in an unstructured environment provided in the above embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, 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 training device for autonomous movement of a tiny robot on the lunar surface in an unstructured environment provided in the above embodiment and the training method embodiment for autonomous movement of a tiny robot on the lunar surface in an unstructured environment belong to the same concept. The specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0093] The embodiment of the present application also provides a computer device, please refer to Figure 3 The computer device includes a processor and a memory, in which at least one instruction, at least one program, a code set or an instruction set is stored, and the at least one instruction, at least one program, a code set or an instruction set is loaded and executed by the processor to implement the training method for autonomous movement of a lunar micro-robot in an unstructured environment provided by the above-mentioned method embodiments.

[0094] An embodiment of the present application also provides a computer-readable storage medium, on which is stored at least one instruction, at least one program, code set or instruction set, and the at least one instruction, at least one program, code set or instruction set is loaded and executed by a processor to implement the training method for autonomous movement of a lunar micro-robot in an unstructured environment provided by the above-mentioned method embodiments.

[0095] An embodiment of the present application also provides a computer program product, which includes a computer program. A processor of a computer device reads the computer program from a computer-readable storage medium, and the processor executes the computer program, so that the computer device executes the training method for autonomous movement of a lunar micro-robot in an unstructured environment as described in any of the above embodiments.

[0096] For the convenience of description, the above system or device is described by dividing it into various modules or units according to its functions. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0097] It can be known from the description of the above implementation methods that those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be essentially or partly contributed to the prior art in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present application or certain parts of the embodiments.

[0098] Finally, it should be noted that, in this article, relational terms such as first, second, third and fourth are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0099] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A training method for autonomous movement of a lunar micro robot in an unstructured environment, characterized in that: The method comprises: Construct a three-dimensional simulation scene for autonomous movement training of small lunar robots; Performing domain randomization processing on key parameters of the robot in the three-dimensional simulation scene according to a preset threshold range, and training and generating an autonomous mobile intelligent model according to the processing results; Construct a non-structural physical scene for testing the autonomous mobility of small lunar robots; The autonomous mobile intelligent model is tested in the non-structural physical scene, and the key parameters are adjusted according to the test results to obtain the test results of the autonomous mobile intelligent model.

2. The method according to claim 1, characterized in that The three-dimensional simulation scene and the non-structural physical scene both include a simulated lunar surface, a deep space background, a landing assembly and a lunar surface micro-robot.

3. The method according to claim 1, characterized in that The process of performing domain randomization processing on key parameters of the robot in the three-dimensional simulation scene according to a preset threshold range, and training and generating an autonomous mobile intelligent model according to the processing results, includes: According to the ideal working state of the micro-robot on the lunar surface and the movement ability of the micro-robot, the domain randomization threshold range of the parameters in the initial simulation scene is determined; The key parameters of each training session are initialized and set within the threshold range, and the robot is trained for multiple rounds according to the initialization results to obtain the autonomous mobile intelligent model.

4. The method according to claim 3, characterized in that Determining the domain randomization threshold range of the parameters in the initial simulation scene according to the ideal working state of the micro-robot on the lunar surface and the movement ability of the micro-robot includes: According to the ideal working position of the robot (x d ,y d ,z d ), determine the value of the robot's target position (x goal ,y goal ,z goal ), where x goal The value range of x is: d +(x d_err_min ,x d_err_max ), y goal The value range of y is: d +(y d_err_min ,y d_err_max ), z goal The value range of z is: d +(z d_err_min ,z d_err_max ); err_min is the minimum error; err_max is the maximum error; Determine the robot's target posture Euler angle as Where, the roll angle φ goal is 0°, the pitch angle θ goal is 0°, yaw angle The random range is According to the robot target position, determine the value of the robot's initial position (x init ,y init ,z init ), where x init The value range of x is: goal ±(x init_err_min ,x init_err_max );y init The value range of y is: goal ±(y init_err_min ,y init_err_max );z init The value range of z is: goal ±(z init_err_min ,z init_err_max ); Determine the robot's initial posture Euler angle as Where, the roll angle φ init The random range is (φ init_min ,φ init_max )°, pitch angle θ init The random range is (θ init_min ,θ init_max )°, yaw angle The random range is The ideal arrival position (x desired ,y desired ,z desired ), determine the value of the robot's actual arrival position (x real ,y real ,z real ), where x real The value range of x is: desired ±(0,x desired_err_max );y real The value range of y is: desired ±(0,y desired_err_max ) real The value range is: desired ±(0,z desired_err_max ); Euler angles of the ideal arrival posture according to the robot's execution of control instructions Determine the value of the Euler angle of the actual arrival posture Among them, φ real The value range of φ is: desired ±(0,φ desired_err_max );θ real The value range of θ is: desired ±(0,θ desired_err_max ); The value range of is:

5. The method according to claim 1, characterized in that The method of testing the autonomous mobile intelligent model by using the non-structural physical scene and adjusting the key parameters according to the test results to obtain the test results of the autonomous mobile intelligent model includes: Initializing the environmental parameters of the non-structural physical scene, and testing the environmental adaptability of the autonomous mobile intelligent model according to the initialization results; wherein the environmental parameters include the thickness of the simulated lunar soil, the placement position of the robot, the simulated lighting conditions, and the installation positions of the components of the landing assembly; Determine whether the test result meets the preset end condition; if so, end the test to obtain the test result of the autonomous mobility model; otherwise, continue the test under this working condition until the preset end condition is met; The key parameters are reset and the test is repeated after the environmental parameters are updated until the test of all working conditions is completed.

6. The method according to claim 5, characterized in that The environmental adaptability includes illumination adaptability, terrain adaptability, target image adaptability and initial image adaptability.

7. A training device for autonomous movement of lunar microrobots in an unstructured environment, characterized in that: The device comprises: The first modeling module is used to construct a three-dimensional simulation scene for autonomous movement training of lunar micro-robots; A training module, used to perform domain randomization processing on key parameters of the robot in the three-dimensional simulation scene according to a preset threshold range, and train and generate an autonomous mobile intelligent model according to the processing results; The second modeling module builds a non-structural physical scene for the autonomous mobility test of the lunar micro-robot; The testing module is used to test the autonomous mobile intelligent model in the non-structural physical scene and adjust the key environmental parameters to obtain the test results of the autonomous mobile intelligent model.

8. A computer device, characterized in that: The computer device includes a memory and a processor, the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to implement the steps of any one of the methods described in claims 1-6.

9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method described in any one of claims 1 to 6 are implemented.

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

Images