Quadruped cross-domain environment detection robot based on reinforcement learning and control method

Abstract

The invention provides a reinforcement learning-based four-footed cross-domain environment detection robot and a control method.The robot comprises a waterproof control cabin, a skeleton device and an integrated device, the waterproof control cabin is used for control and detection of the robot, and the skeleton device is used for fixing the waterproof control cabin and the integrated device; the integrated device is used for conversion of land, ocean and sky forms of the robot. The robot controls the skeleton device and the integrated device to work through the waterproof control cabin so as to complete detection tasks under sea, land and air and complex terrains. According to the invention, the form and the operation mode can be freely switched in different environments, and the detection requirements under various complex terrains such as land, sea and air can be met; the reinforcement learning algorithm enables the robot to perform self-optimization according to feedback of different terrains, and the adaptability and the operation efficiency of the robot are improved.

Description

Technical Field

[0001] The present invention relates to the field of environment detection robots, and in particular to a quadruped-like cross-domain environment detection robot based on reinforcement learning and a control method thereof. Background Art

[0002] Unmanned detection systems can use sensors to collect and analyze environmental data, and are widely used in disaster site detection and emergency rescue, environmental data collection, agriculture, construction, mountainous terrain, marine surveys, etc. With the continuous advancement of science and technology, especially the rapid development of unmanned systems and artificial intelligence, modern detection tasks gradually require unmanned detection systems to be able to move and detect efficiently and flexibly in a variety of complex environments.

[0003] At present, the unmanned detection system has the following main problems: first, most of the existing unmanned detection systems can only work under a single terrain. For example, land robots, marine robots and aerial robots usually need to be designed and deployed separately, which not only increases the cost, but also limits the efficiency and flexibility of task execution; second, in the face of extreme and changeable detection environments, the existing unmanned detection systems cannot adapt to the rapid changes in terrain and environment. For example, when faced with multiple detection path types such as flat, bumpy, and stepped, a land robot that only has the ability to move on a single road type has low reliability and efficiency in completing detection tasks; in addition, the detection of multi-terrain environments often requires multiple different types of equipment to work together. In the face of complex and changeable detection environments, the accuracy and real-time performance of control instructions are very high, and the coordination between equipment is complex and difficult to manage, which brings considerable challenges to practical applications. Therefore, developing a multi-terrain integrated unmanned detection system that can adapt to land, sea and air environments at the same time is a technical problem that needs to be solved in this field. Summary of the invention

[0004] The purpose of the present invention is to provide a reinforcement learning-based quadruped cross-domain environment detection robot and control method to solve the technical problems mentioned in the above background technology regarding adaptability, cost, adaptability and real-time response in multi-terrain detection.

[0005] In order to achieve the above object, the specific technical solution adopted by the present invention is:

[0006] On the one hand, the present invention provides a reinforcement learning-based quadruped cross-domain environment detection robot, the robot comprising: a waterproof control cabin, a skeleton device and an integrated device; the waterproof control cabin is fixed to the skeleton device, and the integrated device is respectively fixed to the left front side, the right front side, the left rear side, and the right rear side of the skeleton device;

[0007] The waterproof control cabin is used for controlling and detecting the robot, the skeleton device is used for fixing the waterproof control cabin and the integrated device, and the integrated device is used for converting the robot's land, sea and sky forms;

[0008] Optionally, the waterproof control cabin includes a controller, the controller includes a micro single-chip computer and a storage module, the micro single-chip computer includes a brain-like chip and a mobile controller, and the brain-like chip is preset with a robot control model;

[0009] Optionally, the robot control model can calculate and output the moving path, operation mode, speed and angle of each joint motor and servo of the detection robot in real time based on the environmental terrain, its own position, map information, task starting point, task target point, obstacle position, air pressure and temperature acquired or collected in real time, and the mobile controller controls the morphology and displacement parameters of the detection robot based on the above-mentioned moving path, operation mode, speed and angle of each joint motor and servo;

[0010] Optionally, different operating modes correspond to different forms, including flight mode, land mode, and sea mode;

[0011] Optionally, the integrated device includes a first joint, a second joint motor, a second joint, a third joint motor, a flight motor, a flight wing, a third joint, a mobile motor, and a composite tire; the folding servo is rotationally fixed to the first joint, the first joint is provided with a boss for fixing the second joint motor, the second joint motor is rotationally fixed to the second joint, the second joint is provided with a through hole for fixing the third joint motor, the second joint is provided with a bracket for fixing the flight motor, the flight motor is rotationally connected to the flight wing, the flight wing is a wing-shaped structure, and the second joint bracket is provided with a thin-walled boss for flight protection; the third joint motor is rotationally connected to one side of the third joint, the other side of the third joint is fixedly connected to the mobile motor, the composite tire is rotationally connected to the mobile motor, and the composite tire is provided with a rotor for underwater propulsion.

[0012] According to another aspect of the present invention, a control method for a quadruped-like cross-domain environment detection robot based on reinforcement learning is also provided, comprising the following steps:

[0013] S1: Real-time acquisition or collection of environmental terrain, own position, map information, mission starting point, mission target point, obstacle location, air pressure and temperature information;

[0014] S2: The robot control model calculates and outputs the movement path, operation mode, speed and angle of each joint motor and servo based on the above information;

[0015] S3: The mobile controller controls the shape and displacement parameters of the detection robot based on the above-mentioned moving path, operation mode, speed and angle of each joint motor and servo;

[0016] S4: The detection robot moves and detects in the sky, sea, and land by switching different forms, and collects and stores surrounding environmental information in real time;

[0017] Optionally, the operating mode includes flight mode, land mode, and ocean mode;

[0018] Optionally, the control process of the flight mode includes: the robot control model of the controller brain-like chip issues a flight mode command, and controls the moving motors and the flight motors on the left front side, the left rear side, the right front side and the right rear side of the skeletal device to stop rotating so that the composite tire and the flying wing stop moving, the controller controls the folding servo and the second joint motor to make the second joint parallel to the ground, and controls the third joint motor to make the third joint parallel to the side of the cabin; the movement of the flight mode includes moving forward, backward, turning left, turning right, ascending and descending;

[0019] Optionally, the land mode includes a land quadruped mode and a land high-speed mode, and the land quadruped mode includes a flat ground driving mode and a walking mode;

[0020] Optionally, the ocean mode includes a sea surface mode and an underwater mode.

[0021] The present invention provides a reinforcement learning-based quadruped-like cross-domain environment detection robot and a control method, comprising: a waterproof control cabin, a skeleton device, and an integrated device. The waterproof control cabin is used for controlling and detecting the robot, the skeleton device is used for fixing the waterproof control cabin and the integrated device, and the integrated device is used for converting the robot's land, sea, and sky forms. The robot controls the skeleton device and the integrated device through the waterproof control cabin to complete detection tasks on land, sea, air, and complex terrains.

[0022] The present invention has the following technical effects:

[0023] (1) The robot can transform between land, sea and air through an integrated device, enabling it to perform exploration tasks on a variety of complex terrains, including land, sea and air, thereby improving the flexibility of task execution and meeting different exploration needs;

[0024] (2) By analyzing the collected environmental information, the robot can accurately match and flexibly switch its working mode, such as four-legged on land, high-speed on land, on the sea surface, under the sea, etc., to achieve multiple functions such as walking on land, floating on the sea, and flying in the air, so as to better adapt to the current detection environment and improve its task execution efficiency and reliability;

[0025] (3) Integrating multiple functions into one device breaks the limitation of traditional single-function devices, avoids the need to design and deploy multiple types of detection robots separately, reduces the types and quantity of detection system equipment, saves resources and costs, and improves the overall reliability and economy of the system;

[0026] (4) Combined with reinforcement learning algorithms, the robot can optimize itself based on real-time feedback from different terrains, continuously improving its adaptability and operational efficiency in various environments, ensuring that it can respond intelligently in unknown or changing environments and perform detection tasks efficiently and accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0028] Figure 1 It is a schematic diagram of the flight mode of the reinforcement learning-based quadruped cross-domain environment detection robot described in the present invention;

[0029] Figure 2 It is a schematic diagram of the bottom structure of the reinforcement learning-based quadruped-like cross-domain environment detection robot of the present invention;

[0030] Figure 3 It is a structural schematic diagram of the waterproof control cabin of the reinforcement learning-based quadruped cross-domain environment detection robot of the present invention;

[0031] Figure 4 It is a schematic diagram of the structure of the skeleton device of the reinforcement learning-based quadruped cross-domain environment detection robot of the present invention;

[0032] Figure 5 It is a structural schematic diagram of an integrated device of a quadruped-like cross-domain environment detection robot based on reinforcement learning according to the present invention;

[0033] Figure 6 It is a schematic diagram of the terrestrial quadruped mode of the quadruped-like cross-domain environment detection robot based on reinforcement learning described in the present invention;

[0034] Figure 7 It is a schematic diagram of the high-speed land mode of the quadruped-like cross-domain environment detection robot based on reinforcement learning according to the present invention;

[0035] Figure 8 It is a schematic diagram of the sea surface mode of the reinforcement learning-based quadruped cross-domain environment detection robot of the present invention;

[0036] Fig. 9It is a schematic diagram of the underwater mode of the reinforcement learning-based quadruped-like cross-domain environment detection robot described in the present invention.

[0037] Description of reference numerals:

[0038] Waterproof control cabin 1, cabin head 101, waterproof ring 102, cabin body 103, sensor 104, gimbal second joint motor 105, gimbal first joint 106, visual sensor 107, gimbal first joint motor 108, gimbal base 109, gimbal second joint 110, gimbal third joint motor 111, gimbal third joint 112, controller 113, skeleton device 2, skeleton frame 201, floating device 202, folding servo 203, integrated device 3, first joint 301, second joint motor 302, second joint 303, third joint motor 304, flight motor 305, flight wing 306, third joint 307, moving motor 308, composite tire 309. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical scheme and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. The specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in each embodiment of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0040] In the following description, terms such as "inside", "outside", "up", "down", "left", "right", etc. that indicate directions or positional relationships are only used to facilitate the description of the embodiments and simplify the description, and do not indicate or imply that the referred device or element must have a specific direction, be constructed and operate in a specific direction. Therefore, it should not be understood as a limitation of the present invention.

[0041] In order to solve the technical problems mentioned in the background technology section, the specific technical solution adopted by the present invention is:

[0042] A quadruped-like cross-domain environment detection robot based on reinforcement learning, as shown in the attached manual. Figure 1-2 As shown, it comprises: a waterproof control cabin 1, a skeleton device 2 and an integrated device 3; the waterproof control cabin 1 is fixed to the skeleton device 2, and the integrated device 3 is respectively fixed to the left front side, the right front side, the left rear side and the right rear side of the skeleton device;

[0043] The waterproof control cabin 1 is used for controlling and detecting the robot, the skeleton device 2 is used for fixing the waterproof control cabin 1 and the integrated device 3, and the integrated device 3 is used for converting the robot's land, sea and sky forms.

[0044] As the instruction manual Figure 3-4As shown, the waterproof control cabin 1 includes a cabin head 101, a waterproof ring 102, a cabin body 103, a sensor 104, a pan / tilt second joint motor 105, a pan / tilt first joint 106, a visual sensor 107, a pan / tilt first joint motor 108, a pan / tilt base 109, a pan / tilt second joint 110, a pan / tilt third joint motor 111, a pan / tilt third joint 112, and a controller 113;

[0045] Specifically, the cabin head 101 is provided with a thin-walled boss for detection by the visual sensor 107, the cabin head 101 is provided with a through hole fixed to the cabin body 103, the waterproof ring 102 is fixed to the top of the cabin body 103 for waterproof sealing of the cabin head 101 and the cabin body 103, the top of the cabin body 103 is provided with a through hole fixedly connected to the cabin head 101 and the waterproof ring 102, the rear part of the cabin body 103 is provided with a waterproof signal hole for controlling the servo and motor signal lines, the cabin body 103 is a film groove type structure, the controller 113 is fixed inside the cabin body 103, the sensor 104 is fixed inside the cabin body 103, the gimbal base 109 is fixed inside the cabin body 103, the gimbal third joint motor 111 is fixed to the top of the gimbal base 109, and the bottom of the gimbal third joint 112 is connected to the gimbal third joint motor 11 1 is rotatably connected, the third joint 112 is provided with a boss, the second joint motor 105 of the pan-tilt head is fixed to the boss, the top of the boss is provided with a through hole for the rotation axis of the second joint motor 105 of the pan-tilt head to rotate, the second joint motor 105 of the pan-tilt head is rotatably connected to the second joint 110 of the pan-tilt head, bosses are provided on both sides of the second joint 110 of the pan-tilt head, one side of the boss is provided with a through hole for the first joint 106 of the pan-tilt head to be rotatably fixed, one side of the boss is rotatably connected to the first joint motor 108 of the pan-tilt head, the first joint motor 108 of the pan-tilt head is fixed to the first joint 106 of the pan-tilt head, the first joint 106 of the pan-tilt head is provided with a through hole for the first joint motor 108 of the pan-tilt head to be fixed, the first joint 106 of the pan-tilt head is provided with a groove for the first joint motor 108 of the pan-tilt head to be fixed, and the visual sensor 107 is fixed to the front side of the first joint 106 of the pan-tilt head;

[0046] The sensor 104 includes a posture sensor, an air pressure sensor, a satellite navigation sensor and a humidity sensor;

[0047] The controller 113 includes a power supply, a micro single chip computer, a communication module, and a storage module; the power supply can supply power to each functional module of the detection robot; the storage module can store data information collected during the operation of the detection robot;

[0048] The micro single-chip computer includes a brain-like chip and a mobile controller;

[0049] A robot control model is preset in the brain-like chip, and the robot control model is obtained by training historical sample data through machine learning. It can calculate and output the moving path, operation mode, speed and angle of each joint motor and servo of the detection robot in real time based on the environmental terrain, its own position, map information, task starting point, task target point, obstacle position, air pressure and temperature acquired or collected in real time. The mobile controller controls the shape and displacement parameters of the detection robot based on the above-mentioned moving path, operation mode, speed and angle of each joint motor and servo;

[0050] To further achieve accurate control of the robot, the controller periodically obtains the robot's operating data and iteratively trains the robot control model based on reinforcement learning to continuously improve its adaptability and operational efficiency in various environments, ensuring that it can make intelligent responses in unknown or changing environments and perform detection tasks efficiently and accurately.

[0051] Optionally, the controller 113 can also adjust and control the operation state of the detection robot by receiving control instructions issued by the user through a remote terminal based on the communication module; the remote terminal can be a PC, a mobile terminal, etc.;

[0052] As the instruction manual Figure 4 and 5 As shown, the skeleton device 2 includes a skeleton frame 201, a floating device 202, and a folding steering gear 203;

[0053] The skeleton frame 201 is a hollow structure, and is provided with through holes for connecting the cabin head 101 and the cabin body 103. The skeleton frame 201 is provided with bosses at the left front, right front, left rear and right rear for fixed connection with the folding steering gear 203. The floating device 202 is fixed to the bottom of the skeleton frame 201.

[0054] The integrated device 3 includes a first joint 301, a second joint motor 302, a second joint 303, a third joint motor 304, a flight motor 305, a flight wing 306, a third joint 307, a mobile motor 308, and a composite tire 309; the folding servo 203 is rotationally fixed to the first joint 301, the first joint 301 is provided with a boss for fixing the second joint motor 302, the second joint motor 302 is rotationally fixed to the second joint 303, the second joint 303 is provided with a through hole for fixing the third joint motor 304, the second joint 303 is provided with a bracket for fixing the flight motor 305, the flight motor 305 is rotationally connected to the flight wing 306, the flight wing is a wing-shaped structure, and the second joint 303 bracket is provided with a thin-walled boss for flight protection; the third joint motor 304 is rotationally connected to one side of the third joint 307, the other side of the third joint 307 is fixedly connected to the mobile motor 308, the composite tire 309 is rotationally connected to the mobile motor 308, and the composite tire 309 is provided with a rotor for underwater propulsion;

[0055] Optionally, the second joint 303 bracket is provided with a plurality of through holes for reducing the weight of the robot.

[0056] According to another aspect of the present invention, a control method for a quadruped-like cross-domain environment detection robot based on reinforcement learning is also provided, comprising the following steps:

[0057] S1: Real-time acquisition or collection of environmental terrain, own position, map information, mission starting point, mission target point, obstacle location, air pressure and temperature information;

[0058] S2: The robot control model calculates and outputs the movement path, operation mode, speed and angle of each joint motor and servo based on the above information;

[0059] S3: The mobile controller controls the shape and displacement parameters of the detection robot based on the above-mentioned moving path, operation mode, speed and angle of each joint motor and servo;

[0060] S4: The detection robot moves and detects in the sky, sea, and land by switching between different forms, and collects and stores surrounding environmental information in real time.

[0061] As the instruction manual Figure 6-9 As shown, the different operation modes correspond to different forms, including flight mode, land mode, and ocean mode;

[0062] The land mode includes a land quadruped mode and a land high-speed mode, wherein the land high-speed mode moves more smoothly and at a faster speed, while the land quadruped mode has a stronger obstacle avoidance capability and a slower speed;

[0063] The land quadruped mode includes a flat ground driving mode and a walking mode;

[0064] The ocean mode includes a sea surface mode and a sea bottom mode;

[0065] The following is a detailed introduction to the control process of different operating modes:

[0066] The control process of the flight mode is as follows: the robot control model of the brain-like chip of the controller 113 issues a flight mode command, and controls the mobile motors 308 and the flight motors 305 on the left front side, the left rear side, the right front side and the right rear side of the skeletal device 2 to stop rotating so that the composite tire 309 and the flying wing 306 stop moving, the controller 113 controls the folding servo 203 and the second joint motor 302 to make the second joint 303 parallel to the ground, and controls the third joint motor 304 to make the third joint 307 parallel to the side of the cabin 103;

[0067] The movements of the flight mode include forward, backward, left turn, right turn, ascent and descent; in the forward / backward control, the speed of the front / rear flight motor 305 of the skeletal device 2 is reduced, and the speed of the rear / front flight motor 305 is increased, so as to realize forward / backward control; in the left / right turn control, the speed of the left / right flight motor 305 of the skeletal device 2 is reduced, thereby driving the speed of the flight wing 306 to reduce, and the speed of the right / left flight motor 305 is increased, thereby driving the speed of the flight wing 306 to increase, so as to realize left / right turn control; in the ascent / descent control, the speed of the left front, left rear, right front and right rear flight motors 305 of the skeletal device 2 is increased / decreased, thereby driving the speed of the flight wing 306 to increase / decrease, so as to realize ascent / descent control;

[0068] The control process of the land high-speed mode is as follows: the robot control model of the brain-like chip of the controller 113 issues a land high-speed mode command, and controls the moving motors 308 and the flying motors 305 on the left front side, the left rear side, the right front side and the right rear side of the skeletal device 2 to stop rotating so that the composite tire 309 and the flying wing 306 stop moving, the controller 113 controls the left front side, the left rear side, the right front side and the right rear side folding servos 203 and the second joint motor 302 to make the second joint 303 fold upward and perpendicular to the ground, and controls the third joint motor 304 to adjust the third joint 307 to achieve the adjustment of the height of the mobile chassis;

[0069] The movement of the land high-speed mode includes forward, backward, left turn and right turn; for the forward / backward control, the left and right moving motors 308 of the skeleton device 2 rotate forward / reversely to drive the composite tire 309 to rotate forward / reversely to propel the robot forward / backward, thus realizing the forward / backward control; for the left / right turn control, the left moving motor 308 of the skeleton device 2 rotates forward / reversely to drive the composite tire 309 to move backward / forward, and the right moving motor 308 rotates forward / reversely to drive the composite tire 309 to move forward / backward, thus realizing the left / right turn control;

[0070] The control process of the land quadruped mode is as follows: the robot control model of the brain-like chip of the controller 113 sends a land quadruped mode command, and controls the moving motors 308 and the flying motors 305 on the left front side, left rear side, right front side and right rear side of the skeletal device 2 to stop rotating so that the composite tire 309 and the flying wing 306 stop moving; the controller 113 controls the left front side, left rear side, right front side and right rear side folding servos 203 to make the second joint 303 fold downward and be perpendicular to the ground; the controller 113 controls the left front side, left rear side, right front side and right rear side second joint motors 302 to make the second joint 303 rotate outward to a position parallel to the side of the cabin 103; the controller 113 controls the left front side, left rear side, right front side and right rear side third joint motors 304 to make the third joint 307 rotate downward, and realize the robot's quadruped standing;

[0071] The flat ground driving mode includes forward, backward, left turn and right turn; for the forward / backward control, the left and right moving motors 308 of the skeleton device 2 rotate forward / reversely to drive the composite tire 309 to rotate forward / reversely to propel the robot forward / backward, thus realizing the forward / backward control; for the left / right turn control, the left moving motor 308 of the skeleton device 2 rotates forward / reversely to drive the composite tire 309 to move backward / forward, and the right moving motor 308 rotates forward / reversely to drive the composite tire 309 to move forward / backward, thus realizing the left / right turn control;

[0072] The walking modes include forward, backward, left turn and right turn. The robot control model sends instructions to the controller 113 through real-time environmental data. The controller 113 controls the movement of the first joint 301, the second joint 303 and the third joint 307 by controlling the folding steering gear 203, the second joint motor 302 and the third joint motor 304, so as to realize the robot moving forward, backward, left turn and right turn.

[0073] The control process of the sea surface mode is as follows: the robot control model of the brain-like chip of the controller 113 issues a sea surface mode command, and controls the flight motors 305 on the left front side, the left rear side, the right front side and the right rear side of the skeleton device 2 to reduce the speed so as to reduce the speed of the flight wing 306, the robot lands on the sea surface, and floats on the sea surface through the floating device 202, the controller 113 controls the flight motor 305 to stop rotating, and then controls the folding servo 203 of the skeleton device 2 to make the second joint 303 fold downward and perpendicular to the sea surface;

[0074] The movement of the sea surface mode includes left turn, right turn, forward and backward; for the left turn / right turn control, the speed of the left / right flight motor 305 of the skeleton device 2 is reduced, thereby driving the speed of the flying wing 306 to reduce, and the speed of the right / left flight motor 305 is increased, thereby driving the speed of the flying wing 306 to increase, thereby realizing the left turn / right turn control; for the forward / backward control, the left / right flight motor 305 of the skeleton device 2 rotates forward / reverse, thereby driving the flying wing 306 to rotate forward / reverse to propel the robot forward / backward, thereby realizing the forward / backward control;

[0075] The control process of the underwater mode is as follows: the robot control model of the brain-like chip of the controller 113 issues an underwater mode command, and controls the flight motors 305 on the left front side, left rear side, right front side and right rear side of the skeleton device 2 to reduce the speed so as to reduce the speed of the flight wing 306, the robot lands on the sea surface, and floats on the sea surface through the floating device 202, the controller 113 controls the flight motor 305 to stop rotating, the controller 113 controls the flight motor 305 to stop rotating, controls the third joint motor 304 to make the third joint 307 parallel to the side of the cabin 103, and then controls the skeleton device 2 to control the rear folding servo 203 to make the second joint 303 fold downward and perpendicular to the sea surface, and the controller 113 controls the front folding servo 203 and the second joint motor 302 to make the second joint 303 parallel to the ground;

[0076] The movements of the underwater mode include ascending, descending, turning left, turning right, advancing and retreating; in the ascending / descending control, the flight motor 305 on the front side of the skeleton device 2 rotates forward / reversely, thereby driving the flight wing 306 to rotate forward / reversely, and the robot is pushed up / down to achieve ascending / descending control; in the left / right turn control, the speed of the flight motor 305 on the front side of the skeleton device 2 remains unchanged, the speed of the left rear side / right rear side flight motor 305 decreases, thereby driving the speed of the flight wing 306 to decrease, and the speed of the right rear side / left rear side flight motor 305 increases, thereby driving the speed of the flight wing 306 to increase, thereby achieving left / right turn control; in the forward / backward control, the speed of the flight motor 305 on the front side of the skeleton device 2 remains unchanged, and the left rear side and right rear side flight motors 305 rotate forward / reversely, thereby driving the flight wing 306 to rotate forward / reversely to propel the robot forward / backward, thereby achieving forward / backward control.

[0077] The present invention provides a reinforcement learning-based quadruped-like cross-domain environment detection robot and a control method, comprising: a waterproof control cabin, a skeleton device, and an integrated device. The waterproof control cabin is used for controlling and detecting the robot, the skeleton device is used for fixing the waterproof control cabin and the integrated device, and the integrated device is used for converting the robot's land, sea, and sky forms. The robot controls the skeleton device and the integrated device through the waterproof control cabin to complete detection tasks on land, sea, air, and complex terrains.

[0078] The present invention can effectively solve the limitations of existing technologies in multi-terrain detection by integrating the imitation of quadruped morphology and sea, land and air conversion capabilities, and combining reinforcement learning algorithms to improve the robot's autonomous adaptability. This integrated, multifunctional robot can not only simplify equipment deployment and save costs, but also improve the execution efficiency and adaptability of detection tasks, meet the real-time response requirements in complex terrain and unknown environments, and thus has broad application prospects.

[0079] The above contents are only the technical ideas of the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical ideas proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A quadruped-like cross-domain environment detection robot based on reinforcement learning, characterized in that: The robot comprises: a waterproof control cabin (1), a skeleton device (2) and an integrated device (3); the waterproof control cabin (1) is fixed to the skeleton device (2), and the integrated device (3) is respectively fixed to the left front side, the right front side, the left rear side and the right rear side of the skeleton device; The waterproof control cabin (1) is used for controlling and detecting the robot, the skeleton device (2) is used for fixing the waterproof control cabin (1) and the integrated device (3), and the integrated device (3) is used for converting the robot's land, sea and sky forms.

2. The reinforcement learning-based quadruped cross-domain environment detection robot according to claim 1 is characterized in that: The waterproof control cabin (1) comprises a controller (113), the controller (113) comprises a micro single-chip computer and a storage module, the micro single-chip computer comprises a brain-like chip and a mobile controller, and the brain-like chip is preset with a robot control model.

3. The reinforcement learning-based quadruped cross-domain environment detection robot according to claim 2 is characterized in that: The robot control model can calculate and output the movement path, operation mode, speed and angle of each joint motor and servo of the detection robot in real time based on the environmental terrain, its own position, map information, task starting point, task target point, obstacle position, air pressure and temperature acquired or collected in real time. The mobile controller controls the morphology and displacement parameters of the detection robot based on the above-mentioned movement path, operation mode, speed and angle of each joint motor and servo.

4. The reinforcement learning-based quadruped cross-domain environment detection robot according to claim 3 is characterized in that: Different operating modes correspond to different forms, and the operating modes include flight mode, land mode, and ocean mode.

5. The reinforcement learning-based quadruped cross-domain environment detection robot according to claim 2 is characterized in that: The integrated device (3) comprises a first joint (301), a second joint motor (302), a second joint (303), a third joint motor (304), a flying motor (305), a flying wing (306), a third joint (307), a moving motor (308), and a composite tire (309); the folding steering engine (203) is rotationally fixed to the first joint (301), the first joint (301) is provided with a boss for fixing the second joint motor (302), the second joint motor (302) is rotationally fixed to the second joint (303), and the second joint (303) is provided with a boss for fixing the third joint motor (304) is fixed with a through hole, the second joint (303) is provided with a bracket for fixing a flight motor (305), the flight motor (305) is rotationally connected to a flight wing (306), the flight wing is a wing-shaped structure, and the bracket of the second joint (303) is provided with a thin-walled boss for flight protection; the third joint motor (304) is rotationally connected to one side of the third joint (307), the other side of the third joint (307) is fixedly connected to a mobile motor (308), the composite tire (309) is rotationally connected to the mobile motor (308), and the composite tire (309) is provided with a rotor for underwater propulsion.

6. A control method for a quadrupedal cross-domain environment detection robot based on reinforcement learning applied to the quadrupedal cross-domain environment detection robot based on reinforcement learning as described in any one of claims 1 to 5, characterized in that: The following steps are included: S1: Real-time acquisition or collection of environmental terrain, own position, map information, mission starting point, mission target point, obstacle location, air pressure and temperature information; S2: The robot control model calculates and outputs the movement path, operation mode, speed and angle of each joint motor and servo based on the above information; S3: The mobile controller controls the shape and displacement parameters of the detection robot based on the above-mentioned moving path, operation mode, speed and angle of each joint motor and servo; S4: The detection robot moves and detects in the sky, sea, and land by switching between different forms, and collects and stores surrounding environmental information in real time.

7. The method according to claim 6, characterized in that The operation modes include flight mode, land mode and sea mode.

8. The method according to claim 7, characterized in that The control process of the flight mode includes: the robot control model of the brain-like chip of the controller (113) issues a flight mode instruction, and controls the moving motors (308) and the flight motors (305) on the left front side, the left rear side, the right front side and the right rear side of the skeletal device (2) to stop rotating, thereby stopping the composite tire (309) and the flying wing (306) from moving; the controller (113) controls the folding steering gear (203) and the second joint motor (302) to make the second joint (303) parallel to the ground, and controls the third joint motor (304) to make the third joint (307) parallel to the side of the cabin (103); the movement of the flight mode includes moving forward, backward, turning left, turning right, ascending and descending.

9. The method according to claim 7, characterized in that: The land mode includes a land quadruped mode and a land high-speed mode, and the land quadruped mode includes a flat ground driving mode and a walking mode.

10. The method according to claim 9, characterized in that The ocean mode includes a sea surface mode and an underwater mode.

Images