A deformable small quadrotor quadruped flying and climbing robot and a control method thereof

Abstract

This invention discloses a deformable miniature quadrotor quadrupedal climbing robot and its control method. The climbing robot includes a fuselage, four rotor assemblies, and four leg assemblies. The fuselage includes a mounting base, a climbing control module, an adjustment module, and a battery. The adjustment module includes a power module and an electronic speed controller (ESC). The leg assemblies achieve at least two degrees of freedom of rotation and include at least two leg segments, joint drive servos, and a wheel-leg switching assembly that can switch between feet or wheels. Each leg assembly is distributed on the mounting base and is driven by a corresponding joint drive servo to achieve horizontal oscillation. The rotor assemblies are connected one-to-one with the leg assemblies and include propellers and motors. The propellers are driven to rotate by the motors, which are electrically connected to the ESCs. The robot can autonomously switch between air and ground modes, generating motion modes in different environments, improving the stability and diversity of air-to-ground movement, and exhibiting good maneuverability and adaptability in complex environments, meeting the needs of miniaturized reconnaissance.

Description

Technical Field

[0001] This invention belongs to the field of robotics technology, specifically relating to a deformable small quadrotor quadrupedal climbing robot and its control method. Background Technology

[0002] Miniature robots represent a high-precision, multidisciplinary technology. Their small size, light weight, and high mobility demonstrate significant potential advantages and application value in various modern application fields, making them a cutting-edge scientific topic in advanced countries. Currently, research on miniature robots is still in its early stages, especially the research and design of centimeter-level robots. There is still a significant gap between these areas and true miniaturization and intelligence. Strict limitations on the robot's own mass, size, and actuator power present considerable challenges. As high-risk land and air scenarios become increasingly complex, relying solely on a single movement pattern is insufficient to meet the operational requirements of real-world tasks. Inspired by nature, insects, as highly mobile and adaptable animals, mostly possess both flight and crawling locomotion. Imitating their external structure and movement characteristics is an effective method to improve the robot's ability to coordinate aerial and ground movements.

[0003] In the prior art, the paper Zhang R., Wu Y., Zhang L., Xu C., and Gao F., Autonomous and Adaptive Navigation for Terrestrial-Aerial Bimodal Vehicles, IEEE Robotics and Automation Letters, 2022, 7(2):3008-3015, proposes a quadcopter-based wheeled robot, which includes two passive wheels and a tiltable quadcopter, and can autonomously switch between flight and rolling motion modes through motion planning and control algorithms. Chinese Patent No. CN108502044B discloses a combined and separable rotor and legged mobile operation robot, including a multi-rotor flight mechanism, a multi-legged walking operation mechanism, and a combined and separated mechanism for combining and separating the multi-rotor flight mechanism and the multi-legged walking operation mechanism. The combined separation mechanism includes an upper connecting module and a lower connecting module, which are respectively installed at the bottom of the multi-rotor flight mechanism and the top of the multi-legged walking operation mechanism. Through the locking cooperation between the two, it can realize aerial flight, support surface flying and climbing, land climbing and walking, and corresponding operation functions. It can also complete multi-mode air-ground collaborative operations through the combination and separation of the robot mechanism.

[0004] All the flying and climbing robots mentioned above have amphibious mobility, but they are modular structures. Modular flying and climbing robots use simple modular configurations, but their structures are relatively complex, heavy, large, have low integration, and high air resistance, which greatly reduces their flight maneuverability. Summary of the Invention

[0005] The purpose of this invention is to address the above-mentioned problems by proposing a deformable small quadrotor quadrupedal climbing robot and its control method. The robot can adaptively change its motion mode according to environmental conditions and has the advantages of being small and lightweight, having low air resistance, good concealment, good obstacle crossing performance, and long endurance. It has good mobility and adaptability in complex land and air operation environments and can meet the needs of miniaturized reconnaissance and operation in complex land and air environments.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] This invention proposes a deformable, small quadrotor, quadrupedal climbing robot, comprising a fuselage, four rotor assemblies, and four leg assemblies, wherein:

[0008] The fuselage includes a mounting base, and a flight control module, an adjustment module, and a battery, all mounted on the mounting base. The adjustment module includes a power module and an ESC. The battery, power module, flight control module, and ESC are electrically connected in sequence.

[0009] The leg assembly, used to achieve rotation with at least two degrees of freedom, includes at least two leg segments, at least two joint drive servos, and a wheel-leg switching assembly. Each joint drive servo is connected to a leg segment in a one-to-one correspondence and is electrically connected to the fly-climb control module. Each leg segment is connected in sequence and driven to rotate by the corresponding joint drive servo. The wheel-leg switching assembly includes a roller and a foot. The foot is rotatably connected to the rotation axis of the roller. By rotating the foot, the end of the leg assembly is switched to either a foot or a roller. Each leg assembly is rectangularly distributed on the mounting base and is driven by the corresponding joint drive servo to achieve horizontal swinging around the mounting base.

[0010] The rotor assembly includes a propeller and a motor. The propeller is driven to rotate by the motor. The motor corresponds one-to-one with the leg assembly and is fixed to any leg section of the leg assembly except for the leg section connected to the mounting base. The motor is also connected to the electronic speed controller. In the in-flight mode, the propeller is in a horizontal state, and the leg assembly switches between four flight motion modes: "T", "H", "O", and "X".

[0011] Preferably, the mounting base includes multiple connecting rods and a first support plate, a second support plate, and a third support plate arranged sequentially and parallel to each other. The first support plate and the second support plate are connected. The two ends of the connecting rods are respectively connected to the second support plate and the third support plate. The climbing control module is mounted on the first support plate, the adjustment module is mounted on the second support plate, and the battery is located between the second support plate and the third support plate.

[0012] Preferably, the rotor assembly further includes a limiting support plate connected to the motor, used to limit the rotation of the leg section connected to the rotor assembly.

[0013] Preferably, there are two leg segments, including a thigh segment and a lower leg segment, and two joint drive servos, including a hip joint drive servo and a knee joint drive servo. The hip joint drive servo is connected to the mounting base and is used to drive the thigh segment to swing horizontally around the mounting base. The knee joint drive servo is used to drive the lower leg segment to swing up and down around the thigh segment. The wheel-leg switching assembly is located at the end of the lower leg segment.

[0014] Preferably, the horizontal swing range of the femoral segment is 180°, and the vertical swing range of the calf segment is 90°.

[0015] Preferably, the propellers are staggered along the height direction in a horizontal state.

[0016] Preferably, the propeller is also equipped with a cylindrical protective cover, and the propeller propulsion enables the cylindrical protective cover to move in a wheel-like or adsorption-like manner.

[0017] Preferably, the end of the foot is provided with a claw or adhesive mechanism.

[0018] Preferably, the deformable small quadrotor quadrupedal climbing robot is equipped with at least one of a visual sensor, an auditory sensor, an olfactory sensor, and a tactile sensor, and each sensor is electrically connected to the climbing control module.

[0019] A control method for a deformable small quadrotor quadrupedal climbing robot includes the following steps:

[0020] S1. Construct the quadrotor dynamics model and the quadruped dynamics model of the climbing robot respectively;

[0021] S2. Determine the propeller lift coefficient K of the quadrotor dynamics model under different flight motion modes using an online system identification method. T ;

[0022] S3. Determine the flight motion model for the in-flight flight mode. The flight motion model includes a static flight motion model and a dynamic flight motion model. The static flight motion model uses a "feedforward + PD feedback" control method to realize the "X" type flight motion mode of the quadrotor dynamics model. The dynamic flight motion model uses a lift compensation algorithm to realize the "T", "H", or "O" type flight motion modes of the quadrotor dynamics model.

[0023] The "feedforward + PD feedback" control method satisfies the following formula:

[0024]

[0025] In the formula, u(t) is the input of the quadrotor dynamics model, x(t) is the state variable of the quadrotor dynamics model, and e(t) is the error of the quadrotor dynamics model. K is a constant coefficient related to the quadrotor dynamics model. p K is the proportional coefficient of the controller. d The differential coefficients of the controller, For feedforward term, For PD feedback items;

[0026] The lift compensation algorithm satisfies the following formula:

[0027]

[0028]

[0029] T=k(φ)w 2

[0030] In the formula, T is the lift of a single propeller, φ is the shielding angle after propeller deformation, ω is the angular velocity of the propeller, and k(φ) is the shielding angle related function. For the compensated desired motor speed, k is a scaling factor that depends on the occlusion angle correlation function k(φ). T It is the lift coefficient and is equal to k (φ=0);

[0031] S4. A two-layer CPG network and a nearest-neighbor coupling model are used to control the foot trajectory of the quadrupedal dynamics model. The two-layer CPG network includes eight oscillators and is divided into a consciousness layer for controlling the hip joint and a behavior layer for controlling the knee joint. The nearest-neighbor coupling model is as follows:

[0032]

[0033] Wherein, rotation matrix Defined as:

[0034]

[0035] In the formula, i = 1, 2, 3, 4, j = 1, 2, m = 1, 2, 3, 4, n = 1, 2, x mn and y mn Let k' represent the oscillator state variable of the nth joint of the mth leg component, where k' is the coupling strength and equals 0.1. This represents the phase difference between the mn-th oscillator and the ij-th oscillator;

[0036] S5. Train a two-layer CPG network based on the DDPG reinforcement learning algorithm to obtain the parameters of the trained behavior layer model. The reward function r of DDPG reinforcement learning is... t The formula is as follows:

[0037]

[0038] Among them, K v K is a positive velocity weighting coefficient. e A positive torque weighting coefficient, v x Let τ be the horizontal velocity and τ be the joint torque. Joint angular velocity;

[0039] S6. The flying and climbing robot locates itself and builds a map based on real-time collected environmental information to plan its path. It then adaptively selects a flight motion model or a trained two-layer CPG network to control the flying and climbing robot to achieve aerial flight or ground crawling motion.

[0040] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0041] 1) This robot combines an aerial rotor assembly and a ground-based leg assembly, integrating the rotor assembly directly onto the leg assembly to form a single-piece structure, creating a quadcopter with four rotatable leg segments. This enables the robot to perform land-air collaborative operations (such as walking, flying, climbing, manipulating, and transporting objects). It combines the mobility of a flying robot with the terrain adaptability of a legged robot, enabling it to achieve large-scale, long-distance global observation in the air and precise positioning in a small-scale, close-range environment on the ground. This solves the problems of poor stealth of flying robots and slow movement speed of crawling robots, improving the robot's mobility and adaptability in complex land-air operating environments. It also achieves lightweight, miniaturization, integration, and intelligence, with low air resistance, good obstacle-crossing performance, and long endurance, meeting the needs of miniaturized reconnaissance and operations in complex land-air environments.

[0042] 2) The deformable leg assembly, in conjunction with the rotor assembly, generates different motion modes during flight, such as “T”, “H”, “O”, and “X”. It can adaptively change the corresponding flight motion mode according to environmental conditions, enabling special operation tasks such as traversing narrow and restricted areas and observing targets at very close range. It also has the functions of grabbing and vertical wall climbing.

[0043] 3) Compared with existing amphibious robots with a single wheel or a single leg, this device is equipped with a wheel-leg switching component. By switching the end of the leg component to either a wheel or a foot, it has efficient and flexible movement speed and ground obstacle crossing ability.

[0044] 4) By establishing flight motion models under different motion modes in the air flight mode and using a two-layer CPG network trained based on the DDPG reinforcement learning algorithm, the flying and climbing robot can achieve autonomous switching between air and ground, generating motion modes in different environments, improving the stability and diversity of air and ground motion, and realizing intelligent control. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the structure of the climbing robot of the present invention;

[0046] Figure 2 This is a schematic diagram of the fuselage structure of the present invention;

[0047] Figure 3 This is a schematic diagram of the structure of the leg assembly of the present invention when it is switched to a roller;

[0048] Figure 4 This is a schematic diagram of the structure of the leg component of the present invention when it is switched to the foot.

[0049] Figure 5 This is a schematic diagram of the structure of the climbing robot of the present invention when it uses rollers for sliding;

[0050] Figure 6 This is a schematic diagram of the "H"-shaped structure of the climbing robot of the present invention in aerial flight mode;

[0051] Figure 7 This is a schematic diagram of the "O"-shaped structure of the climbing robot of the present invention in aerial flight mode;

[0052] Figure 8 This is a schematic diagram of the ground crawling mode structure of the flying robot of the present invention;

[0053] Figure 9 This is a schematic diagram of the dynamic deformation and crossing of the climbing robot of the present invention;

[0054] Figure 10 This is a flowchart of the control method for the deformable small quadrotor quadrupedal climbing robot of the present invention.

[0055] Explanation of reference numerals in the attached drawings: 1. Fuselage; 2. Rotor assembly; 3. Leg assembly; 11. First support plate; 12. Second support plate; 13. Third support plate; 14. Flight control module; 15. Adjustment module; 16. Battery; 17. Connecting rod; 121. Mounting base; 21. Propeller; 22. Motor; 23. Limiting support plate; 31. Hip joint drive servo; 32. Thigh joint; 33. Knee joint drive servo; 34. Lower leg joint; 35. Wheel-leg switching assembly; 351. Roller; 352. Foot. Detailed Implementation

[0056] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0057] It should be noted that when a component is referred to as being "connected" to another component, it can be directly connected to the other component or there may be an intervening component. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein in the specification of this application is for the purpose of describing particular embodiments only and is not intended to limit the application.

[0058] Example 1:

[0059] like Figure 1-9 As shown, a deformable small quadrotor quadrupedal climbing robot includes a body 1, four rotor assemblies 2, and four leg assemblies 3, wherein:

[0060] The fuselage 1 includes a mounting base, and a flight control module 14, an adjustment module 15 and a battery 16, all mounted on the mounting base. The adjustment module 15 includes a power module and an ESC. The battery 16, the power module, the flight control module 14 and the ESC are electrically connected in sequence.

[0061] The leg assembly 3 is used to realize rotation with at least two degrees of freedom. It includes at least two leg segments, at least two joint drive servos, and a wheel-leg switching assembly 35. Each joint drive servo is connected to a leg segment and electrically connected to the fly-climb control module. Each leg segment is connected in sequence and driven to rotate by the corresponding joint drive servo. The wheel-leg switching assembly 35 includes a roller 351 and a foot 352. The foot 352 is rotatably connected to the rotation axis of the roller 351. By rotating the foot 352, the end of the leg assembly 3 is switched to either the foot 352 or the roller 351. Each leg assembly 3 is rectangularly distributed on the mounting base and is driven by the corresponding joint drive servo to realize horizontal swinging around the mounting base.

[0062] The rotor assembly 2 includes a propeller 21 and a motor 22. The propeller 21 is driven to rotate by the motor 22. The motor 22 corresponds one-to-one with the leg assembly 3 and is fixed to any leg section of the leg assembly 3 except for the leg section connected to the mounting base. The motor 22 is also connected to the electronic speed controller. In the flight mode, the propeller 21 is in a horizontal state, and the leg assembly 3 switches between four flight motion modes: "T", "H", "O" and "X".

[0063] To achieve lightweight design, each leg section can utilize a hollow frame. For example, the leg section connected to roller 351 can employ an integrated double-layer hollow arm. The rotation axis of roller 351 is located in the gap between the double-layer hollow arms and is rotatably connected to them. Roller 351 is a driven wheel. Foot 352 is rotatably connected to the rotation axis of roller 351. Alternatively, the rotation axis of foot 352 can be coaxially connected to the rotation axis of roller 351. By rotating foot 352, the end of leg assembly 3 can be switched between foot 352 and roller 351. Foot 352 can be of any shape and can be manually switched by engaging with the double-layer hollow arm for limiting movement during use or retrieval. An electronic speed controller (ESC) is used to control the motor to complete the specified speed and movement.

[0064] Without increasing the complexity of the mechanical structure or reducing the controllability of the platform, this robot adopts a lighter and more integrated unibody structure: the leg components feature a hollow design while ensuring rigidity, and each leg incorporates multiple active joints (driven by servo motors) and one passive joint (wheel-leg switching component) with degrees of freedom; the rotor components are directly integrated into the leg components to form a unibody structure, constituting a quadcopter with four rotatable leg joints. The four leg components are symmetrically distributed along one diagonal of the fuselage. Rotor structures are added to the active joints to achieve dynamically deformable flight motion; auxiliary wheel structures are added to the passive joints to achieve rapid movement on smooth surfaces. Therefore, this flying and climbing robot can autonomously adjust to switch between different land and air environments, demonstrating flexible maneuverability and environmental adaptability.

[0065] By dynamically deforming in the air to generate different flight motion modes (such as "T", "H", "O", "X" etc.), the "T" motion mode can get closer to the target, the "H" motion mode can realize the grasping function, and the "four legs to four claws" grasping function can be extended to realize the grasping function. It can be applied to different task scenarios (such as walking, flying in narrow space, climbing vertical walls, ultra-close-range reconnaissance, object grasping and transportation, etc.), which greatly improves the robot's environmental adaptability and overcomes the disadvantages of poor stealth of single flying robots and slow movement speed of single crawling robots. The movement is smooth and flexible and has low energy consumption.

[0066] On land, when the robot's propeller angle is adjusted to horizontal, the robot can activate its flight mode. It achieves flight through the coordinated operation of its four rotor components. In the air, it can freely choose between non-deformation and dynamic deformation. Dynamic deformation produces "T," "H," "O," and "X" shapes to adapt to different environments. Figure 9 As shown: a represents an "X" shape, b represents an "O" shape, c represents an "H" shape, and d represents a "T" shape. The gray boxes represent obstacles, and a to d sequentially illustrate the deformation states of the flying robot during dynamic traversal. In this embodiment, the robot's takeoff weight is less than 800g, and its overall deformation size can be reduced by 10cm in the circumferential direction. In the air, the robot can freely descend vertically and quickly adjust to ground crawling mode, achieving crawling through the cooperation of its four leg components. Since the rollers 351 are passive wheels, ground crawling is achieved by rotating the feet 352 to switch the ends of the leg components 3 to feet 352. When crawling vertically, it can dynamically deform into a "T" shape and move under the drive of the four rotor components. Therefore, this flying robot can autonomously switch between complex land and air environments, dynamically deform and traverse narrow and restricted environments, and its insect-like leg structure enables obstacle crossing on rugged terrain. It should be noted that the robot can also achieve water movement by adjusting the propellers 21 to a vertical position through a waterproof design.

[0067] It can meet the needs of miniaturized reconnaissance, adapt to the operational needs of complex land and air environments, and perform a variety of security tasks such as rescue and information gathering in complex and high-risk environments (such as jungles, mountains, urban building complexes, and remote islands). It also has potential applications in agricultural protection, forest fire prevention, tunnel inspection and geological exploration, high-altitude building and equipment inspection, post-disaster search and rescue, reconnaissance, positioning, tracking, patrol, air-to-ground mobile signal base stations, operations in confined spaces such as tunnels / pipelines / caves, and hazardous gas detection.

[0068] In one embodiment, the mounting base includes multiple connecting rods 17 and a first support plate 11, a second support plate 12, and a third support plate 13 arranged sequentially and parallel to each other. The first support plate 11 and the second support plate 12 are connected, and the two ends of the connecting rods 17 are respectively connected to the second support plate 12 and the third support plate 13. A climbing control module 14 is mounted on the first support plate 11, an adjustment module 15 passes through the second support plate 12, and a battery 16 is located between the second support plate 12 and the third support plate 13. It should be noted that the mounting base can also have any structural shape.

[0069] In one embodiment, the rotor assembly 2 further includes a limiting support plate 23 connected to the motor 22, used to limit the rotation of the leg section connected to the rotor assembly 2. The limiting support plate 23 is connected between the corresponding motor 22 and the leg section, and protects the rotor assembly 2 from damage through the limiting function.

[0070] In one embodiment, there are two leg segments, including a thigh segment 32 and a lower leg segment 34. There are also two joint drive servos, including a hip joint drive servo 31 and a knee joint drive servo 33. The hip joint drive servo 31 is connected to the mounting base and drives the thigh segment 32 to swing horizontally around the mounting base. The knee joint drive servo 33 drives the lower leg segment 34 to swing up and down around the thigh segment 32. A wheel-leg switching assembly 35 is located at the end of the lower leg segment 34. Fixed seats 121 can be provided at the four corners of the mounting base. The hip joint drive servo 31 is installed in the fixed seats 121. The joint drive servos drive the thigh segment 32 and the lower leg segment 34 to achieve two degrees of freedom movement of the leg assembly 3. The number of leg segments and joint drive servos of the leg assembly 3 can be adjusted according to actual needs to meet different application requirements. The rotor assembly 2 is mounted on the lower leg segment 34, and the direction of the rotor assembly 2 can be adjusted by adjusting the lower leg segment 34.

[0071] In one embodiment, the horizontal swing range of the thigh segment 32 is 180°, and the vertical swing range of the calf segment 34 is 90°. Alternatively, the range of motion can be adjusted according to actual needs.

[0072] In one embodiment, the propellers 21 are staggered along the height direction in a horizontal state. In this embodiment, the propellers 21 are on the same horizontal plane in flight mode. Alternatively, the staggered design of the propellers 21 can also help to further reduce the size of the robot while ensuring normal operation.

[0073] In one embodiment, the propeller 21 is further provided with a cylindrical protective cover, and the propeller 21 propels the cylindrical protective cover to achieve wheel-like or adsorption-like motion. For example, when the cylindrical protective cover rotates synchronously with the propeller 21, the propeller 21 is adjusted to a vertical state, and the cylindrical protective cover can switch to a wheel-like function, achieving wheel-like motion under the propulsion of the propeller 21; or when the propeller 21 is adjusted to an inverted state, it achieves adsorption motion under the propulsion of the propeller 21.

[0074] In one embodiment, the end of the foot 352 is provided with a claw mechanism or an adhesive mechanism. The claw mechanism or adhesive mechanism can be a common robot end effector in the prior art, used to realize the grasping or adhesive function, such as a gripper or suction cup, which can realize vertical movement or adsorption on different material surfaces.

[0075] In one embodiment, the deformable miniature quadrotor quadrupedal climbing robot is equipped with at least one of a visual sensor, an auditory sensor, an olfactory sensor, and a tactile sensor, each of which is electrically connected to the climbing control module 14. The type, quantity, and model of each sensor are determined according to actual needs. By setting up sensors, the robot acquires the ability to perceive its environment, thereby improving its level of intelligence.

[0076] Example 2:

[0077] like Figure 10 As shown, a control method for a deformable small quadrotor quadrupedal climbing robot includes the following steps:

[0078] S1. Construct the quadrotor dynamics model and the quadruped dynamics model of the climbing robot respectively;

[0079] S2. Determine the propeller lift coefficient K of the quadrotor dynamics model under different flight motion modes using an online system identification method. T ;

[0080] S3. Determine the flight motion model for the in-flight flight mode. The flight motion model includes a static flight motion model and a dynamic flight motion model. The static flight motion model uses a "feedforward + PD feedback" control method to realize the "X" type flight motion mode of the quadrotor dynamics model. The dynamic flight motion model uses a lift compensation algorithm to realize the "T", "H", or "O" type flight motion modes of the quadrotor dynamics model.

[0081] The "feedforward + PD feedback" control method satisfies the following formula:

[0082]

[0083] In the formula, u(t) is the input of the quadrotor dynamics model, x(t) is the state variable of the quadrotor dynamics model, and e(t) is the error of the quadrotor dynamics model. K is a constant coefficient related to the quadrotor dynamics model. p K is the proportional coefficient of the controller. d The differential coefficients of the controller, For feedforward term, For PD feedback items;

[0084] The lift compensation algorithm satisfies the following formula:

[0085]

[0086]

[0087] T=k(φ)w 2

[0088] In the formula, T is the lift of a single propeller, φ is the shielding angle after propeller deformation, ω is the angular velocity of the propeller, and k(φ) is the shielding angle related function. For the compensated desired motor speed, k is a scaling factor that depends on the occlusion angle correlation function k(φ). T It is the lift coefficient and is equal to k (φ=0);

[0089] S4. A two-layer CPG network and a nearest-neighbor coupling model are used to control the foot trajectory of the quadrupedal dynamics model. The two-layer CPG network includes eight oscillators and is divided into a consciousness layer for controlling the hip joint and a behavior layer for controlling the knee joint. The nearest-neighbor coupling model is as follows:

[0090]

[0091] Wherein, rotation matrix Defined as:

[0092]

[0093] In the formula, i = 1, 2, 3, 4, j = 1, 2, m = 1, 2, 3, 4, n = 1, 2, x mn and y mn Let k' represent the oscillator state variable of the nth joint of the mth leg component, where k' is the coupling strength and equals 0.1. This represents the phase difference between the mn-th oscillator and the ij-th oscillator;

[0094] S5. Train a two-layer CPG network based on the DDPG reinforcement learning algorithm to obtain the parameters of the trained behavior layer model. The reward function r of DDPG reinforcement learning is... t The formula is as follows:

[0095]

[0096] Among them, K v K is a positive velocity weighting coefficient. e A positive torque weighting coefficient, v x Let τ be the horizontal velocity and τ be the joint torque. Joint angular velocity;

[0097] S6. The flying and climbing robot locates itself and builds a map based on real-time collected environmental information to plan its path. It then adaptively selects a flight motion model or a trained two-layer CPG network to control the flying and climbing robot to achieve aerial flight or ground crawling motion.

[0098] The autonomous land-air switching control process of the flying-climbing robot mainly includes motion planning and motion control. The robot collects information about its environment through onboard sensors. After receiving the information from the sensors, the flying-climbing control module achieves localization and builds a map. Path planning is then performed based on localization and map building. The trajectory (including land and air movement paths) can be optimized using existing path planning methods. The flying-climbing robot moves according to the planned trajectory, taking into account terrain changes, and performs cross-domain movement trajectory planning and control. Ground modes can be prioritized to reduce the robot's energy consumption. For air and land movements, a quadcopter dynamics model of the flying-climbing robot is built on the Gazebo platform, and a quadruped dynamics model is built on the MATLAB platform, respectively.

[0099] (1) The air flight mode adopts a static flight control method with "feedforward + PD feedback" control mode or a deformable dynamic flight control method based on lift compensation algorithm to control the quadrotor dynamic model.

[0100] This flying robot typically has four types of flight attitudes, including "X", "O", "H", and "T" shapes. Flight control can be divided into static quadrotor control (non-deformable) and dynamic deformable control (dynamic deformation).

[0101] Under static quadrotor control, a "feedforward + PD feedback" control law based on a dynamic model is proposed:

[0102]

[0103] Where u(t) is the input of the quadrotor dynamics model, x(t) is the state variable of the quadrotor dynamics model, and e(t) is the error of the quadrotor dynamics model. K is a constant coefficient in the quadrotor dynamics model. p K is the proportional coefficient of the controller. d The differential coefficients of the controller, For feedforward term, This is the PD feedback term. It involves decomposing the overall control into model-dependent components. and the model-independent part (K) p and K d It can be directly applied to the motion control of static quadrotors.

[0104] Under dynamic deformable control, an online system identification method simplifies the modeling process of a deformable quadrotor, and a lift compensation algorithm is used to adjust the rotor's motor speed to overcome lift changes caused by dynamic rotor deformation. The online system identification method can employ conventional methods from existing technologies, such as least squares method and subspace identification method.

[0105] To address the lift loss issue caused by flight attitude deformation in climbing robots, a geometric compensation strategy is employed to adjust propeller lift. The propeller lift coefficient K is determined through online system identification for different flight attitudes. T The lift model of a single propeller can be simplified as follows:

[0106] T=k(φ)ω 2

[0107] Where T is the lift of a single propeller, φ is the obstruction angle of the propeller, ω is the angular velocity of the propeller, and k(φ) is the obstruction angle correlation function. For details, please refer to: A. Fabris, K. Kleber, D. Falanga and D. Scaramuzza, "Geometry-aware Compensation Scheme for Morphing Drones," 2021 IEEE International Conference on Robotics and Automation (ICRA), 2021, pp. 592-598.

[0108] The greater the propeller obstruction angle, the greater the lift loss. The research objective of geometric compensation algorithms is to adjust the rotor motor speed of the climbing robot to ensure that the desired lift is maintained before and after deformation. Assuming the known flight mode of the climbing robot is "O", it can be regarded as a symmetrical quadrotor smaller than the "X" mode. Due to the lift loss caused by the overlap between the propeller and the fuselage, the angular velocity of the rotor must be increased accordingly. The propeller angular velocity model after lift compensation (lift compensation model) can be expressed as:

[0109]

[0110]

[0111] in, For the compensated desired motor speed, k is a scaling factor that depends on the occlusion angle correlation function k(φ). T It is the lift coefficient and is equal to k (φ=0).

[0112] (2) The ground crawling mode is based on the central pattern generator (two-layer CPG network) and the nearest neighbor coupling model + reinforcement learning method to control the quadruped dynamics model.

[0113] The leg component 3 is a bionic foot that resembles the movement behavior of animals. It generally adopts a rhythmic movement mode, which is controlled by a central pattern generator (CPG) located in the spinal cord. This control process is controlled by the central nervous system in layers.

[0114] To reduce the time and space complexity of reinforcement learning for optimizing high-dimensional joint modes, a two-layer CPG (8 oscillators) topology network is used to generate different motion modes. The first layer of the CPG (hip joint, e.g., corresponding to the thigh segment) serves as the awareness layer, responsible for fixing the robot's basic motion modes (e.g., diagonal mode, trotting mode, jumping mode, etc.). The modal parameters of this layer are set through kinematic analysis of the quadruped robot to generate the desired basic motion modes, which can be achieved using existing technologies. The second layer of the CPG (knee joint, e.g., corresponding to the lower leg segment) serves as the behavior layer, responsible for controlling the robot's limb behavior. A reinforcement learning method based on Deep Deterministic Policy Gradient (DDPG) is used to fine-tune the CPG parameters (e.g., amplitude, frequency, and phase difference) of this layer, optimizing the robot's actual modal performance to adapt to complex environments.

[0115] The environmental adaptability of a climbing robot is measured by two metrics: the robot's forward speed and its energy consumption. Therefore, the reward function r of reinforcement learning... t This can be simplified to two things: rewarding greater speed and penalizing higher energy consumption.

[0116]

[0117] Among them, K v K is a positive velocity weighting coefficient. e A positive torque weighting coefficient, v x Let τ be the horizontal velocity and τ be the joint torque. K represents the joint angular velocity. v ·v x This represents a speed reward, which motivates the robot to move forward as fast as possible and iterates continuously until the robot converges to a certain positive speed reward value. This represents an energy penalty term to optimize the robot's energy consumption.

[0118] The hierarchical CPG topology network and nearest neighbor coupling model (with the coupling parameter characteristics of a single-leg structure symmetry) reduce the CPG parameters that need to be tuned in reinforcement learning from high dimension to low dimension. Therefore, compared with directly learning the commands of each joint of the robot, this approach is more suitable for the practical application of the robot in dynamic environments.

[0119] This robot combines an aerial rotor assembly and a ground-based leg assembly, integrating the rotor assembly directly onto the leg assembly to form a single, integrated structure. This creates a quadcopter with four rotatable leg segments, enabling the robot to perform collaborative land-air operations (such as walking, flying, climbing, manipulating, and transporting objects). It combines the mobility of a flying robot with the terrain adaptability of a legged robot, allowing for wide-range, long-distance global observation in the air and precise, short-range positioning on the ground. This solves the problems of poor stealth in flying robots and slow movement speed in crawling robots, improving the robot's maneuverability and adaptability in complex land-air environments. It also achieves lightweight, miniaturization, integration, and intelligence, with low air resistance, good obstacle-crossing performance, and long endurance, meeting the needs of miniaturized reconnaissance and operations in complex land-air environments. The robot utilizes deformable leg components and... The rotor assembly generates different motion modes during flight, such as "T", "H", "O", and "X" patterns. It adaptively changes the corresponding flight motion mode according to environmental conditions, enabling special operations such as traversing narrow and restricted areas and observing targets at extremely close range. It also has grasping and vertical wall-climbing capabilities. Compared with existing amphibious robots with a single wheel or a single leg, this device is equipped with a wheel-leg switching assembly. By switching the end of the leg assembly to either a wheel or a foot, it has efficient and flexible movement speed and ground obstacle-crossing ability. By establishing flight motion models under different motion modes in the air and using a two-layer CPG network trained based on the DDPG reinforcement learning algorithm, the flying and climbing robot can autonomously switch between air and ground, generating motion modes in different environments, improving the stability and diversity of air and ground movement, and achieving intelligent control.

[0120] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0121] The embodiments described above are merely specific and detailed examples of the embodiments described in this application, and should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the scope of protection of this application. Therefore, the scope of protection of this patent application should be determined by the appended claims.

Claims

1. A deformable, small quadrotor, quadrupedal flying and climbing robot, characterized in that: The deformable miniature quadrotor quadrupedal climbing robot includes a body (1), four rotor components (2) and four leg components (3), wherein: The fuselage (1) includes a mounting base, and a flight control module (14), an adjustment module (15) and a battery (16) all mounted on the mounting base. The adjustment module (15) includes a power module and an ESC. The battery (16), the power module, the flight control module (14) and the ESC are electrically connected in sequence. The leg assembly (3) is used to realize rotation of at least two degrees of freedom, including at least two leg segments, at least two joint drive servos and a wheel-leg switching assembly (35). Each joint drive servo is connected to a leg segment in a one-to-one correspondence and is electrically connected to the fly-climb control module. Each leg segment is connected in sequence and driven to rotate by the corresponding joint drive servo. The wheel-leg switching assembly (35) includes a roller (351) and a foot (352). The foot (352) is rotatably connected to the rotation axis of the roller (351). By rotating the foot (352), the end of the leg assembly (3) is switched to the foot (352) or the roller (351). Each leg assembly (3) is rectangularly distributed on the mounting base and is driven by the corresponding joint drive servo to realize horizontal swing around the mounting base. The rotor assembly (2) includes a propeller (21) and a motor (22). The propeller (21) is driven to rotate by the motor (22). The motor (22) corresponds one-to-one with the leg assembly (3) and is fixed to any of the leg sections of the leg assembly (3) except for the leg section connected to the mounting base. The motor (22) is also electrically connected to the ESC. In the flight mode, the propeller (21) is in a horizontal state. The leg assembly (3) switches between four flight motion modes: "T", "H", "O", and "X". The "X" flight motion mode is achieved by using a "feedforward + PD feedback" control method. The "feedforward + PD feedback" control method satisfies the following formula: ； In the formula, u(t) is the input of the quadrotor dynamics model, x(t) is the state variable of the quadrotor dynamics model, and e(t) is the error of the quadrotor dynamics model. K is a constant coefficient related to the quadrotor dynamics model. p K is the proportional coefficient of the controller. d The differential coefficients of the controller, For feedforward term, This is a PD feedback item.

2. The deformable miniature quadrotor quadrupedal climbing robot as described in claim 1, characterized in that: The mounting base includes multiple connecting rods (17) and a first support plate (11), a second support plate (12), and a third support plate (13) arranged sequentially and parallel to each other. The first support plate (11) and the second support plate (12) are connected. The two ends of the connecting rods (17) are connected to the second support plate (12) and the third support plate (13) respectively. The climbing control module (14) is installed on the first support plate (11). The adjustment module (15) passes through the second support plate (12). The battery (16) is located between the second support plate (12) and the third support plate (13).

3. The deformable miniature quadrotor quadrupedal climbing robot as described in claim 1, characterized in that: The rotor assembly (2) also includes a limiting support plate (23) connected to the motor (22) for limiting the rotation of the leg section connected to the rotor assembly (2).

4. The deformable miniature quadrotor quadrupedal climbing robot as described in claim 1, characterized in that: The leg segment consists of two segments, including a thigh segment (32) and a lower leg segment (34). The joint drive servo consists of two segments, including a hip joint drive servo (31) and a knee joint drive servo (33). The hip joint drive servo (31) is connected to the mounting base and is used to drive the thigh segment (32) to swing horizontally around the mounting base. The knee joint drive servo (33) is used to drive the lower leg segment (34) to swing up and down around the thigh segment (32). The wheel-leg switching assembly (35) is located at the end of the lower leg segment (34).

5. The deformable miniature quadrotor quadrupedal climbing robot as described in claim 4, characterized in that: The horizontal swing range of the thigh segment (32) is 180°, and the vertical swing range of the calf segment (34) is 90°.

6. The deformable miniature quadrotor quadrupedal climbing robot as described in claim 1, characterized in that: Each of the propellers (21) is staggered along the height direction in a horizontal state.

7. The deformable miniature quadrotor quadrupedal climbing robot as described in claim 1, characterized in that: The propeller (21) is also provided with a cylindrical protective cover, and the propeller (21) propels the cylindrical protective cover to achieve wheel-like motion or adsorption motion.

8. The deformable miniature quadrotor quadrupedal climbing robot as described in claim 1, characterized in that: The foot (352) is provided with a claw or adhesive mechanism at its end.

9. The deformable miniature quadrotor quadrupedal climbing robot as described in claim 1, characterized in that: The deformable small quadcopter quadrupedal climbing robot is equipped with at least one of a visual sensor, an auditory sensor, an olfactory sensor, and a tactile sensor, and each of the sensors is electrically connected to the climbing control module (14).

10. A control method for a deformable small quadrotor quadrupedal climbing robot, characterized in that: The control method for the deformable miniature quadrotor quadrupedal climbing robot includes the following steps: S1. Construct the quadrotor dynamics model and the quadruped dynamics model of the climbing robot respectively; S2. Determine the propeller lift coefficient K of the quadrotor dynamics model under different flight motion modes using an online system identification method. T ; S3. Determine the flight motion model for the in-flight flight mode. The flight motion model includes a static flight motion model and a dynamic flight motion model. The static flight motion model uses a "feedforward + PD feedback" control method to realize the "X" type flight motion mode of the quadrotor dynamics model. The dynamic flight motion model uses a lift compensation algorithm to realize the "T", "H", or "O" type flight motion modes of the quadrotor dynamics model, wherein: The "feedforward + PD feedback" control method satisfies the following formula: ； In the formula, u(t) is the input of the quadrotor dynamics model, x(t) is the state variable of the quadrotor dynamics model, and e(t) is the error of the quadrotor dynamics model. K is a constant coefficient related to the quadrotor dynamics model. p K is the proportional coefficient of the controller. d The differential coefficients of the controller, For feedforward term, For PD feedback items; The lift compensation algorithm satisfies the following formula: ； ； In the formula, T is the lift of a single propeller, ϕ is the shielding angle after the propeller deforms, ω is the angular velocity of the propeller, and k(ϕ) is the shielding angle related function. For the compensated desired motor speed, k is a scaling factor that depends on the occlusion angle correlation function k(ϕ). T It is the lift coefficient and is equal to k (ϕ=0); S4. A two-layer CPG network and a nearest-neighbor coupling model are used to control the foot movement trajectory of the quadrupedal dynamics model. The two-layer CPG network includes eight oscillators and is divided into a consciousness layer for controlling the hip joint and a behavior layer for controlling the knee joint. The nearest-neighbor coupling model is as follows: ； Wherein, rotation matrix Defined as: ； In the formula, i = 1, 2, 3, 4, j = 1, 2, m = 1, 2, 3, 4, n = 1, 2, x mn and y mn Let k' represent the oscillator state variable of the nth joint of the mth leg component, where k' is the coupling strength and equals 0.

1. This represents the phase difference between the mn-th oscillator and the ij-th oscillator; S5. Train a two-layer CPG network based on the DDPG reinforcement learning algorithm to obtain the trained behavior layer model parameters. The reward function r of the DDPG reinforcement learning is... t The formula is as follows: ； Among them, K v K is a positive velocity weighting coefficient. e A positive torque weighting coefficient, v x For horizontal velocity, For joint torque, Joint angular velocity; S6. The flying and climbing robot locates itself and builds a map based on real-time collected environmental information to plan its path. It then adaptively selects a flight motion model or a trained two-layer CPG network to control the flying and climbing robot to achieve aerial flight or ground crawling motion.

Images