All-terrain amphibious quadruped robot system and control method based on full-drive control capability

Abstract

The application discloses a land-air amphibious quadruped robot system and a control method based on full-drive control capability. The system comprises a body, four mechanical legs equipped with drive units, and four air power duct units respectively connected to the hip joint parts of the corresponding mechanical legs. The core lies in that the drive functions are fully reused. When flying, the mechanical leg drive units are reused as thrust vector adjusting mechanisms to adjust the direction of the duct thrust. The control method comprises the following steps: establishing a unified full-drive dynamics model, deciding a motion mode, and distributing optimal joint torque and duct instructions in real time based on the model and an optimization algorithm. Through drive depth reuse and a unified control framework, the application solves the structural redundancy problem, realizes high-mobility ground gaits, full-attitude agile flight and smooth transition between land and air modes.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of robot technology and unmanned system, and particularly relates to a hybrid robot with ground moving and flying capabilities, and particularly relates to a land-air amphibious quadruped robot system achieving full-drive control through vector thrust and a control method thereof. BACKGROUND

[0002] In recent years, quadruped robots and unmanned aerial vehicles have made significant progress. Quadruped robots such as Spot of Boston Dynamics have shown excellent mobility in terms of walking, running and jumping on the ground, but their mobility is limited to two-dimensional ground and cannot effectively deal with obstacles such as cliffs, high walls or wide ditches. Unmanned aerial vehicles, especially multi-rotor unmanned aerial vehicles, have excellent three-dimensional spatial mobility and can easily fly over ground obstacles, but their flight usually relies on static stable aerodynamic layout, which is high in energy consumption, short in endurance and lacks effective physical interaction ability in the air, and cannot achieve actions such as pushing doors and pressing.

[0003] To integrate the advantages of both, some existing land-air amphibious robot solutions use simple structure stacking, such as adding an independent rotor module on the back of a quadruped robot (CN116374041B). This solution has several inherent disadvantages: the flight and ground movement systems are independent of each other, the structure is redundant, the weight is large, and the efficiency is low; the additional flight module also seriously affects the stability and flexibility of the ground movement.

[0004] To solve the above problems, further research focuses on the reuse of structure and function, aiming to improve the integration and efficiency of the system, but existing solutions still have obvious limitations in the depth and way of "reuse".

[0005] For example, Chinese patent CN219947806U discloses a "variable attitude cross-medium multi-environment robot". The technical core of this solution is its "power switching mechanism", which includes "mobile engagement components" and "fixed engagement components", which realize time-sharing reuse of power sources through a complex mechanical clutch (spline clamping plate and clamping groove meshing or separating), and specifically use a single drive motor to drive the flight propeller or the ground / water dual-purpose wheel propeller in time-sharing mode. Although this design achieves reuse of power sources, it has the following inherent problems:

[0006] Low system reliability and dynamic response speed: the introduction of the mechanical switching system has a complex structure, which increases the potential failure points, and the delay of the mechanical switching process makes it unable to meet the needs of high dynamic motion scenarios.

[0007] Insufficient control accuracy: its attitude adjustment mechanism is mainly used for macro switching between flight, ground and other preset modes, rather than real-time and continuous thrust vector adjustment during flight, which limits its air mobility.

[0008] Another example is the "MULTI-MODAL MOBILITY UNMANNED VEHICLE" disclosed in US12344407B2. This scheme proposes to set the propeller coaxially with the wheel and adjust the overall orientation of the wheel / paddle composite unit through the movement of the mechanical leg. However, its "reuse" level is shallow, only attitude mechanism reuse, and there is obvious driving redundancy. Its claim explicitly states that "the wheel and the propeller are driven by different actuators". This means that it needs to configure two independent sets of drive motors for ground movement and flight respectively. More importantly, this scheme fails to achieve deep reuse of driving functions: the leg joint motor is only used for macro attitude setting between different motion modes, and is not reused to achieve real-time thrust vector control during flight, which limits its flight attitude control ability and agility. In addition, this scheme uses wheeled ground movement, which limits its application to relatively flat ground and cannot take advantage of the high passability of legged robots in complex terrain.

[0009] In addition, Chinese patent CN114801613A discloses a deformable quadrotor quadruped flying and climbing robot, which fixes the rotor assembly on the leg assembly. Although this scheme also involves the use of leg structures, the degree of driving reuse is limited, and the flight performance is significantly constrained.

[0010] The core limitation is that the "air flight mode" of this scheme is limited to the propeller always being parallel to the body plane, i.e. the lift direction is always perpendicular to the body plane. The leg servo is only used for macro shape switching between different motion modes (such as "X" type, "H" type) to pass through narrow areas, and is not reused to achieve real-time attitude adjustment and thrust vector control. When the propeller is in a non-horizontal state (such as vertical or inverted), the robot is defined as performing "wheeled motion" or "suction motion", not flight.

[0011] In summary, this scheme is essentially a traditional quadrotor with variable configuration, and its flight control relies on differential, which cannot achieve high maneuverability actions such as side flight and inverted flight, and does not achieve deep reuse of driving functions.

[0012] In addition, there are also some robot schemes for amphibious operations. For example, there is a technology scheme that discloses a variable vector propulsion walking dual-mode amphibious robot. This scheme installs a total of seven propellers on the body and each leg to achieve the division and cooperation of lift and horizontal propulsion in water, supporting the movement of the robot in underwater and water surface environments.

[0013] However, such water-land solutions are optimized for high-density, high-damping hydrodynamic environments. Its system operates under the buoyancy support of water, and the main resistance that the propeller overcomes is fluid resistance. This "buoyancy support" and "propulsion decoupling" design is completely unable to adapt to the fundamental physical challenges faced in the "amphibious" field.

[0014] The specific technical field to which the present invention belongs is the "amphibious" field, and the core problem is to achieve instantaneous, nonlinear transition from "ground support" to "active thrust support" in low-density, low-damping air. The robot must rely solely on its propulsion system to generate sufficient lift to overcome 100% of its own gravity, and the low-damping nature of air will dramatically amplify any attitude disturbance, which poses extreme requirements for the coupling, real-time performance, and robustness of control.

[0015] Therefore, the "propulsion function decoupling" layout in the above water-land solution will bring unacceptable defects when applied in the land-air field:

[0016] Structural redundancy and low efficiency: such solutions have too many propulsion devices (such as seven), complex structures, and significant redundancy between propulsion units.

[0017] Control method is fragmented: each propeller mainly relies on independent speed regulation to achieve attitude and direction control, which is inefficient in aerodynamics, and fails to fully utilize the attitude variable characteristics of the leg joints to generate effective propulsion force while maintaining lift.

[0018] In summary, although the existing amphibious robot technology has made some explorations in functional reuse, it still generally has one or more of the following key problems:

[0019] 1) Insufficient drive integration: either relying on complex mechanical switching mechanisms or having redundant configurations of execution motors, all of which require additional mechanical structures (such as rudders) to achieve amphibious functions, failing to achieve true lightweight and efficient driving.

[0020] 2) Limited air mobility: the thrust adjustment mechanism is relatively single, limiting the robot's flight agility and attitude control ability in complex space.

[0021] 3) Difficulty in balancing ground adaptability and air capability: some solutions simplify ground movement methods (such as using wheels) in order to integrate flight capabilities, sacrificing the ability to pass through complex and unstructured terrains.

[0022] 4) Control system fragmentation: different control strategies are usually designed for ground and air, lacking a unified dynamic model and control framework to achieve smooth and robust control of both modes and transition processes.

[0023] The above-mentioned shortcomings reflect the inherent contradiction between the mobility and integration of traditional land-air amphibious robots, and a new solution with high structural integration, deep drive reuse and unified control framework is urgently needed to solve this problem. SUMMARY

[0024] In view of the problems existing in the prior art, especially the fundamental problem of the mobility and integration of traditional land-air amphibious robots, the present application provides a land-air amphibious quadruped robot system based on full-drive control capability, which adopts a brand-new design concept of "full reuse of drive functions".

[0025] "Full reuse of drive functions" here is defined as: giving different core functions to the same set of actuators (including but not limited to leg joint motors) in different motion modes, and realizing real-time, dynamic and continuous adjustment of functions. The core innovation of the present application is that in the ground mode, the leg joint motors serve as the main driving unit for ground movement, while in the flight mode, the same batch of leg joint motors seamlessly switch functions and are reused as the direct driving unit of the thrust vector system, responsible for real-time and accurate adjustment of the pitch and roll direction of the propeller thrust. This is fundamentally different from the prior art (such as CN114801613A). The prior art only uses leg rudders to switch flight configurations (such as changing to "H" type), and its flight mode is strictly limited to the state of propeller in the plane of the fuselage, and cannot realize thrust vector control. The reuse degree of the present application realizes complete integration at the drive level. Compared with the "time-sharing reuse of power sources" of CN219947806U and the "reuse of attitude mechanisms" of US12344407B2, the present application further reuses 12 degrees of freedom of the driving unit, and can still realize independent control of the pitch and roll (4 2=8 degrees of freedom) of the 4 ducted fans without additional rudders. Therefore, the system breaks through the limitation of traditional "horizontal flight" and realizes "full-drive control" of the attitude in the air, significantly improving the structural integration, dynamic control capability and air mobility (such as side-body hovering).

[0026] "Full-drive control" is a core concept in the fields of automation control, robotics, mechanical engineering, etc., which refers to a technical solution that directly and actively controls all independent motion degrees of freedom (Degree of Freedom, DOF) of the controlled object (such as robots, mechanical structures, vehicles, etc.) through independent driving units (such as motors, cylinders, hydraulic motors, etc.), thereby realizing omnidirectional movement. The core is "no underdrive" - each motion dimension that needs to be controlled has a dedicated driving device to provide power and adjustment capability, rather than relying on passive structures (such as gravity, springs) or other degrees of freedom for indirect driving.

[0027] To achieve the above object, the full-drive control amphibious quadruped robot system composition specifically comprises:

[0028] A body 1;

[0029] Four mechanical legs 2, symmetrically arranged on both sides of the body 1;

[0030] Four air power duct units 3, respectively connected with the hip joint parts of the corresponding mechanical legs 2;

[0031] And four drive units 201 mounted on the body 1, respectively corresponding to driving the four mechanical legs 2;

[0032] Among them, the drive function of the drive unit 201 is fully reused:

[0033] In the ground walking mode, the drive unit 201 is used to drive the mechanical leg 2 to realize the multi-degree-of-freedom walking of the robot;

[0034] In the air flight mode, the drive unit 201 is reused to drive the mechanical leg 2 to adjust the posture, and then provide at least two degrees of freedom of the thrust vector adjustment ability for the air power duct unit 3 connected therewith; and through the drive unit 201, the thrust of the four air power duct units 3 is uniformly vector adjusted, realizing the full-drive control of the robot including but not limited to side flight, upside down flight, side body hovering and six-degree-of-freedom motion.

[0035] Preferably, the four mechanical legs 2 are the same in structure, and each mechanical leg 2 further comprises a thigh rod 202, a drive connecting rod 203, a shank rod 204 and a foot end 205.

[0036] Preferably, a pair of mechanical legs 2 are arranged on each side of the body 1.

[0037] Each of the mechanical legs 2 is equipped with a drive unit 201, which is mounted on the body 1. The drive unit 201 is internally divided into three drive parts (for example, three independent motors or rudders), which cooperatively realize three movement degrees of freedom of each leg.

[0038] In a preferred embodiment, the layout and transmission mode of the three drive parts are as follows:

[0039] The first drive part (hip joint abduction / adduction): used to drive the whole mechanical leg 2 to swing around the first axis (for example, the X axis parallel to the forward direction of the body).

[0040] Second and third driving parts (leg flexion and extension): The two driving parts (for example, in a top-down stacked arrangement) are mounted on the output shaft of the first driving part for driving the mechanical leg 2 to perform flexion and extension movement around the second axis (for example, Y axis) in coordination.

[0041] To realize leg flexion and extension, the structure adopts a connecting rod transmission, specifically:

[0042] The output end of the second driving part (for example, the upper motor) is directly connected to one end of the thigh rod 202 for rotating control of the thigh rod 202.

[0043] The other end (i.e., the knee joint) of the thigh rod 202 is rotatably connected to the proximal end of the calf rod 204.

[0044] The output end of the third driving part (for example, the lower motor) drives one end of the driving connecting rod 203.

[0045] The other end of the driving connecting rod 203 is mounted at the distal end of the calf rod 204 (or a specific mounting point for driving the calf rod 204 to rotate up and down around the knee joint).

[0046] The other end of the calf rod 204 is mounted with a foot end 205.

[0047] In one specific embodiment, each driving part of the driving unit 201 can be a joint servo motor. The joint servo motor can be controlled by a PWM (Pulse Width Modulation) signal.

[0048] In another embodiment, each driving part of the driving unit 201 can also be a reduction joint motor.

[0049] In other embodiments, the driving parts of the driving unit 201 can also be units with continuous rotation with air or liquid driving.

[0050] Preferably, the amphibious quadruped robot system also comprises:

[0051] Fixing supports 4, the four air duct units 3 are connected to the hip joint parts of the mechanical legs 2 through the fixing supports 4, respectively.

[0052] A perception module 5, including an on-board inertial measurement unit, a depth camera, and a laser radar, for perceiving the external environment and the robot body condition.

[0053] A high-level decision and gait planning module 6 for receiving motion instructions and environment perception information, generating a desired motion trajectory and foot end landing point sequence.

[0054] A posture controller 7 is in communication connection with the high-level decision and gait planning module, configured to calculate total spatial force and total spatial moment at the center of mass of the robot body required for tracking the desired motion trajectory based on the desired motion trajectory and the current state of the robot body, using a rigid body dynamics model of the robot, and to distribute the total spatial force and total spatial moment to each foot end in contact with the ground to generate target contact force of each foot end;

[0055] A revolute joint controller 8 is in communication connection with the posture controller, configured to calculate and output desired signal instructions for driving each driving unit 201 based on the target contact force of each foot end and the current configuration of the robot by means of a Jacobian transpose matrix.

[0056] The above preferred structures can be selectively adopted according to actual application requirements, and can be arbitrarily combined to achieve the purpose of the present application, and the present application is not limited to the specific combination listed above.

[0057] The system performs full-drive control on the driving unit 201 through the desired signal instructions.

[0058] Preferably, the posture controller 7 comprises:

[0059] A model-based feedforward control unit 701 is configured to perform inverse dynamics calculation using the rigid body dynamics model, taking the desired motion trajectory and the current state as input, and outputting the total spatial force and total spatial moment required at the center of mass of the robot body;

[0060] A contact force distribution optimization unit 702 is configured to take the total spatial force and total spatial moment as an optimization target, and to solve the target contact force of each foot end by means of a quadratic programming algorithm under the condition of satisfying the friction cone constraint and the no-pull force constraint.

[0061] Preferably, the constraint condition for optimization performed by the contact force distribution optimization unit 702 comprises:

[0062] The normal component of the contact force of each foot end is greater than zero;

[0063] The ratio of the lateral component to the normal component of the contact force of each foot end is less than a preset friction coefficient;

[0064] The sum of the target contact forces of all foot ends is equal to the total spatial force;

[0065] The sum of the moments generated by the target contact forces of all foot ends about the center of mass of the robot body is equal to the total spatial moment.

[0066] Preferably, the revolute joint controller 8 is specifically configured to:

[0067] According to the current joint angle of the robot, the Jacobian matrix of each leg under the current configuration is calculated.

[0068] The expected feedforward torque of each joint is calculated by the formula wherein is the joint torque vector, is the transpose of the Jacobian matrix, and f is the target contact force vector of the foot end;

[0069] A PID controller based on joint angle or torque feedback is introduced to generate compensation torque to correct model errors and external disturbances;

[0070] The expected feedforward torque is added to the compensation torque to obtain the final expected torque command.

[0071] Preferably, the high-level decision and gait planning module 6 is also used for real-time online foot end landing point adjustment according to the environment perception information to avoid obstacles or adapt to uneven terrain.

[0072] The application also provides a full-drive control method for a land-air amphibious quadruped robot, which is applied to the land-air amphibious quadruped robot system and includes the following steps:

[0073] State perception step: The attitude, position, and speed information of the body 1, the angle and angular velocity information of each mechanical leg 2 joint, the contact force information of the foot end 205 with the ground, and the rotation speed information of each air power duct unit 3 are obtained in real time through the perception module 5;

[0074] Mode decision step: The high-level decision and gait planning module 6 decides whether the current mode is the ground walking mode, the air flight mode, or the mode transition state according to the task instruction and the environment perception information from the perception module 5, and generates the corresponding body expected motion trajectory;

[0075] Unified dynamics modeling step: A unified full-drive dynamics model including the body 1, the mechanical leg 2, the air power duct unit 3, and the vector angle thereof is established; based on the expected motion trajectory and the robot body condition, the rigid body dynamics model is used to calculate the required total space force and total space torque;

[0076] Optimization distribution step: The attitude controller 7 takes tracking the expected motion trajectory as the target, uses a constrained optimization algorithm to solve the optimal joint torque command, the rotation speed of each air power duct unit 3, and the vector angle command in real time based on the unified dynamics model;

[0077] Execution control step: The joint torque command is output to the driving unit 201, and the duct rotation speed and vector angle command are output to the air power duct unit 3 and the driving unit 201, respectively, after being solved by the inverse algorithm.

[0078] Preferably, the optimization problem in the optimization allocation step is mathematically expressed as:

[0079]

[0080]

[0081]

[0082] wherein, is the joint torque vector; is the nacelle rotation speed vector; , is the nacelle pitch and yaw vector angle vector; M is the mass matrix; C is the Coriolis and centrifugal force term; J is the Jacobian matrix, is the transpose of the Jacobian matrix; F thrust is the nacelle thrust vector; is the external force; q is the generalized coordinate vector (body pose + joint angle); represents the first order derivative of q with respect to time, i.e. the generalized velocity; represents the second order derivative of q with respect to time, i.e. the generalized acceleration, denotes the desired acceleration; W 1, W 2， W 3 is a weight matrix, which can be dynamically adjusted.

[0083] Preferably, the mode transition state adjusts the priority of the foot end contact force constraint and the body attitude tracking constraint in the optimization objective through a weight coefficient, to realize smooth take-off and landing.

[0084] This scheme discards the redundant layout of "vertical lift" and "horizontal propulsion" separation in the amphibious scheme, and instead relies only on four air-powered nacelle fans distributed on four mechanical legs. Through real-time collaborative control of leg joint angles and nacelle thrust, this single power system has all the functions of vertical take-off and landing, horizontal propulsion, and full attitude control, thereby significantly reducing the number of propulsion devices, reducing system mass and power consumption.

[0085] This "structure-power-control" integrated efficient motion mechanism not only simplifies the power layout, but also provides the high response speed and coupled control ability necessary to cope with the extremely unstable working conditions in the land-air transition phase, providing a new implementation approach for multi-modal robots to quickly switch and adaptively maneuver on complex terrain.

[0086] Beneficial effects:

[0087] In summary, the present application has one or more of the following advantages compared with the prior art: Extremely high maneuverability: by reusing the driving units to drive the mechanical legs to adjust the posture, the air power duct unit is provided with at least two degrees of freedom of thrust vector adjustment capability, so that the robot is free from the dependence of the traditional unmanned aerial vehicle on the horizontal posture, and can realize omnidirectional flight in any posture including side flight, inverted flight and side body hovering in the air, and has extremely strong passing ability in narrow and complex space.

[0088] Compact and efficient structure: by reusing the driving units as the thrust vector adjustment mechanism, the additional rudder required in the traditional vector thrust scheme is omitted, the driving function is deeply reused, and the total weight of the system is significantly reduced, and the structural complexity and potential failure points are reduced.

[0089] Unified robust control: based on the unified full-drive dynamics model and the optimization distribution step, the full-drive unified control framework of the present application can seamlessly handle all dynamic conditions in the ground walking, air flight and mode transition stages, and realize fast and robust posture response and anti-interference ability by real-time solving of optimal joint torque and duct command.

[0090] Diverse functions: based on the unified dynamics model and the optimization distribution, the air power duct unit is not only used for flight, but also can be called in the ground walking mode, and the thrust can be used to assist in realizing super large angle climbing, increasing the foot end pressure to resist slipping, or adjusting the air posture during jumping, thereby expanding the ground maneuvering boundary of the robot. BRIEF DESCRIPTION OF DRAWINGS

[0091] Figure 1 It is a physical schematic diagram of the amphibious quadruped robot;

[0092] Figure 2 It is a system control schematic diagram of the amphibious quadruped robot;

[0093] Figure 3 It is a system framework diagram of the amphibious quadruped robot;

[0094] Figure 4 It is a composition diagram of the mechanical leg;

[0095] Figure 5 It is a flowchart of the control method;

[0096] Figure 6 It is a thrust and foot end force change curve diagram in the transition control stage.

[0097] 1 is a body; 2 is a mechanical leg; 3 is an air power duct unit; 4 is a fixed support; 5 is a sensing module; 6 is a high-level decision and gait planning module; 7 is a posture controller; 8 is a rotating joint controller;

[0098] 201 is a driving unit; 202 is a thigh rod; 203 is a driving connecting rod; 204 is a shank rod; 205 is a foot end;

[0099] 701 is a feedforward control unit; 702 is a contact force distribution optimization unit. DETAILED DESCRIPTION

[0100] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments of the present application.

[0101] The example embodiments according to the present application will be described in detail below with reference to the drawings. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments of the present application, and it should be understood that the present application is not limited to the example embodiments described herein.

[0102] Embodiment 1

[0103] Reference Figures 1-4 This embodiment aims to illustrate the system composition, which provides a land-air amphibious quadruped robot system based on full-drive control capability, and the system takes an electrically driven quadruped robot with 12 degrees of freedom (3 joints per leg) as a platform.

[0104] The specific composition includes:

[0105] A body 1;

[0106] Four mechanical legs 2 symmetrically arranged on both sides of the body 1, each mechanical leg 2 being equipped with a driving unit 201, the driving unit 201 being mounted on the body 1 and corresponding to driving the four mechanical legs 2 respectively, a servo steering engine capable of precise torque control being selected and equipped with a high-precision encoder and a torque sensor, for driving the mechanical leg 2 to realize ground walking, and for adjusting the posture of the mechanical leg 2 in the flight state; the driving unit 201 is internally divided into three driving parts, which cooperatively realize three movement degrees of freedom of each leg.

[0107] The body 1 and the mechanical leg 2 adopt a lightweight carbon fiber and aluminum alloy frame;

[0108] Four air duct units 3, duct propellers are selected, electric duct fans with a diameter of 150mm are selected, the maximum thrust of a single unit can reach 2kgf, and they are respectively connected with the hip joint parts of the corresponding mechanical legs 2, which provide thrust to realize the flight of the robot, and the thrust direction is adjusted through the attitude change of the mechanical legs 2, thereby realizing the adjustment of the flight attitude.

[0109] The robot system realizes the control of the ground walking and flight attitude through multiplexing of the drive unit 201.

[0110] Preferably, the four mechanical legs 2 are structurally identical, and each mechanical leg 2 further includes a thigh rod 202, a drive connecting rod 203, a shank rod 204, and a foot end 205.

[0111] The layout and transmission mode of the three drive parts are as follows:

[0112] The first drive part (hip joint abduction / adduction): used to drive the entire mechanical leg 2 to swing around the first axis (for example, the X axis parallel to the forward direction of the robot body) for abduction / adduction.

[0113] The second and third drive parts (leg flexion / extension): The two drive parts (for example, arranged in a stacked manner) are installed on the output shaft of the first drive part, and are used to cooperatively drive the mechanical leg 2 to perform flexion / extension movement around the second axis (for example, the Y axis).

[0114] To realize leg flexion / extension, the structure adopts a connecting rod transmission, specifically:

[0115] The output end of the second drive part (for example, the upper motor) is directly connected to one end of the thigh rod 202, and the thigh rod 202 is rotationally controlled;

[0116] The other end of the thigh rod 202 (i.e., the knee joint) is rotationally connected to the proximal end of the shank rod 204;

[0117] The output end of the third drive part (for example, the lower motor) drives one end of the drive connecting rod 203;

[0118] The other end of the drive connecting rod 203 is installed at the distal end of the shank rod 204 (or a specific mounting point, used to drive the shank rod 204 to rotate up and down around the knee joint).

[0119] The other end of the shank rod 204 is installed with the foot end 205.

[0120] The foot end 205 is made of soft silicone or elastomer materials such as polyurethane, to increase the friction force during ground walking and provide impact cushioning during landing.

[0121] Each driving part of the driving unit 201 can be a joint servo.

[0122] The joint servo can be controlled by a PWM (Pulse Width Modulation) signal.

[0123] Each driving part of the driving unit 201 can also be a reduction joint motor.

[0124] The driving part of the driving unit 201 can also be a unit with continuous rotation with air or liquid driving.

[0125] Preferably, the amphibious quadruped robot system also comprises:

[0126] A fixed support 4, the four air duct units 3 are connected to the hip joint part of the mechanical leg 2 through the fixed support 4 respectively;

[0127] A perception module 5, including an on-board inertial measurement unit, a depth camera and a laser radar, for perceiving the external environment and the robot body condition;

[0128] A high-level decision and gait planning module 6, for receiving motion instructions and environment perception information, generating a desired motion trajectory and foot landing point sequence;

[0129] A posture controller 7, in communication connection with the high-level decision and gait planning module, for calculating the total spatial force and total spatial moment required at the body center of mass for tracking the desired motion trajectory based on the desired motion trajectory and the robot body condition, using the rigid body dynamics model of the robot; and optimizing the distribution of the total spatial force and total spatial moment to each foot end in contact with the ground, to generate the target contact force of each foot end;

[0130] A rotating joint controller 8, in communication connection with the posture controller, configured to calculate and output the expected signal instructions for driving each driving unit 201 based on the target contact force of each foot end and the current configuration of the robot through the Jacobian transpose matrix;

[0131] The system performs full-drive control on the driving unit 201 through the expected signal instructions.

[0132] The posture controller 7 comprises:

[0133] A model-based feedforward control unit 701, for performing inverse dynamics calculation using the rigid body dynamics model, taking the desired motion trajectory and current state as input, and outputting the total spatial force and total spatial moment required at the body center of mass;

[0134] The contact force distribution optimization unit 702 is configured to solve the target contact forces of each foot end by a quadratic programming algorithm, with the total spatial force and the total spatial force moment as optimization objectives, under the conditions of satisfying the friction cone constraint and the no-pulling force constraint.

[0135] The constraint conditions for the optimization performed by the contact force distribution optimization unit 702 include:

[0136] The normal component of the contact force of each foot end is greater than zero;

[0137] The ratio of the transverse component to the normal component of the contact force of each foot end is less than a preset friction coefficient;

[0138] The sum of the target contact forces of all foot ends is equal to the total spatial force;

[0139] The sum of the moments generated by the target contact forces of all foot ends about the center of mass of the robot is equal to the total spatial force moment.

[0140] The rotational joint controller 8 is in communication connection with the attitude controller, and is configured to calculate and output the expected signal instructions for driving each driving unit 201 by a Jacobian transpose matrix based on the target contact forces of each foot end and the current configuration of the robot;

[0141] The rotational joint controller 8 is specifically configured to:

[0142] Calculate the Jacobian matrix of each leg under the current configuration according to the current joint angle of the robot;

[0143] Calculate the expected feedforward torque of each joint by the formula wherein is the joint torque vector, is the transpose of the Jacobian matrix, and f is the target contact force vector of the foot end;

[0144] Introduce a PID controller based on joint angle or torque feedback to generate a compensation torque to correct model errors and external disturbances;

[0145] Add the expected feedforward torque and the compensation torque to obtain the final expected torque instruction.

[0146] The high-level decision and gait planning module 6 is also used for real-time online foot end landing point adjustment according to the environment perception information, to avoid obstacles or adapt to uneven terrain.

[0147] Embodiment 2

[0148] With reference to Figure 5For the second embodiment of the present application, this embodiment is intended to illustrate the "full-drive control method" framework, which includes a "mode decision step" for deciding whether the robot should be in the ground walking mode, the air flight mode, or the mode transition state. The embodiment provides a full-drive control method based on the amphibious quadruped robot, comprising the following steps:

[0149] State perception step: Real-time acquisition of the attitude, position, and speed information of the body 1, the angle and angular velocity information of each mechanical leg 2 joint, the contact force information of the foot end 205 with the ground, and the rotation speed information of each air duct unit 3 through the perception module 5;

[0150] Mode decision step: The high-level decision and gait planning module 6 decides whether to be in the ground walking mode, the air flight mode, or the mode transition state according to the task instruction and the environmental perception information from the perception module 5, and generates the corresponding body desired motion trajectory;

[0151] Unified dynamics modeling step: Establish a unified full-drive dynamics model including the body 1, the mechanical leg 2, the air duct unit 3, and the vector angle thereof; based on the desired motion trajectory and the robot body condition, calculate the required total space force and total space moment using the rigid body dynamics model; pre-derive and code the symbolic dynamics model in the MATLAB / Simulink or C++ environment, and calculate the real-time state parameters online.

[0152] Establish a complete equation (formula 1) including all degrees of freedom (body 6DOF + 12 joints):

[0153]

[0154] q: Generalized coordinate vector (body pose + joint angle). : represents the first order derivative of q with respect to time, that is, the generalized velocity. represents the second order derivative of q with respect to time, that is, the generalized acceleration.

[0155] M: Generalized mass matrix of the system.

[0156] C: Coriolis force and centrifugal force matrix.

[0157] G: Gravity term.

[0158] B: Mapping matrix of joint torque.

[0159] : Joint torque vector.

[0160] , : Transposition of the contact Jacobian matrix and foot end contact force.

[0161] , : transpose of the thrust Jacobian matrix and the jet thrust vector. The Jacobian matrix is a function of joint angles and vector angles , .

[0162] : external forces.

[0163] Equation 1 is the rigid body unified dynamics equation (also known as the motion equation), which is established based on Lagrange mechanics:

[0164] The left side of the equation is the "passive force" of the system:

[0165] is the mass matrix x acceleration, i.e. the inertial force (the tendency of the system to resist acceleration);

[0166] is the Coriolis force and centrifugal force (force generated in the rotating coordinate system);

[0167] is the gravity.

[0168] The right side of the equation is the "active force" applied:

[0169] is the torque provided by the driving unit 201 through the transmission system at the joint of the mechanical leg 2 (T );

[0170] is the contact force generated by the foot end contacting the ground (F );

[0171] is the thrust generated by the air dynamic jet (F ).

[0172] This formula shows that "the motion (acceleration ) of the robot at any moment is determined by its own inertia, gravity, motor force, ground contact force and jet thrust". It is the basis for all subsequent optimization physical constraints.

[0173] Optimization distribution step: based on the unified dynamics model, the optimal joint torque command and the speed and vector angle command of each air dynamic jet unit 3 are solved in real time by using a constrained optimization algorithm with the goal of tracking the desired motion trajectory by the posture controller 7. The optimization problem in the optimization distribution step is mathematically expressed as (Equation 2):

[0174]

[0175]

[0176]

[0177] where, is the joint torque vector; is the nacelle rotation speed vector; , is the nacelle pitch and yaw vector angle; M is the mass matrix; C is the Coriolis and centrifugal force terms; J is the Jacobian matrix, is the transpose of the Jacobian matrix; F thrust is the nacelle thrust vector; is the external force; q is the generalized coordinate vector (body pose + joint angles); represents the first order derivative of q with respect to time, i.e. the generalized velocity; represents the second order derivative of q with respect to time, i.e. the generalized acceleration, denotes the desired acceleration; W 1, W 2， W 3 is the weight matrix, which can be dynamically adjusted.

[0178] performing a control step: outputting the joint torque command to the drive unit 201, and outputting the nacelle rotation speed and vector angle command to the air duct unit 3 and the drive unit 201 respectively after inverse solution.

[0179] Equation 2 defines the generalized mathematical problem of the "optimal distribution" step:

[0180] is the constraint condition, which is a simplified writing of the physical model ( represents the joint torque, represents the external force), reiterating that optimization must be carried out within the framework of physical laws.

[0181] is the generalized optimization goal, represents "minimization". That is, the controller needs to find a set of optimal control commands so that the weighted sum of the following three items is minimized:

[0182] : generalized position tracking error, i.e. to make the actual position / attitude of the robot ( ) as close as possible to the desired position / attitude ( );

[0183] : Joint torque loss, i.e. use motor as little as possible (energy saving);

[0184] : Duct speed loss, i.e. use duct fan as little as possible (energy saving);

[0185] Hardware constraints. Make sure all instructions are within the safe working range of the motor and duct.

[0186] In equation 2 is a generalized expression of tracking error. In a more specific, preferred implementation for high-frequency real-time control, the objective is realized as the error of tracking "desired acceleration" and "actual acceleration", and a high-efficiency QP solver (such as OSQP or qpOASES) is used to solve the optimization problem online, with a running frequency no less than 500Hz. At this time, the optimization problem is more detailed in mathematical form as (equation 3):

[0187]

[0188] is the desired motion trajectory (desired acceleration); is the actual acceleration.

[0189] min is the minimum value, max is the maximum value, both are constraint conditions of the above equation.

[0190] W 1, W 2， W 3 is the weight matrix, dynamically adjusted. For example, in the air, the weight of attitude tracking W 1 increases; on the ground, the weight of joint torque loss W 2 increases.

[0191] Equation 3 defines the quadratic programming (QP) problem actually executed by the controller (QP solver):

[0192] In equation 3:

[0193] is the acceleration tracking error. This is a specific implementation of the generalized "tracking error". The controller calculates a "desired acceleration" ( ), and tries to make the "actual acceleration" ( ) strictly track it.

[0194] : Same meaning as equation 2 (minimize energy consumption), but using a more formal mathematical norm writing.

[0195] Constraints: More complete constraints for solving the QP problem are listed here, more detailed than equation 2, including:

[0196] Unified dynamics equation: i.e. equation 1 must hold.

[0197] Hardware saturation.

[0198] Friction cone constraint. This is the core of ground walking: make sure the foot end contact force ( ) is within the friction cone, to prevent the robot from slipping.

[0199] : If the foot end is off the ground, the contact force must be zero.

[0200] Embodiment 3

[0201] Referring to Figure 6 , this is the third embodiment of the present application, which aims to illustrate the specific optimization control method of "ground / flight mode switching", which optimizes the full-drive control method based on the amphibious quadruped robot in embodiment 2, and includes the following steps for the mode transition state:

[0202] Smoothly adjust the priority of foot end contact force constraint and body attitude tracking constraint in the optimization objective through weight coefficient, to realize smooth take-off and landing.

[0203] As Figure 6 shown, the smoothness of the transition is realized by adaptive adjustment of the weight in the optimization objective:

[0204] Take-off: initial state Figure 6 In W _ground (high ground weight). As the command is issued, Figure 6 In W_ aerial (high air weight) gradually increases, and the optimizer automatically transfers the support force from the foot end to the duct. As Figure 6 shown by the curve, during this process, the "total foot end contact force (N)" (cyan solid line) smoothly decreases from about 160 N to 0 N; at the same time, the "total duct thrust (N)" (blue solid line) smoothly increases from 0 N to about 155 N, eventually replacing the foot end contact force as the main support source to overcome gravity. When the foot end contact force is detected to be 0, the system completely enters the flight mode. Landing: the process is reversed. At the same time, increase the constraint on the derivative of the foot end impact force ( F c,z ) to realize soft landing.

[0205] The above technical solution only reflects the preferred technical solution of the present application, and some changes made by the skilled in the art to some parts thereof also reflect the principle of the present application and are within the protection scope of the present application.

Claims

1. A full drive control capability based amphibious quadruped robot system, characterized in that, Comprise: A body (1); Four mechanical legs (2) symmetrically arranged on both sides of the body (1); Four air power duct units (3) respectively connected with the hip joint parts of the corresponding mechanical legs (2); And four drive units (201) mounted on the body (1) respectively corresponding to driving the four mechanical legs (2); Wherein, the driving function of the drive unit (201) is fully reused: In the ground walking mode, the drive unit (201) is used to drive the mechanical leg (2) to realize the multi-degree-of-freedom walking of the robot; In the air flight mode, the drive unit (201) is reused to drive the mechanical leg (2) to adjust the posture, thereby providing at least two degrees of freedom of the thrust vector adjustment capability for the air power duct unit (3) connected therewith; and by the drive unit (201) to the thrust of the four air power duct units (3) Unified vector adjustment, realize the full drive control of the robot full attitude stable flight and six degrees of freedom motion, the full attitude includes side flight, upside down flight and side body hovering; Wherein, the amphibious quadruped robot system is driven by the following full drive control method, the full drive control method comprises the following steps: State sensing step: through the sensing module (5), the attitude, position, speed information of the body (1), the angle, angular velocity information of each mechanical leg (2) joint, the contact force information of the foot end (205) and the ground, and the rotating speed information of each air power duct unit (3) are obtained in real time; Mode decision step: according to the task instruction and the environment sensing information from the sensing module (5), the high-level decision and gait planning module (6) decides whether it is in the ground walking mode, the air flight mode or the mode transition state, and generates the corresponding body expected motion trajectory; Unified dynamics modeling step: a unified full drive dynamics model including the body (1), the mechanical leg (2), the air power duct unit (3) and the vector angle thereof is established; based on the expected motion trajectory and the robot body condition, the required total space force and total space moment are calculated by using the rigid body dynamics model; Optimization allocation step: by the attitude controller (7), taking tracking the expected motion trajectory as the target, based on the unified dynamics model, the optimal joint torque command and the rotating speed and vector angle command of each air power duct unit (3) are solved in real time by using the optimization algorithm with constraints; Execution control step: the joint torque command is output to the drive unit (201), and the duct rotating speed and vector angle command solved by the optimization allocation step are respectively output to the air power duct unit (3) and the drive unit (201) after inverse calculation; Wherein, the optimization problem in the optimization allocation step is mathematically expressed as: wherein is the joint torque vector; is the nacelle rotational speed vector; , is the nacelle pitch and yaw vector angle vector; M is the mass matrix; C is the Coriolis and centrifugal force terms; J is the Jacobian matrix, is the transpose of the Jacobian matrix; F thrust is the nacelle thrust vector; is the external forces; q is the generalized coordinate vector; represents the first order derivative of q with respect to time, i.e. the generalized velocity; represents the second order derivative of q with respect to time, i.e. the generalized acceleration, denotes the desired acceleration; W 1, W 2， W 3 is a weight matrix, which can be dynamically adjusted.

2. The amphibious quadruped robot system of claim 1, wherein, Further comprising: Fixed support (4), the four air power duct units (3) are respectively connected to the hip joint parts of the mechanical legs (2) through the fixed support (4); Sensing module (5) comprising on-board inertial measurement unit, depth camera and laser radar, for sensing the external environment and the robot body condition; a high-level decision and gait planning module (6) configured to receive motion instructions and environment perception information, generate a desired motion trajectory and a sequence of foot placements; a posture controller (7) in communication with the high-level decision and gait planning module, configured to calculate, based on the desired motion trajectory and a robot body condition, a total spatial force and a total spatial moment at a center of mass of the robot body required for tracking the desired motion trajectory using a rigid body dynamics model of the robot, and distribute the total spatial force and the total spatial moment to each foot in contact with the ground to generate a target contact force for each foot. a rotational joint controller (8) in communication with the posture controller, configured to calculate and output a desired signal instruction for driving each driving unit (201) based on the target contact force of each foot and a current configuration of the robot by using a Jacobian transpose matrix. The system controls the driving units (201) by using the desired signal instruction. Each of the four mechanical legs (2) has the same structure, and each mechanical leg (2) further includes a thigh link (202), a driving link (203), a shank link (204), and a foot (205).

3. The amphibious quadruped robot system of claim 1, wherein, The posture controller (7) includes:

4. The amphibious quadruped robot system of claim 2, wherein, a model-based feedforward control unit (701) configured to perform inverse dynamics calculation using the rigid body dynamics model, take the desired motion trajectory and a current state as inputs, and output the total spatial force and the total spatial moment at the center of mass of the robot body required for tracking the desired motion trajectory; a contact force distribution optimization unit (702) configured to take the total spatial force and the total spatial moment as an optimization target, solve the target contact force of each foot by using a quadratic programming algorithm under the conditions of satisfying a friction cone constraint and a no-pull force constraint. The constraint conditions for optimization performed by the contact force distribution optimization unit (702) include:

5. The amphibious quadruped robot system of claim 4, wherein, a normal component of the contact force of each foot is greater than zero; a ratio of a lateral component to the normal component of the contact force of each foot is less than a preset friction coefficient; a sum of the target contact forces of all the feet is equal to the total spatial force; a sum of moments generated by the target contact forces of all the feet about the center of mass of the robot body is equal to the total spatial moment. The rotational joint controller (8) is specifically configured to:

6. The amphibious quadruped robot system of claim 2, wherein, calculate a Jacobian matrix of each leg under a current configuration according to a current joint angle of the robot; introduce a PID controller based on joint angle or torque feedback to generate a compensation torque for correcting model errors and external disturbances; The desired feedforward torques for each joint are calculated using the formula where is the joint torque vector, is the transpose of the Jacobian matrix, and f is the target contact force vector at the foot end. add the desired feedforward torque and the compensation torque to obtain a final desired torque instruction. The high-level decision and gait planning module (6) is further configured to perform real-time online foot placement adjustment according to the environment perception information to avoid obstacles or adapt to uneven terrain.

7. The amphibious quadruped robot system of claim 2, wherein, The mode transition state adjusts a priority of a foot contact force constraint and a robot body posture tracking constraint in an optimization target by using a weight coefficient, and realizes smooth takeoff and landing.

8. The amphibious quadruped robot system of claim 1, wherein, ​

Images