Three-dimensional wall surface adaptive cross-wall transition and gait control method for bionic foot climbing robot

Abstract

The application provides a three-dimensional wall surface adaptive cross-wall transition and gait control method for a bionic foot climbing robot, and belongs to the field of robot control. In order to solve the problems of insufficient mobility and stability, difficult autonomous transition across the wall surface and weak adaptability to complex surfaces of the existing adsorption type foot climbing robot when moving on a three-dimensional wall surface, the application constructs a stability constraint criterion based on three-point adsorption, ensures that at least three foot ends are reliably adsorbed and the support area contains the gravity center projection throughout the process; adopts a two-stage movement mechanism to decouple continuous movement into the alternating cycle of foot end displacement and body translation; realizes forward gait, self-rotation gait and wall transfer gait, so that the robot can move flexibly on a single wall surface and can autonomously and smoothly transition between different wall surfaces in a three-dimensional space; integrates closed-loop control based on foot end force sense and body posture sensing to monitor the adsorption state in real time and compensate for posture deviation, thereby significantly improving the movement stability and environmental adaptability.

Description

Technical Field

[0001] This invention relates to the field of robot control technology, and more specifically, to a three-dimensional wall-adaptive cross-wall transition and gait control method for a biomimetic legged climbing robot. Background Technology

[0002] With the rapid development of science and technology in my country, the robotics industry has grown rapidly in recent years. Climbing robots, capable of clinging to walls, have demonstrated significant application value in areas such as high-risk environment inspection and cleaning of large facilities (e.g., photovoltaic panels, wind turbine blades, and nuclear power plant containment structures). Among them, legged climbing robots have attracted much attention due to their discrete footholds and strong adaptability to unstructured surfaces.

[0003] However, existing legged climbing robots face significant challenges when moving in three-dimensional space. First, when moving on a single wall (such as a vertical wall), the trade-off between stability and efficiency must be resolved. Simple alternating gaits may lead to instability due to insufficient adhesion points, while overly conservative gaits result in sluggish movement. Second, in complex three-dimensional environments, such as transitioning from the ground to a wall, from a wall to a ceiling, or moving between walls at different angles, the climbing robot needs to coordinate its multiple degrees of freedom to plan complex gaits that satisfy mechanical stability constraints while achieving smooth transitions. Furthermore, due to the uncertainty of the material and flatness of the working surface, purely open-loop gait control is often unreliable, necessitating the introduction of proprioception for closed-loop adjustment.

[0004] Currently, research on gait control methods for three-dimensional wall transitions of legged climbing robots is insufficient. Existing solutions are either limited to planar motion or have complex transition gaits and low reliability. There is a lack of a systematic control method that takes into account stability, efficiency and adaptability. Summary of the Invention

[0005] The technical problem to be solved by this invention is:

[0006] To address the issues of insufficient mobility and stability of existing adhesive-type legged climbing robots when moving across three-dimensional walls, difficulty in autonomous transition across walls, and weak adaptability to complex surfaces.

[0007] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows:

[0008] This invention provides a three-dimensional wall-adaptive cross-wall transition and gait control method for a biomimetic legged climbing robot, comprising the following steps:

[0009] S100. Construct a kinematic constraint model based on the three-point adsorption stability criterion to ensure that the bionic legged climbing robot always has at least three foot ends that can reliably adsorb onto the wall to form adsorption points during the movement process, and the support area formed by the three adsorption points includes the projection of the robot's center of gravity onto the wall.

[0010] S200 employs a two-stage motion mechanism for gait planning and execution. The first stage is the foot displacement stage, which controls the suction cups of the robot's legs to detach from the wall and move along a preset trajectory to the target pose before re-attaching, while maintaining the body pose. The second stage is the body translation stage, which controls all legs to remain in the attached state and drives the body to move to the target pose.

[0011] S300. Based on the task requirements, and using the motion constraint model constructed in step S100 and the two-stage motion mechanism in step S200, plan and execute at least one of the following: forward gait, self-rotation gait, and wall transfer gait, so as to realize the robot's movement on a single wall or adaptive transition between different walls in three-dimensional space.

[0012] Furthermore, the biomimetic legged climbing robot includes at least three legs, with suction cups disposed at the ends of each leg, and a thin-film pressure sensor disposed inside each leg.

[0013] Furthermore, the biomimetic legged climbing robot includes four legs.

[0014] Furthermore, in step S100, the DH parameter method is used to model a single leg. Based on the transformation matrix of a single joint, the accumulated foot pose is:

[0015]

[0016] in, 0 P4 refers to the position of the foot at point P. i-1 T i It is obtained from the following calculation formula:

[0017]

[0018] in, This indicates the length of the connecting rod, which is the distance between the common perpendiculars of the two joint axes; It represents the torsion angle of the link, that is, the rotation angle about the common perpendicular from axis i-1 to axis i; This represents the link offset, which is the distance along joint axis i from the intersection of the common perpendicular and the axis to the next common perpendicular. This represents the joint angle, which is the rotation angle around the joint axis i from link i-1 to link i.

[0019] Starting with one leg, proceed sequentially... 01 T1 to3 The product of T4 finally yields the pose of foot P;

[0020] Inverse kinematics is solved iteratively using a custom-damped least squares method. The angles of the four joints are inferred from the target pose at the foot end, and the position and attitude error functions are constructed and combined into the total error. ;

[0021]

[0022] in, These are the weighting coefficients. For ankle joint angle, These represent the target angle and current angle of the ankle joint, respectively; x, y, and z represent the three coordinate components in the base coordinate system of each leg, curr is the current position, and tar is the target position;

[0023] A Jacobian matrix is ​​constructed based on numerical differentiation to describe the influence of joint angles on foot position. The joint angles are then iteratively corrected, including...

[0024] Damping least squares joint correction for:

[0025]

[0026] in, for Jacobian matrix The damping coefficient is... for identity matrix , These are the target at the foot and the current position, respectively.

[0027] The joint angles have been updated to:

[0028]

[0029] in, For the updated joint angles, For the current joint angle, This represents the change in joint angle.

[0030] Until the total error satisfies the condition that the position error is less than 0.1 mm and the posture error is less than 0.1°, the final joint angle is stored in the joint parameter array of the robot model to drive the foot movement.

[0031] Furthermore, in step S100, the mathematical criterion for stability is based on the three-point adsorption criterion, assuming... Let be the projection point of the fuselage's center of gravity onto the supporting wall. The coordinates of the three adsorption feet projected onto the wall surface are shown; these coordinates are calculated in real-time during planning. The supporting triangle formed is determined, and the following criteria are followed: Whether it is located inside the supporting triangle, if If the robot is located inside the supporting triangle, it is considered stable. If the robot is located outside the supporting triangle, it is determined to be unstable.

[0032] Further, in step S200,

[0033] S210, During the foot displacement phase, including,

[0034] S211. Command Issuance and Adhesion Release: The main controller inside the bionic legged climbing robot loads gait parameters to verify whether the bionic legged climbing robot is in the foot displacement completion state. If there are still feet that have not completed the displacement within the gait cycle, the controller controls the bionic legged climbing robot to delay and wait for that foot. At the same time, it reads the data of the thin-film pressure sensor integrated in that foot and confirms that the pressure value has dropped below the preset detachment threshold. If the pressure value drops below the threshold, it is determined that the adhesion has been reliably released, and the process proceeds to step S212. If the pressure does not drop to the threshold, the abnormal handling process is initiated.

[0035] S212, Trajectory Generation and Joint Driving: Based on the current foot position and target position, a smooth spatial trajectory including lifting, horizontal movement and landing is planned using a cubic polynomial interpolation method, ensuring that the velocity and acceleration of the trajectory are zero at the start and end points; The main controller calculates the trajectory into the angle sequence of the corresponding four joints according to the inverse kinematics solution in step S100, and drives the servo motor to execute.

[0036] S213. Foot placement and adsorption confirmation: After the foot moves to the target position, the main controller sends a command to the foot and reads the data from the membrane pressure sensor again to confirm that the pressure value has risen above the preset reliable adsorption threshold and remains stable, thus determining that adsorption has been reliably established.

[0037] Repeat steps S211-S213 until all feet have moved, then proceed to the next stage;

[0038] S220, during the fuselage relocation phase, including,

[0039] S221, Target pose meter, which calculates the next target pose that the computer needs to move to according to gait requirements;

[0040] S222, Joint Angle Solution and Driving: In this stage, the positions of all foot ends in the world coordinate system are kept constant, using kinematic relationships... ,in Represents the position vector of the foot in the world coordinate system. It's about yaw angle. The rotation matrix, It is the position vector of the foot in the fuselage coordinate system. It represents the relative position of the origin of the fuselage coordinate system in the world coordinate system; the inverse solution yields the new positions of each foot in the fuselage coordinate system under the new fuselage pose. Subsequently, inverse kinematics is used again to solve and drive the fuselage to move to the target pose by the sequence of joint angle changes, and the main controller coordinates the servos to execute this sequence synchronously.

[0041] S223, Phase complete. After all joints have moved into position, the fuselage migration phase is complete; the system status is updated, awaiting the start of the next gait cycle.

[0042] Further, in step S300,

[0043] S310, For the forward-moving state, it is used for efficient linear movement within a single wall surface, including a cycle of sequential single-leg shifting-intermittent fuselage following;

[0044] S320. For its own rotational gait, a strategy of single-leg fixed axis - multi-leg arc shift - step-by-step rotation is adopted to realize the turning of the bionic legged climbing robot around the foot of one of its own legs.

[0045] Using the kinematic constraint model in step S100, the maximum rotation angle that the robot can achieve in a single rotation is pre-calculated under the constraint of fixed axis foot end and limit constraints of other foot joint angles. If the target total rotation angle is greater than the maximum rotation angle in a single rotation, it is decomposed into multiple steps, each step rotating by the maximum rotation angle in a single rotation, and the last step rotating by an angle smaller than the maximum rotation angle in a single rotation to complete the entire rotation action. Each step includes moving the foot end arc and rotating the body around the axis foot end by the maximum rotation angle in a single rotation.

[0046] S330. The wall-to-wall gait for ground A and wall B includes,

[0047] S331, Symmetrical Distribution Stage

[0048] First, determine the desired configuration when the robot is fully positioned on wall B, and then deduce the ideal footholds for its two front legs on wall B. Before the transfer begins, the robot remains on ground A; by controlling the leg joints, the position of the two hind legs on ground A is adjusted so that their projection points... and The two planes are geometrically symmetrical about the intersection line; the two front legs are transferred to the target point B on the wall through the foot-end displacement stage in sequence; at this time, a stable cross-wall support is formed with the two feet on the wall B and the two feet on the ground A;

[0049] S332, during the center of gravity transfer phase, the fuselage changes from parallel to the ground A to parallel to the wall B; through inverse kinematics numerical solution, a continuous sequence of 16 joint angles changing with time is obtained, driving the fuselage to complete a smooth rotation;

[0050] S333, Leg retraction phase: After the body rotation is completed, the body configuration and foot position are symmetrical with respect to the body configuration and position before the rotation about the intersection line O of ground A and wall B. The two front legs located on wall B are extended forward through the foot displacement phase until they are symmetrical with respect to the initial position of the two hind legs on ground A about the intersection line O. Then the remaining two hind legs are transferred from ground A to the pre-calculated symmetrical position on wall B, finally completing the stable four-legged standing of the entire bionic legged climbing robot on wall B.

[0051] Compared with the prior art, the beneficial effects of the present invention are:

[0052] (1) Significantly improved motion stability and safety. By enforcing the "three-point adsorption" geometric static criterion, a rigid safety boundary is provided for all movements of the robot on non-horizontal surfaces such as vertical walls and ceilings, improving the utilization rate of foot adsorption force and fundamentally avoiding the risk of tipping or slipping of multi-legged robots due to insufficient support points or center of gravity shift. Combined with a force-sensing-based real-time monitoring and protection mechanism for adsorption status, the reliability of adsorption during movement is further ensured.

[0053] (2) Enhanced three-dimensional spatial mobility and transition capabilities. Based on the two-stage motion mechanism planning of "forward movement", "self-rotation" and "wall transfer" gaits, a complete set of motion primitives has been formed, which enables the robot to not only move flexibly and accurately in a single wall, but also to autonomously and smoothly complete complex transitions between different spatial planes (such as ground-wall-ceiling), greatly expanding its autonomous workspace.

[0054] (3) The complexity and reliability of the control system are optimized. The two-stage decoupling mechanism of "foot displacement" and "fuselage migration" decomposes the high-degree-of-freedom continuous spatial motion into a series of standardized sub-tasks executed in sequence, which greatly reduces the difficulty of real-time planning and coordinated control and improves the predictability and robustness of system execution.

[0055] (4) Breakthroughs in environmental adaptability and operational robustness. By integrating foot force sensing and body posture perception into a closed-loop control module, the robot can perceive its own state and external interactions (such as changes in adhesion force and posture disturbances) in real time, and actively adjust its joints and compensate for stability. This enables the invention to effectively cope with uncertainties in real-world environments such as changes in wall roughness and slight unevenness, and enhances its practical value in unstructured scenarios.

[0056] (5) It provides a universal mobile platform foundation for multifunctional operations. The high stability and high mobility achieved by this invention make the robot chassis almost without any up-and-down movement during the movement. It can serve as a reliable platform and, by integrating different functional modules (such as cleaning, inspection, and spraying devices), can be widely used in various high-risk or extreme operation scenarios such as high-altitude building maintenance, large industrial facility inspection, and disaster site reconnaissance. Attached Figure Description

[0057] Figure 1 This is a schematic diagram of the hardware and software architecture of the biomimetic legged climbing robot in an embodiment of the present invention;

[0058] Figure 2 This is a schematic diagram illustrating the principle of the three-point adsorption stability criterion in an embodiment of the present invention;

[0059] Figure 3 This is a flowchart illustrating the two-stage motion mechanism of the foot displacement stage and the fuselage migration stage in an embodiment of the present invention.

[0060] Figure 4 This is a schematic diagram of the kinematic model of a single leg of the biomimetic legged climbing robot and the definition of the DH parameter coordinate system in an embodiment of the present invention;

[0061] Figure 5 This is a schematic diagram of the motion sequence of the forward progress state planned based on the kinematic model and two-stage mechanism within a complete cycle in an embodiment of the present invention;

[0062] Figure 6 This is a schematic diagram of a self-rotational gait sequence planned based on a kinematic model and a two-stage mechanism in an embodiment of the present invention;

[0063] Figure 7 This is a schematic diagram of a wall transfer gait sequence planned based on a kinematic model and a two-stage mechanism in an embodiment of the present invention;

[0064] Figure 8 This is a structural diagram of the biomimetic legged climbing robot in an embodiment of the present invention;

[0065] Figure 9 This is a schematic diagram of a biomimetic legged climbing robot equipped with an extension component in an embodiment of the present invention;

[0066] Figure 10 This is a cross-sectional structural diagram of the foot end component of the biomimetic legged climbing robot in an embodiment of the present invention;

[0067] Figure 11 This is a schematic diagram illustrating the principle and connection of the airflow adsorption system of the biomimetic legged climbing robot in an embodiment of the present invention.

[0068] Figure 12This is an architecture diagram of the hardware control system for the biomimetic legged climbing robot in an embodiment of the present invention;

[0069] Figure 13 This is a schematic diagram of the interface of a web-based robot remote telemetry and teleoperation platform in an embodiment of the present invention.

[0070] Explanation of reference numerals in the attached figures:

[0071] 1. Main body; 2. Quadrupedal mechanical structure; 3. Negative pressure adsorption system; 4. Body sensing unit; 5. Expansion module; 101. Upper plate of the body; 102. Lower plate of the body; 103. Expansion module interface; 201. Thigh component; 202. Lower leg component A; 203. Lower leg component B; 204. Foot end component; 205. Thin-film pressure sensor; 301. Vacuum pump; 302. Busbar; 303. Solenoid valve; 304. Suction cup; 401. Main controller; 402. Power management module; 403. Battery; 404. IMU; 405. STM32 status acquisition board; 406. Camera; 501. DC motor; 502. Expansion module main controller; 503. Limit switch; 504. Cleaning brush; 505. Cleaning sponge; 506. Main body of cleaning module. Detailed Implementation

[0072] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0073] Specific Implementation Plan 1: Combining Figure 4 , 8 As shown in Figure -10, a biomimetic legged climbing robot is presented. This robot is a quadrupedal, negative pressure suction wall-climbing robot with redundant degrees of freedom; each leg has four rotational degrees of freedom. The biomimetic legged climbing robot includes a body and legs, specifically a main body 1, a quadrupedal mechanical structure 2, a negative pressure suction system 3, a body sensing unit 4, and an expansion module 5. The main body 1 includes a parallel upper plate 101 and a lower plate 102. An expansion module interface 103 (for connecting to the expansion module 5) is located below the lower plate 102. Figure 9 As shown, the expansion module interface 103 is connected to the cleaning module body 506. A negative pressure adsorption system 3 is provided between the upper plate 101 and the lower plate 102 of the robot body. A four-legged mechanical structure 2 is provided at each of the four corners of the connection between the upper plate 101 and the lower plate 102 of the robot body. The four-legged mechanical structure 2 is the leg of the negative pressure adsorption wall-climbing robot. A body sensing unit 4 is provided on the upper plate 101.

[0074] Each quadrupedal mechanical structure 2 includes a thigh component 201, a lower leg component A 202, a lower leg component B 203, and a foot end component 204 connected in sequence. The foot end component 204 is equipped with a thin film pressure sensor 205. The thigh component 201, lower leg component A 202, lower leg component B 203, and foot end component 204 can rotate 270° relative to each other. The connection between the thigh component 201 and the upper plate 101 and the lower plate 102 of the machine body can also rotate 270°.

[0075] Combination Figure 11 As shown, the negative pressure adsorption system 3 includes a vacuum pump 301, a manifold 302, a solenoid valve 303, and a suction cup 304. The suction cup 304 is located at the end of each foot component 204 and is used to ensure that the bionic legged climbing robot can crawl by negative pressure adsorption in cooperation with the quadrupedal mechanical structure 2. The opening and closing of the air pump 301 and the four solenoid valves 303 are controlled by the GPIO signal of the main controller 401 in the body sensing unit 4. The air pump 301 is connected to the four solenoid valves 303 through a manifold 302. When the control signal K controls the solenoid valve 303 to open, P and A are connected, that is, the suction cup 304 is connected to the air pump 301, and a negative pressure is drawn between the suction cup 304 and the wall to achieve stable adsorption. When the control signal K controls the solenoid valve to close, A and O are connected, that is, the suction cup 304 is connected to the atmosphere, and the negative pressure is released instantly to achieve adsorption release.

[0076] Combination Figure 12As shown, the body perception unit 4 includes a main controller 401, a power management module 402, a battery 403, an IMU 404, an STM32 status acquisition board 405, and a camera 406. The battery 403 powers the entire bionic legged climbing robot through the power management module 402. The STM32 status acquisition board 405 acquires attitude information and force distribution information of each foot provided by the IMU 404. The camera 406 acquires image or video information. The main controller 401 is a Raspberry Pi 4B. The IMU 404 is a body inertial measurement unit. The battery 403 is a 12V lithium battery connected to the power management module 402. The entire system includes a vacuum pump 301, a solenoid valve 303, an expansion module main controller 502, a DC motor 501, a thigh component 201, a lower leg component A 202, a lower leg component B 203, and servos and their drive boards (PCA9685 PWM drivers) in the foot component 204. The power supply is provided by the battery 403. The current output of the battery 403 is acquired by the ACS712 chip in the 405STM32 status acquisition board. At the same time, the 405STM32 status acquisition board acquires and encodes the pressure information of the thin film pressure sensor 205, and sends it to the main controller 401 (Raspberry Pi 4B) via serial port. The IMU 404 and camera 406 are directly connected to the main controller 401, and send the calculated posture information and the video stream acquired by the camera 406 to the main controller 401. The main controller 401 controls the start, stop and speed of the DC motor 501 through GPIO, and senses whether the cleaning brush 504 and cleaning sponge 505 have reached the target position through the limit switch 503 (when the cleaning brush 504 and cleaning sponge 505 reach the target position (i.e., the limit point), the limit switch will be triggered); at the same time, the main controller 401 can also obtain user commands from the Web platform and synchronously upload the robot status information to the Web platform.

[0077] Combination Figure 9 As shown, the expansion module 5 includes a DC motor 501, an expansion module main controller 502, a limit switch 503, a cleaning brush 504, a cleaning sponge 505, and a cleaning module body 506. That is, the bionic legged climbing robot of the present invention is a cleaning robot. The cleaning module body 506 is connected to the expansion module interface 103. The cleaning sponge 505 and the cleaning brush 504 can move cyclically under the drive of the DC motor 501. A limit switch 503 is provided at the limit position. When the limit switch is triggered, the DC motor 501 drives the cleaning sponge 505 and the cleaning brush 504 to change direction. The expansion module main controller 502 obtains the motor drive signal from the autonomous controller 401, amplifies the power, and drives the DC motor 501 to move according to the motor drive signal from the main controller 401.

[0078] It also includes a motion decoupling and gait planning module, an inverse kinematics solving module, and a closed-loop control module. The motion decoupling and gait planning module is responsible for receiving pre-loaded actions from the local control terminal and user instructions from the remote monitoring platform, generating gait motion plans and providing them to the inverse kinematics solving module for inverse kinematics solving. After constraint and verification, the generated motion sequence is output to the closed-loop control module. The closed-loop control module controls the 16 joints of the quadrupedal mechanical structure 2 through feedback of the body state information; at the same time, it controls the air passage opening and closing of the negative pressure adsorption system 3 according to the timing requirements.

[0079] The above control algorithm runs as a modular Python program on the main controller 401. It processes the raw sensor data from the body sensing unit 4 through the state acquisition unit running on the STM32 state acquisition board 405. The processed body state information is packaged and sent to the main controller 401 through serial communication, and copied to the local control terminal and the remote monitoring platform.

[0080] If extension module 5, such as the cleaning function, is loaded, the closed-loop control module is also responsible for controlling extension module 5; the local control terminal interface provides real-time monitoring and debugging functions; simultaneously, a Flask-based web-based remote monitoring platform is built, utilizing WebSocket to achieve bidirectional low-latency communication, presenting the robot's multimodal data, including video streams, 3D poses, and sensor curves, to the remote operator in real time, and issuing advanced task commands, such as performing wall transfer and path-based cleaning, verifying the method's ability to support teleoperation and semi-autonomous operation; specifically,

[0081] Combination Figure 9 As shown, the expansion module 5, which implements the cleaning function, is an independent functional unit. Its start / stop and speed can be controlled by the GPIO and PWM interfaces of the main controller 401. When performing a cleaning task, the main controller 401 decomposes the cleaning area into a series of coverage paths consisting of forward walking states and its own rotational gait. When the control system performs a straight-line movement segment, it automatically starts the DC motor 501 of the cleaning module in conjunction with the water supply system (not installed on the robot) to perform cleaning. Before turning or performing a wall transfer gait, the cleaning function is paused. This collaborative planning of tasks and movements ensures work efficiency and system safety, fully demonstrating the practical value of this invention as a general-purpose mobile work platform.

[0082] Specific Implementation Plan Two: Combining Figures 1 to 12 As shown, this invention provides a three-dimensional wall-adaptive cross-wall transition and gait control method for a biomimetic legged climbing robot, comprising the following steps:

[0083] S100. Construct a motion constraint model based on the three-point adsorption stability criterion to ensure that at least three adsorption actuators are reliably adsorbed onto the wall during the robot's movement, and that the support area formed by the three adsorption points includes the projection of the robot's center of gravity onto the wall.

[0084] Specifically, including,

[0085] Combination Figure 4 As shown, in order to achieve precise control, a kinematic constraint model of the robot is established; the Denavit-Hartenberg (DH) parametric method is used to model a single leg, and the coordinate system and parameter definitions are shown in Table 1. This kinematic constraint model is the basis for subsequent foot trajectory planning and inverse kinematics solution.

[0086] Table 1 Single Leg DH Parameter Table

[0087]

[0088] in, This indicates the length of the connecting rod, which is the distance between the common perpendiculars of the two joint axes; It represents the torsion angle of the link, that is, the rotation angle about the common perpendicular from axis i-1 to axis i; This represents the link offset, which is the distance along joint axis i from the intersection of the common perpendicular and the axis to the next common perpendicular. This represents the joint angle, which is the rotation angle around the joint axis i from link i-1 to link i.

[0089] The formula for calculating foot pose based on the transformation matrix of a single joint is as follows:

[0090]

[0091] in, 0 P4 refers to the position of the foot at point P. i-1 T i It is obtained from the following calculation formula:

[0092]

[0093] Starting from the O-XYZ coordinates of a single leg, the corresponding base coordinate system directions are: Figure 4 middle Along the axis, the pose of foot P is finally obtained by successively multiplying the four matrices T1 to T4 (the matrices include rotation and translation operations); in the entire robot's body coordinate system, the pose of foot P needs to be determined by the mounting matrix of the leg and transformed into the body coordinate system.

[0094] Inverse kinematics is solved iteratively using a custom damped least squares method. The angles of the four joints are inferred from the foot target pose, and the position and attitude error functions are constructed and combined into the total error. ;

[0095]

[0096] in, Weighting coefficients ( ), Ankle joint angle (characterizing posture) The target angle and current angle of the ankle joint are respectively; x, y, and z represent the three coordinate components in the base coordinate system of each leg, curr is the current position, and tar is the target position. The error is constructed in the form of a weighted sum of squares. ;

[0097] A Jacobian matrix is ​​constructed based on numerical differentiation to describe the influence of joint angles on foot position. The joint angles are then iteratively corrected.

[0098] Damping least squares joint correction for:

[0099]

[0100] in, for Jacobian matrix The damping coefficient is... for identity matrix , These are the target and the current position, respectively.

[0101] The joint angles have been updated to:

[0102]

[0103] in, For the updated joint angles, For the current joint angle, This represents the change in joint angle.

[0104] Until the total error satisfies the condition that the position error is less than 0.1 mm and the posture error is less than 0.1°, the final joint angle is stored in the joint parameter array of the robot model to drive the foot movement.

[0105] Combination Figure 2 As shown, the mathematical criterion for stability is based on the three-point adsorption criterion; let... Let be the projection point of the fuselage's center of gravity onto the supporting wall. The coordinates of the three adsorption feet projected onto the wall surface are given; the system calculates these coordinates in real time during planning. The supporting triangle formed is determined, and the following criteria are followed: Whether the gait sequence is located inside the triangle is a prerequisite for its safe execution.

[0106] The S200 employs a two-stage motion mechanism for gait planning and execution. The first stage is the foot displacement stage: the suction actuator of the target leg is controlled to detach from the wall and move along a preset trajectory to the target pose before re-attaching, while maintaining the body pose. The second stage is the body translation stage: all legs are controlled to remain in the suction state, and the body is driven to move to the target pose.

[0107] Specifically, including,

[0108] Combination Figure 3 As shown, all gaits follow an alternating cycle of foot movement phase and fuselage movement phase;

[0109] S210, During the foot displacement phase, including,

[0110] S211, Command Issuance and Adhesion Release: The main controller 401 loads gait parameters to verify whether the bionic legged climbing robot is in the foot displacement completion state. If there are still feet that have not completed the displacement within the gait cycle, a closing command is sent to the solenoid valve 303 corresponding to the target leg, and a delay is performed. At the same time, the data of the thin-film pressure sensor 205 integrated in the foot is read to confirm that the pressure value has dropped below the preset release threshold. It is determined that the adhesion has been reliably released and proceeds to step S212. If the pressure has not dropped to the threshold, the abnormal handling process is entered, that is, the subsequent foot displacement action is temporarily stopped, the staff is reminded to check in time, and after the pressure drops normally to below the threshold, the abnormal state is exited and the subsequent actions continue.

[0111] S212, Trajectory Generation and Joint Driving: Based on the current foot position and target position, a smooth spatial trajectory including lifting, horizontal movement and descent is planned within a certain time interval using a cubic polynomial interpolation method, ensuring that the velocity and acceleration of the trajectory are zero at the start and end points to reduce impact; the main controller 401 solves the trajectory into the angle sequence of the corresponding four joints according to the inverse kinematics solution, and drives the servo motor to execute through PWM signals;

[0112] S213, Foot placement and adsorption confirmation: After the foot moves to the target position, the main controller 401 sends a command to open the corresponding solenoid valve 303, reads the data of the membrane pressure sensor 205 again, confirms that the pressure value rises to above the preset reliable adsorption threshold and remains stable, and determines that adsorption has been reliably established.

[0113] Repeat steps S211-S213 until all feet have moved, then proceed to the next stage. During the foot movement stage, the body posture is locked and remains unchanged, and only the foot position changes. This has the advantage of making the body more stable and providing good stability for cleaning and other working modes.

[0114] S220, during the fuselage relocation phase, including,

[0115] S221. Target pose calculation: Based on gait requirements, calculate the next target pose that the computer needs to move to.

[0116] S222, Joint Angle Calculation and Driving: In this stage, the positions of all foot ends in the world coordinate system are constrained to remain unchanged, and kinematic relationships are used to solve the problem. ,in Represents the position vector of the foot in the world coordinate system. It's about yaw angle. The rotation matrix rotates the "vector in the fuselage coordinate system" to the "direction in the world coordinate system"; It is the position vector of the foot in the fuselage coordinate system. It represents the relative position of the origin of the body coordinate system in the world coordinate system; the inverse solution yields the new position of each foot in the body coordinate system (the origin of the coordinate system is at the geometric center of the robot, the height of the origin coincides with the plane formed by the origins of the base coordinate systems of the four feet, the x-axis is along the camera's viewpoint, the z-axis points upwards, and the y-axis is determined by the right-hand rule) under the new body pose. Subsequently, inverse kinematics is used again to solve and drive the fuselage to move to the target pose, and the sequence of angle changes of all 16 joints required is used. The main controller 401 coordinates all servos to execute this sequence synchronously.

[0117] S223, Phase Complete: After all joints have moved into position, the fuselage migration phase is complete; the system status is updated, awaiting the start of the next gait cycle;

[0118] S300. Based on the task requirements, and using the motion constraint model constructed in step S100 and the two-stage motion mechanism in step S200, plan and execute at least one of the following: forward gait, self-rotation gait, and wall transfer gait, so as to realize the robot's movement on a single wall or adaptive transition between different walls in three-dimensional space.

[0119] Specifically, including,

[0120] S310, for the forward-moving state, it is used for efficient linear movement within a single wall surface. Its core is the cyclical pattern of sequential single-leg shifting followed by intermittent fuselage following; combined with Figure 5Figures (a) to (l) in the diagram show the various stages of the forward movement in chronological order. The horizontal axis is the x-axis, the vertical axis is the y-axis, and the ordinate is the N-axis. The x-axis has five discrete coordinate points: -10, -5, 0, 5, and 10; the y-axis has seven discrete coordinate points: -6, -4, -2, 0, 2, 4, and 6; and the N-axis has four discrete coordinate points: -2, 0, 2, and 4. Throughout the entire movement, the fuselage height remains constant, with only the foot position changing. The fuselage remains stable, demonstrating good functional adaptability and stability for working modes requiring a fixed fuselage height, such as cleaning.

[0121] S311, Foot displacement stage: Figure 5 (a) through (j) sequentially demonstrate the foot displacement phases, ensuring that at least three feet remain stably attached during each movement, with one leg being transferred at a time, such as... Figure 5 (a) to (c) show the movement of the left foreleg (corresponding to the start, middle, and end, respectively). After the four legs are transferred in this way, the foot displacement phase ends.

[0122] S312, Fuselage migration stage: The horizontal displacement vector of the single-leg shift and the subsequent fuselage follow-up displacement vector satisfy that the single-leg shift is greater than the subsequent fuselage follow-up, and the directions form a certain angle, which is used to create space for fuselage translation and avoid motion interference; Figure 5 (j) to (l) in the diagram illustrate the process of the fuselage translation phase (corresponding to the start, middle, and end, respectively).

[0123] S320. For self-rotational gait, used to achieve precise turning of the biomimetic legged climbing robot around the foot of one of its own legs, a strategy of single-leg fixed axis - multi-leg arc shift - step-by-step rotation is adopted; usually, the forefoot or hindfoot located on the same side as the rotation direction is selected as the pivot foot, combined with... Figure 10 As shown, the foot end component 204 is integrally formed based on a nested 3D printing method; a 1mm air gap is left between the suction cup 304 and the foot end component 204, which allows relative rotation between the suction cup 304 and the foot end component 204; this ensures that the suction cup 304 at the foot end remains stably attached during the rotation around the axis foot; at the same time, a thin film pressure sensor 205 is embedded during the printing process;

[0124] Using the kinematic constraint model in step S100, the maximum rotation angle that the bionic legged climbing robot can achieve in a single operation is pre-calculated under the constraint of fixed axis foot end and other foot joint angle limits. For example, in this embodiment, the maximum rotation angle in a single operation is calculated to be 21°. If the target total rotation angle is greater than the above maximum rotation angle in a single operation, it is decomposed into multiple steps. Each step rotates the aforementioned maximum rotation angle in a single operation, and the last step rotates an angle smaller than the maximum rotation angle in a single operation to complete the entire rotation action.

[0125] Each step involves: the moving foot sequentially completing a certain arc of displacement, followed by the fuselage rotating around the axis of the foot at a single maximum angle; this mechanism ensures that the rotation process remains within kinematic and stability constraints, combined with... Figure 6 Figures (a) to (l) in the diagram show the various stages of the forward movement in chronological order. The horizontal axis is the x-axis, the vertical axis is the y-axis, and the vertical axis is the N-axis. The x-axis has five discrete coordinate points: -15, -10, -5, 0, and 5; the y-axis has four discrete coordinate points: -15, -10, -5, and 0; and the N-axis has four discrete coordinate points: -0, 2, 4, and 6. The fuselage height remains constant throughout the entire movement process, with only the position of the feet changing.

[0126] S321, Foot displacement stage: combined with Figure 6 (a) to (i) in the diagram illustrate the foot displacement stages in chronological order. Each movement process ensures that at least three feet remain stably attached, with one leg being transferred at a time. Using the left foreleg foot as a fixed axis of rotation, the fuselage and the remaining legs follow a circular trajectory centered on this axis, forming a segment of an arc in space. Figure 6 (a) to (c) in the figure show the movement process of the right foreleg (corresponding to the start, middle and end respectively). After the right foreleg, right hind leg and left hind leg are transferred in sequence, the foot displacement phase ends.

[0127] S322, Fuselage relocation phase: combined with Figure 6 (i) to (l) in the diagram illustrate the process of the fuselage translation phase (corresponding to the start, two intermediate phases, and the end, respectively); the fuselage height remains constant throughout the entire fuselage rotation process;

[0128] S330. For wall-transfer gait, this is the core complex gait for achieving three-dimensional spatial crossing, requiring the vehicle to maintain a certain height from both walls during the transfer to allow space for other loads; combined with Figure 7 As shown, taking the transition from the ground (denoted as plane A) to a vertical wall (denoted as plane B) as an example, its planning and execution are decomposed into three stages. The horizontal axis is the x-axis, the vertical axis is the y-axis, and the vertical axis is the N-axis. The x-axis has five discrete coordinate points: -15, -10, -5, 0, and 5; the y-axis has three discrete coordinate points: -5, 0, and 5; and the N-axis has four discrete coordinate points: 0, 5, 10, and 15.

[0129] S331, Symmetrical Distribution Stage

[0130] First, determine the desired configuration of the bionic legged climbing robot when it is fully positioned on wall B, and then deduce the ideal footholds of its two front legs on surface B. ;

[0131] Before the transfer began, the bionic legged climbing robot remained on ground A; combined with Figure 7As shown in (a) to (c), by controlling the joints, the positions of the two hind legs on the ground A are adjusted so that their projection points... and The intersection of the two planes is geometrically symmetrical; this establishes a mechanically balanced state of center of gravity shift and support.

[0132] Combination Figure 7 As shown in (d) to (f), the two forelegs are sequentially transferred to the target point on surface B through a foot-end displacement stage; at this point, a stable cross-wall support is formed with both feet on surface B and both feet on surface A, as shown in (d) to (f). Figure 7 (f) in the middle;

[0133] S332, the phase of shifting the center of gravity.

[0134] This stage is a special fuselage translation stage, with the goal of transforming the fuselage from parallel to plane A to parallel to plane B. This requires the fuselage to simultaneously rotate around an axis parallel to the intersection line O (the line where planes A and B intersect) and translate closer to plane B. This planning method ensures that the height above the two walls during the transfer is sufficient to install other expansion modules, such as cleaning modules, enabling load-bearing transfers. Since the positions of the four legs are locked in the world coordinate system, the aforementioned fuselage pose transformation is converted into complex coordinated movements of the leg joints. Through inverse kinematics numerical solutions, a continuous sequence of 16 joint angles changing over time is obtained, driving the fuselage to smoothly complete the rotation. Figure 7 (f) to (h) in the diagram illustrate this transition process, through the careful design of the symmetrical distribution stages, combined with Figure 7 The poses of (f) to (h) are completely symmetrical about the intersection line O. Therefore, it is not difficult to know from the principle of reversible motion that this embodiment can allow the bionic legged climbing robot to move freely back and forth between surface A and surface B.

[0135] S333, Leg retraction phase.

[0136] After the fuselage rotation is completed, the fuselage configuration and foot position are symmetrical about the intersection line O to the fuselage configuration and position before the rotation. The two front legs, located on plane B, are then extended forward through a foot-position shifting phase until they are symmetrical about the intersection line O to the initial positions of the two hind legs on plane A. Figure 7 The process is briefly illustrated in (h) to (i), the purpose of which is to provide the two hind legs with room to move and sufficient torque; then the remaining two hind legs are transferred from surface A to a pre-calculated symmetrical position on surface B, combined with... Figure 7 As shown in (i) to (l), the entire bionic legged climbing robot finally achieves stable four-legged standing on surface B;

[0137] While the present invention has been disclosed above, its scope of protection is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, and all such changes and modifications will fall within the scope of protection of the present invention.

Claims

1. A three-dimensional wall-adaptive cross-wall transition and gait control method for a biomimetic legged climbing robot, characterized in that, Includes the following steps: S100. Construct a kinematic constraint model based on the three-point adsorption stability criterion to ensure that the bionic legged climbing robot always has at least three foot ends that can reliably adsorb onto the wall to form adsorption points during the movement process, and the support area formed by the three adsorption points includes the projection of the robot's center of gravity onto the wall. S200 employs a two-stage motion mechanism for gait planning and execution. The first stage is the foot displacement stage, which controls the suction cups (304) on the robot's legs to detach from the wall and move along a preset trajectory to the target pose before re-attaching, while maintaining the body pose unchanged. The second stage is the body translation stage, which controls all legs to remain in the attached state and drives the body to move to the target pose. S300. Based on the task requirements, and using the motion constraint model constructed in step S100 and the two-stage motion mechanism in step S200, plan and execute at least one of the following: forward gait, self-rotation gait, and wall transfer gait, so as to realize the robot's movement on a single wall or adaptive transition between different walls in three-dimensional space.

2. The three-dimensional wall-adaptive cross-wall transition and gait control method for a biomimetic legged climbing robot according to claim 1, characterized in that: The biomimetic legged climbing robot includes at least three legs, with a suction cup (304) at the end of each leg, and a thin-film pressure sensor (205) inside each leg.

3. The three-dimensional wall-adaptive cross-wall transition and gait control method for a biomimetic legged climbing robot according to claim 2, characterized in that: The biomimetic legged climbing robot has four legs.

4. The three-dimensional wall-adaptive cross-wall transition and gait control method for a biomimetic legged climbing robot according to claim 3, characterized in that: In step S100, the DH parameter method is used to model a single leg. Based on the transformation matrix of a single joint, the cumulative foot pose is: in, 0 P4 refers to the position of the foot at point P. i-1 T i It is obtained from the following calculation formula: in, This indicates the length of the connecting rod, which is the distance between the common perpendiculars of the two joint axes; It represents the torsion angle of the link, that is, the rotation angle about the common perpendicular from axis i-1 to axis i; This represents the link offset, which is the distance along joint axis i from the intersection of the common perpendicular and the axis to the next common perpendicular. This represents the joint angle, which is the rotation angle around the joint axis i from link i-1 to link i. Starting with a single leg, proceed sequentially... 01 T1 to 3 The product of T4 finally yields the pose of foot P; Inverse kinematics is solved iteratively using a custom-damped least squares method. The angles of the four joints are inferred from the target pose at the foot end, and the position and attitude error functions are constructed and combined into the total error. ; in, These are the weighting coefficients. For ankle joint angle, These represent the target angle and current angle of the ankle joint, respectively; x, y, and z represent the three coordinate components in the base coordinate system of each leg, curr is the current position, and tar is the target position; A Jacobian matrix is ​​constructed based on numerical differentiation to describe the influence of joint angles on foot position. The joint angles are then iteratively corrected, including... Damped least squares joint correction for: in, for Jacobian matrix The damping coefficient is... for identity matrix , These are the target at the foot and the current position, respectively. The joint angles have been updated to: in, For the updated joint angles, For the current joint angle, This represents the change in joint angle. Until the total error satisfies the condition that the position error is less than 0.1 mm and the posture error is less than 0.1°, the final joint angle is stored in the joint parameter array of the robot model to drive the foot movement.

5. The three-dimensional wall-adaptive cross-wall transition and gait control method for a biomimetic legged climbing robot according to claim 4, characterized in that: In step S100, the mathematical criterion for stability is based on the three-point adsorption criterion, assuming... Let be the projection point of the fuselage's center of gravity onto the supporting wall. The coordinates of the three adsorption feet projected onto the wall surface are shown; these coordinates are calculated in real-time during planning. Form a supporting triangle and determine... Whether it is located inside the supporting triangle, if If the robot is located inside the supporting triangle, it is considered stable. If the robot is located outside the supporting triangle, it is determined to be unstable.

6. The three-dimensional wall-adaptive cross-wall transition and gait control method for a biomimetic legged climbing robot according to claim 5, characterized in that: In step S200, S210, During the foot displacement phase, including, S211, Command Issuance and Adhesion Release: The main controller (401) in the bionic legged climbing robot loads gait parameters to verify whether the bionic legged climbing robot is in the foot displacement completion state. If there are still feet that have not completed the displacement within the gait cycle, the robot controls the foot of the bionic legged climbing robot to delay and wait. At the same time, it reads the data of the thin film pressure sensor (205) integrated in the foot and confirms that the pressure value has dropped below the preset detachment threshold. It is determined that the adhesion has been reliably released and proceeds to step S212. If the pressure has not dropped to the threshold, it enters the abnormal handling process. S212, Trajectory Generation and Joint Driving: Based on the current foot position and target position, a smooth spatial trajectory including lifting, horizontal movement and landing is planned using a cubic polynomial interpolation method, ensuring that the velocity and acceleration of the trajectory are zero at the start and end points; The main controller (401) calculates the trajectory into the angle sequence of the four joints according to the inverse kinematics solution of step S100, and drives the servo motor to execute. S213, Foot placement and adsorption confirmation: After the foot moves to the target position, the main controller (401) sends a command to the foot and reads the data from the thin film pressure sensor (205) again to confirm that the pressure value has risen above the preset reliable adsorption threshold and remains stable, and determines that adsorption has been reliably established. Repeat steps S211-S213 until all feet have moved, then proceed to the next stage; S220, during the fuselage relocation phase, including, S221, Target pose meter, which calculates the next target pose that the computer needs to move to according to gait requirements; S222, Joint Angle Solution and Driving: In this stage, the positions of all foot ends in the world coordinate system are kept constant, using kinematic relationships... ,in Represents the position vector of the foot in the world coordinate system. It's about yaw angle. The rotation matrix, It is the position vector of the foot in the fuselage coordinate system. It is the relative position of the origin of the fuselage coordinate system in the world coordinate system; The inverse kinematics method yields the new positions of each foot in the fuselage coordinate system under the new fuselage pose. ; Subsequently, the sequence of joint angle changes required to move the fuselage to the target pose is solved again using inverse kinematics, and the main controller (401) coordinates the servo motors to execute this sequence synchronously. S223, Phase complete. After all joints have moved into position, the fuselage migration phase is complete; the system status is updated, awaiting the start of the next gait cycle.

7. The three-dimensional wall-adaptive cross-wall transition and gait control method for a biomimetic legged climbing robot according to claim 6, characterized in that: In step S300, S310, For the forward-moving state, it is used for efficient linear movement within a single wall surface, including a cycle of sequential single-leg shifting-intermittent fuselage following; S320. For its own rotational gait, a strategy of single-leg fixed axis - multi-leg arc shift - step-by-step rotation is adopted to realize the turning of the bionic legged climbing robot around the foot of one of its own legs. Using the kinematic constraint model in step S100, the maximum rotation angle that the robot can achieve in a single rotation is pre-calculated under the constraint of fixed axis foot end and limit constraints of other foot joint angles. If the target total rotation angle is greater than the maximum rotation angle in a single rotation, it is decomposed into multiple steps, each step rotating by the maximum rotation angle in a single rotation, and the last step rotating by an angle smaller than the maximum rotation angle in a single rotation to complete the entire rotation action. Each step includes moving the foot end arc and rotating the body around the axis foot end by the maximum rotation angle in a single rotation. S330. The wall-to-wall gait for ground A and wall B includes, S331, Symmetrical Distribution Stage First, determine the desired configuration when the robot is fully positioned on wall B, and then deduce the ideal footholds for its two front legs on wall B. Before the transfer begins, the robot remains on ground A; by controlling the leg joints, the position of the two hind legs on ground A is adjusted so that their projection points... and The two planes are geometrically symmetrical about the intersection line; the two front legs are transferred to the target point B on the wall through the foot-end displacement stage in sequence; at this time, a stable cross-wall support is formed with the two feet on the wall B and the two feet on the ground A; S332, during the center of gravity transfer phase, the fuselage changes from parallel to the ground A to parallel to the wall B; through inverse kinematics numerical solution, a continuous sequence of 16 joint angles changing with time is obtained, driving the fuselage to complete a smooth rotation; S333, Leg retraction phase: After the body rotation is completed, the body configuration and foot position are symmetrical with respect to the body configuration and position before the rotation about the intersection line O of ground A and wall B. The two front legs located on wall B are extended forward through the foot displacement phase until they are symmetrical with respect to the initial position of the two hind legs on ground A about the intersection line O. Then the remaining two hind legs are transferred from ground A to the pre-calculated symmetrical position on wall B, finally completing the stable four-legged standing of the entire bionic legged climbing robot on wall B.

Images