A method, apparatus, terminal device, and computer-readable storage medium for sliding mode tracking control of a quadruped robot dog for cable inspection.

Abstract

This invention discloses a sliding mode tracking control method, device, terminal equipment, and computer-readable storage medium for a quadruped robot dog used for cable inspection, belonging to the field of intelligent power inspection technology. The method includes: acquiring the desired cable inspection path, desired movement speed, and robot dog operating status; calculating the current kinematic tracking error of the chassis; acquiring the current visual pose deviation, transforming it in a coordinate system, and weightedly fusing it with the kinematic tracking error to obtain the current comprehensive tracking error; inputting the current comprehensive tracking error and desired movement speed into a sliding mode controller based on proportional-integral ratio to generate a sliding surface, substituting it into a saturation function to solve the sliding mode control law, and obtaining the underlying motion control quantity; and driving the robot dog to operate according to this control quantity. This invention eliminates the closed-loop deviation between the heterogeneous visual and chassis data, overcoming the problem of robot dogs easily becoming unstable and failing to track in complex terrain due to the lack of anti-shake and anti-disturbance control.

Description

Technical Field

[0001] This invention relates to the field of intelligent power inspection technology, and in particular to a method, device, terminal equipment, and computer-readable storage medium for sliding mode tracking control of a four-legged robotic dog for cable inspection. Background Technology

[0002] When quadrupedal robotic dogs perform field cable inspection tasks, high-precision trajectory tracking control is the core foundation for ensuring stable operation and autonomous navigation. Existing trajectory tracking systems typically rely solely on the chassis odometer and the vision system onboard the robotic dog to calculate kinematic states, and employ conventional open-loop control or PID regulation mechanisms for low-level motion control.

[0003] However, conventional open-loop control or PID regulation mechanisms for low-level motion control have drawbacks such as poor resistance to nonlinear mechanical disturbances and high-frequency jittering under strong interference compensation. This makes it easy for robot dogs to experience problems such as the inability to coordinate multi-source heterogeneous feedback data, gait instability, and eventual tracking failure when crossing complex terrain in the wild. Summary of the Invention

[0004] This invention provides a sliding mode tracking control method for a quadruped robot dog for cable inspection. It can solve the problems in the prior art that cannot eliminate the closed-loop deviation between visual and chassis heterogeneous data, and lacks a low-level tracking control mechanism that takes into account both anti-interference and anti-shaking, which makes the robot dog prone to gait instability and eventual tracking failure when crossing complex terrain in the wild.

[0005] One embodiment of the present invention provides a sliding mode tracking control method for a quadruped robot dog for cable inspection, comprising: The system acquires the current desired cable inspection path, the current desired motion speed, and the current operating status of the robot dog. Based on the desired cable inspection path and the current operating status of the robot dog, it calculates the current kinematic tracking error of the robot dog chassis. Using the robot dog's onboard camera, it acquires the current visual pose deviation of the robot dog relative to the cable. After transforming the current visual pose deviation to the robot dog's coordinate system, it is weighted and fused with the current kinematic tracking error to obtain the current comprehensive tracking error. The current comprehensive tracking error and the current desired motion speed are input into a sliding mode controller based on proportional-integral ratio to generate a sliding surface. The sliding surface is substituted into a saturation function to reduce chattering to solve the sliding mode control law. Based on the sliding mode control law, the underlying motion control quantity, including linear velocity and angular velocity, is calculated. Based on the underlying motion control quantity, the robot dog's motion is controlled at the next moment.

[0006] Furthermore, the current methods for determining the desired cable inspection path include: If the current time is the initial time, then the static obstacle map of the environment and the known cable line routing information are obtained; based on the static obstacle map and the cable line routing information, the optimal global path that fits the cable line is generated by the ant colony algorithm, and the optimal global path is used as the expected path for the current cable inspection; wherein, the state transition probability heuristic function of the ant colony algorithm includes a path tortuosity penalty term. If the current moment is not the initial moment, the dynamic obstacle information of the surrounding environment is obtained through the airborne camera, and the safe braking distance between the robot dog and each dynamic obstacle is calculated based on the robot dog's current running speed, motor acceleration limit, and the dynamic obstacle information, and a feasible speed space including the robot dog's linear velocity and angular velocity is constructed. Speed ​​sampling is performed within the feasible speed space. The current running state of the robot dog and the various speed combinations obtained by sampling are extrapolated within a preset time window to predict and generate local simulated trajectories under each speed combination. An evaluation function that includes a cable deviation penalty term is used to score the cost value of each local simulated trajectory, and the optimal local trajectory with the highest cost value score is selected as the expected path for the current cable inspection.

[0007] Furthermore, the evaluation function, which includes a cable deviation penalty term, is used to score the cost of each local simulated trajectory, specifically including: For each local simulated trajectory, the following evaluation factors are extracted: heading angle evaluation factor representing the alignment between the trajectory end and the target point, distance evaluation factor representing the closest distance between the trajectory and surrounding obstacles, and speed evaluation factor representing the magnitude of the movement speed. Calculate the vertical distance from the trajectory point on the local simulated trajectory to the optimal global path, and determine the ratio of the vertical distance to the maximum allowable deviation distance as the cable deviation penalty term; The basic evaluation sum is obtained by multiplying the heading angle evaluation factor, the distance evaluation factor, and the speed evaluation factor by their respective weighting coefficients. Calculate the product of the cable deviation penalty term and the corresponding weighting coefficient to obtain the weighted cable deviation penalty term; The cost score of the local simulated trajectory is calculated based on the difference between the basic evaluation value and the weighted cable deviation penalty term.

[0008] Furthermore, the methods for determining the current desired speed of motion include: If the current moment is the initial moment, then the current desired speed is the preset speed; If the current time is not the initial time, the speed combination corresponding to the current cable inspection expected path is taken as the current expected movement speed.

[0009] Furthermore, the step of transforming the current visual pose deviation to the robot dog coordinate system and then weighting and fusing it with the kinematic tracking error to obtain the current comprehensive tracking error specifically includes: obtaining the pre-calibrated lateral offset and installation yaw angle of the airborne camera, and constructing a calibration transformation matrix based on the lateral offset of the airborne camera and the installation yaw angle; Using the calibration transformation matrix, the current visual pose deviation defined in the coordinate system of the airborne camera is converted into the transformed visual pose deviation defined in the coordinate system of the robot dog; wherein, the current visual pose deviation specifically includes the lateral offset of the cable centerline in the image and the heading deviation. The current kinematic tracking error is multiplied by a given fusion weight coefficient to obtain the first weighting term; The converted visual pose deviation is weighted by the complementary weight coefficients corresponding to the fusion weight coefficients to obtain the second weighting term. The first weighted term and the second weighted term are summed to obtain the current comprehensive tracking error.

[0010] Furthermore, a sliding mode surface is constructed using the comprehensive tracking error, and the sliding mode surface is substituted into the saturation function that weakens chattering to solve for the sliding mode control law, specifically including: The sliding mode control law is calculated using the following formula: ; In the formula, This includes the underlying motion control variables, which contain linear velocity and angular velocity. The inverse matrix of the control input coefficient matrix is ​​preset in the chassis kinematic model of the robot dog. For the current desired speed of motion, To track error variables, This is an estimate of the unknown disturbance. This refers to the proportional coefficient of the proportional adjustment term in the sliding mode controller. The integral coefficient of the integral adjustment term in the sliding mode controller is... As an additional input including the saturation function, t represents the current control time; in the actual closed-loop solution of the sliding mode controller, the tracking error variable Extract the value of the current comprehensive tracking error and perform calculations.

[0011] Furthermore, the additional input containing the saturation function is determined as follows: ; In the formula, η is a given sliding surface constant, t is time, and s is the sliding surface constructed based on the current integrated tracking error, and satisfies the following conditions: ; To prevent interference with observation errors and satisfy the following conditions , The boundary layer thickness of the sliding surface. The coefficients of the convergence term, Let be the saturation function for reducing chattering.

[0012] Another embodiment of the present invention provides a quadruped robot dog sliding mode tracking control device for cable inspection, including: a status acquisition module, an error calculation module, a multi-source fusion module, a sliding mode solving module, and a chassis drive module; the status acquisition module is used to acquire the current desired cable inspection path, the current desired movement speed, and the current operating status of the robot dog; The system comprises the following modules: an error calculation module, used to calculate the current kinematic tracking error of the robot dog chassis based on the desired cable inspection path and the robot dog's current operating state; a multi-source fusion module, used to acquire the robot dog's current visual pose deviation relative to the cable using the robot dog's onboard camera, transforming the current visual pose deviation to the robot dog's coordinate system, and then weighting and fusing it with the current kinematic tracking error to obtain the current comprehensive tracking error; a sliding mode solution module, used to input the current comprehensive tracking error and the desired motion velocity into a sliding mode controller based on proportional-integral ratios, generating a sliding mode surface, substituting the sliding mode surface into a saturation function to reduce chattering, solving for the sliding mode control law, and calculating the underlying motion control quantities including linear velocity and angular velocity based on the sliding mode control law; and a chassis drive module, used to control the robot dog's motion at the next moment based on the underlying motion control quantities.

[0013] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the quadruped robot dog sliding mode tracking control method for cable inspection as described in any one of the present invention.

[0014] Another embodiment of the present invention provides a computer-readable storage medium, comprising: a stored computer program, wherein, when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the quadruped robot dog sliding mode tracking control method for cable inspection as described in any one of the present invention.

[0015] The present invention provides the following beneficial effects: It offers a sliding mode tracking control method for a quadruped robot dog used for cable inspection, comprising: acquiring the current desired cable inspection path, the current desired motion speed, and the current operating state of the robot dog; calculating the current kinematic tracking error of the robot dog chassis based on the current desired cable inspection path and the current operating state of the robot dog; acquiring the current visual pose deviation of the robot dog relative to the cable using the robot dog's onboard camera, transforming the current visual pose deviation to the robot dog's coordinate system, and then weighting and fusing it with the current kinematic tracking error to obtain the current comprehensive tracking error; inputting the current comprehensive tracking error and the current desired motion speed into a sliding mode controller based on proportional-integral ratios to generate a sliding surface; substituting the sliding surface into a saturation function to reduce chattering to solve the sliding mode control law, and calculating the underlying motion control quantity including linear velocity and angular velocity based on the sliding mode control law; and controlling the robot dog's motion at the next moment based on the underlying motion control quantity.

[0016] Compared with existing trajectory tracking methods that rely solely on independent sensors for calculation and employ conventional open-loop or PID control, the technical solution provided by this invention directly overcomes the shortcomings of lacking an effective coordinate transformation and fusion mechanism between multi-source heterogeneous sensors by weighted fusion of the current visual pose deviation acquired by the airborne camera after coordinate system transformation and the current kinematic tracking error of the chassis. This completely solves the problems of inaccurate collaborative closed-loop control and severe feedback deviation of heterogeneous feedback data in complex terrain. At the same time, this solution inputs the current comprehensive tracking error into a sliding mode controller based on proportional-integral ratio, generates a sliding surface, and substitutes it into the saturation function to solve the sliding mode control law to calculate the underlying motion control quantity. This underlying closed-loop architecture effectively overcomes the shortcomings of conventional controllers, such as poor resistance to nonlinear mechanical disturbances and the tendency for control commands to exhibit high-frequency chattering under strong interference compensation. It fundamentally solves the problem of gait instability and eventual tracking failure that is prone to occur when the robot dog crosses complex terrain in the wild, thereby achieving high-precision and robust anti-disturbance tracking of the desired path for cable inspection of the quadruped robot dog. Attached Figure Description

[0017] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a sliding mode tracking control method for a quadruped robot dog for cable inspection, provided by an embodiment of the present invention.

[0019] Figure 2This is a schematic diagram of the device structure of a sliding mode tracking control method for a quadruped robot dog for cable inspection provided in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0022] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0023] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0024] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0025] In the description of the embodiments of this application, the term "multiple" refers to two or more (including two), similarly, "multiple sets" refers to two or more (including two sets), and "multiple pieces" refers to two or more (including two pieces).

[0026] In the description of the embodiments of this application, unless otherwise expressly specified and limited, technical terms such as "installation," "connection," "joining," and "fixing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. For those skilled in the art, the specific meaning of the above terms in the embodiments of this application can be understood according to the specific circumstances.

[0027] The existing technology cannot eliminate the closed-loop bias between visual and chassis heterogeneous data, and lacks a low-level tracking control mechanism that takes into account both anti-interference and anti-shaking, which makes the robot dog prone to gait instability and eventual tracking failure when crossing complex terrain in the wild.

[0028] One embodiment of the present invention provides a sliding mode tracking control method for a quadruped robot dog for cable inspection, comprising: Step S1: Obtain the current desired cable inspection path, the current desired movement speed, and the current operating status of the robot dog; In a preferred embodiment, the method for determining the desired path for the current cable inspection includes: If the current time is the initial time, then the static obstacle map of the environment and the known cable line routing information are obtained; based on the static obstacle map and the cable line routing information, the optimal global path that fits the cable line is generated by the ant colony algorithm, and the optimal global path is used as the expected path for the current cable inspection; wherein, the state transition probability heuristic function of the ant colony algorithm includes a path tortuosity penalty term. If the current moment is not the initial moment, the dynamic obstacle information of the surrounding environment is obtained through the airborne camera, and the safe braking distance between the robot dog and each dynamic obstacle is calculated based on the robot dog's current running speed, motor acceleration limit, and the dynamic obstacle information, and a feasible speed space including the robot dog's linear velocity and angular velocity is constructed. Speed ​​sampling is performed within the feasible speed space. The current running state of the robot dog and the various speed combinations obtained by sampling are extrapolated within a preset time window to predict and generate local simulated trajectories under each speed combination. An evaluation function that includes a cable deviation penalty term is used to score the cost value of each local simulated trajectory, and the optimal local trajectory with the highest cost value score is selected as the expected path for the current cable inspection.

[0029] Preferred methods for determining the current desired speed of motion include: If the current moment is the initial moment, then the current desired speed is the preset speed; If the current time is not the initial time, the speed combination corresponding to the current cable inspection expected path is taken as the current expected movement speed.

[0030] Preferably, the step of using an evaluation function that includes a cable deviation penalty term to score the cost of each local simulated trajectory specifically includes: For each local simulated trajectory, the following evaluation factors are extracted: heading angle evaluation factor representing the alignment between the trajectory end and the target point, distance evaluation factor representing the closest distance between the trajectory and surrounding obstacles, and speed evaluation factor representing the magnitude of the movement speed. Calculate the vertical distance from the trajectory point on the local simulated trajectory to the optimal global path, and determine the ratio of the vertical distance to the maximum allowable deviation distance as the cable deviation penalty term; The basic evaluation sum is obtained by multiplying the heading angle evaluation factor, the distance evaluation factor, and the speed evaluation factor by their respective weighting coefficients. Calculate the product of the cable deviation penalty term and the corresponding weighting coefficient to obtain the weighted cable deviation penalty term; The cost score of the local simulated trajectory is calculated based on the difference between the basic evaluation value and the weighted cable deviation penalty term.

[0031] Preferably, in the process of generating the optimal global path that fits the cable line by means of the improved ant colony algorithm if the current time is the initial time, in order to solve the problem that the blind divergent search in the initial stage of the traditional ant colony algorithm leads to extremely slow convergence, this embodiment assigns a priori values ​​to the initial pheromone concentration distribution matrix based on the known geometric direction of the cable line during the algorithm initialization stage (i.e., when the iteration round t=0). The specific priori pheromone initialization formula is constructed as follows: ; In the formula, This represents the initial pheromone concentration on the path from the current node i to the next adjacent optional node j in the environmental raster map; i and j represent the spatial index numbers of two connected nodes in the raster map, respectively. `i,j` represents the global baseline pheromone constant (e.g., a positive constant of 0.1), used to ensure that all unexplored unknown areas in the map have a very small base search probability, preventing the algorithm from stagnating in the early stages. `i,j` represents the indicator function (i.e., a Boolean function) based on the cable direction, used to evaluate the effectiveness of the search direction. Its calculation logic is as follows: extract the edge vector connecting node `i` to node `j`, calculate the angle between this vector and the known cable route vector; if the angle is less than a given angle threshold (e.g., set to 30°, physically meaning the search direction basically follows the cable corridor), then `i,j` = 1; if the angle is greater than or equal to the angle threshold, then `i,j` = 0. Through this formula, the system pre-lays a high-concentration pheromone channel for the ant colony at a macroscopic level.

[0032] Furthermore, in calculating the state transition probability of the improved ant colony algorithm, to suppress the generation of a large number of zigzag paths during the search, this embodiment redesigned the heuristic function and constructed the following formula for calculating the heuristic function, which includes a path tortuosity penalty term: ; In the formula, This represents the heuristic function value when the nth ant moves from the current node i to the next node j in the t-th system iteration. The larger the value, the higher the preference of moving in that direction; n represents the number parameter of the specific ant currently performing path optimization; t represents the iteration round time variable in which the algorithm is currently running. This represents the Euclidean linear distance between grid node i and node j in physical space. Let be the direction angle from node i to node j. This represents the entry direction angle of the nth ant when it enters the current node i, moving along the previous historical path; therefore This represents the actual yaw angle at which the ant makes a turn at node i; μ represents the set path tortuosity penalty coefficient, which is a constant greater than 0 (preferably in the range of 0.5 to 2.0). From this fractional structure, it can be seen that when the ant attempts to make a sharp turn at a large angle, i.e. (… (where the value is extremely large), leading to a heuristic function value The drastic decay suppresses the possibility of the robot dog's route frequently making sudden stops and turns at the mathematical level.

[0033] Preferably, if the current time is not the initial time, a feasible velocity space including the linear velocity and angular velocity of the robot dog is dynamically constructed. Specifically, this is achieved by finding the intersection of the following three physical level boundary velocity sets (i.e. ): (1) Allowable velocity set The construction formula is: ; in, This represents the theoretical speed space limited by the mechanical motor limits of the robot dog itself; V and Ω represent the sampled linear velocity and angular velocity control commands, respectively. and This indicates the minimum (e.g., reversing limit) and maximum linear speed thresholds that the chassis drive motor can output based on its physical structure; and This indicates the minimum and maximum angular velocity rotation thresholds that the chassis can withstand.

[0034] (2) Set of reachable velocities The construction formula is: ; in, This represents the actual velocity space that can be achieved in the next sampling moment after taking into account the inertia and acceleration / deceleration capabilities of the machine body; and This represents the robot dog's actual linear velocity and actual angular velocity at the current moment; and These represent the maximum linear acceleration and maximum angular acceleration limits that the motor can provide (both positive numbers); Δt represents the single-step execution time interval of the control system. The formula for the achievable speed set limits abrupt changes in the robot dog's speed commands, preventing drive wheel slippage caused by issuing commands exceeding the acceleration limits.

[0035] (3) Safety braking speed set The construction formula is: ; In the formula, This represents the baseline safe speed space set to prevent physical collisions with suddenly appearing dynamic obstacles; d(V,Ω) represents the minimum Euclidean distance between the predicted trajectory endpoint and the nearest obstacle when moving along a local trajectory with the current sampling command combination (V,Ω); and These represent the maximum linear deceleration and maximum angular deceleration (absolute values) that the robot dog can provide during emergency braking. This formula is derived from the kinematic uniform deceleration linear formula, strictly ensuring that the selected speed command can guarantee that the robot dog can completely stop before a collision occurs.

[0036] Preferably, in the process of using an evaluation function that includes a cable deviation penalty term to score the cost of each local simulated trajectory in order to select the optimal speed combination, the complete closed-loop mathematical model for cost evaluation is expressed as follows: ; Where F(V,Ω) represents the comprehensive cost score of the local trajectory predicted for a specific combination of motion velocities (V,Ω) currently sampled. The higher the score, the higher the priority of the trajectory being selected. This represents a smoothing normalization function, used to scale the factors with different dimensions after summing within the parentheses to a uniform range of [0,1]. The heading angle evaluation factor is used to measure the geometric angle deviation between the end orientation of the predicted trajectory and the direction of the local target point. The smaller the angle, the higher the score of this factor. This represents the movement speed evaluation factor, which is used to encourage the robot dog to move at the highest possible speed when the environment is safe, in order to improve inspection efficiency. The distance evaluation factor represents the safety distance evaluation factor, which means that the farther the predicted trajectory is from the known obstacle, the higher the score of this factor. , , These are the positive weighting constant coefficients for the three basic gain factors: heading, distance, and speed. A negative penalty weighting coefficient (a positive constant) specifically designed for cable inspection operations; a larger one... It will penalize deviations from the cable's trajectory, thus ensuring that the robot dog does not deviate from the cable inspection path while dynamically avoiding obstacles.

[0037] The cable deviation penalty term represents a unique cable deviation penalty term that defines the maximum tolerance for the robot dog to deviate from the original inspection corridor during local obstacle avoidance. Its specific boundary penalty formula is expressed as follows: ; Where max{·} represents the maximum value operation; This represents the physical coordinates of any sampled prediction point derived by the robot dog on this local simulated trajectory within a two-dimensional horizontal plane. This represents the vertical projected Euclidean distance from the predicted point coordinates to the optimal globally desired cable path; `max` represents the maximum safe distance the robot dog is allowed to deviate from the cable, a constraint imposed by the system in engineering practice (e.g., based on the maximum field of view (FOV) of the airborne inspection camera, it can be calculated and set to 3 meters). This is in case the robot dog is forced to veer significantly to avoid large construction facilities. The value will increase accordingly. This is because the penalty term has a weighting coefficient with a negative sign at the beginning. Therefore, if a candidate obstacle avoidance trajectory becomes excessively circuitous, it will lead to... Approaching The total score F(V,Ω) of this trajectory will be instantly subjected to a severe cliff-like algebraic penalty and be eliminated.

[0038] In this preferred embodiment, by injecting a Boolean function based on the prior geometric features of the cable into the global ant colony planning initialization stage, and innovatively superimposing a negative suppression term for path tortuosity in the probability selection, the limitation of conventional swarm intelligence algorithms that blindly search and cause computing power to be wasted is broken. A macroscopically smooth reference route that closely matches the cable direction can be generated in an extremely short computing cycle. Subsequently, when the robot dog encounters obstacles with sudden dynamic and static interplay during actual movement, a rigorous speed feasible space to avoid collisions is constructed by strictly calculating the intersection dynamic limit, motor acceleration and deceleration windows, and absolute safe braking boundaries. By embedding a division overflow penalty term (i.e., deviation penalty term) bound to the physical maximum observation line-of-sight depth into the DWA local window cost evaluation system, this dual-layer protection mechanism combining global prior guidance and local hard correction fundamentally solves the technical problem of traditional path algorithms causing the robot to deviate from its path and permanently lose the observation line of the power transmission line due to excessive obstacle avoidance in open fields. Ultimately, while achieving high dynamics and millisecond-level smooth obstacle avoidance, the robot dog's physical trajectory is always constrained within the safe line-of-sight distance of the cable path. This ensures the robot dog's chassis mobility in complex field conditions and maximizes the continuity and reliability of the power inspection task, providing a continuous expected reference flow with high confidence and high anti-disturbance potential for subsequent low-level sliding mode attitude control.

[0039] Step S2: Based on the expected path of the current cable inspection and the current operating status of the robot dog, calculate the current kinematic tracking error of the robot dog chassis; In a preferred embodiment, the current kinematic tracking error of the robot dog chassis is specifically achieved by extracting the vector difference between the target's desired coordinates and the chassis's real-time coordinates. During the robot dog's movement, its actual kinematic model is constrained by both its own mechanical structure and the external environment. In this embodiment, the mathematical model defining the robot dog's chassis kinematic state equation is as follows: ; In the formula, t represents the current continuous operation control time of the system; x(t) represents the current operating state variable of the robot dog, specifically a two-dimensional column vector. ,in and These represent the longitudinal and lateral actual physical position coordinates of the robot dog's chassis center in the world coordinate system at the current moment (usually calculated by the wheel-leg odometer and IMU inertial navigation unit built into the robot dog's chassis). This represents the derivative of the state variable x(t) with respect to time, and its physical meaning is the actual instantaneous velocity vector of the robot dog's chassis in the two-dimensional plane; The control input represents the control input vector issued by the system to the chassis at the current moment, specifically including the desired combination of linear velocity V and angular velocity Ω, i.e. ; The matrix represents the pre-calibrated control input coefficients used to map the linear angular velocity at the command level into the physical displacement components of the chassis in the Cartesian coordinate system; f(t) represents the estimated term of unknown physical disturbances encountered by the robot dog when moving in the wild, such as slippage of the chassis feet, sudden changes in mechanical friction, and wind load when crossing muddy ditches or moss-covered slopes. These nonlinear external disturbances, which cannot be accurately described by conventional kinematic formulas, are uniformly abstracted and quantified into disturbance terms. f(t)

[0040] Based on the above operational state definition, the system further extracts the expected target location corresponding to the current time t from the current expected cable inspection path, denoted as . At this point, the system calculates the current kinematic tracking error e(t) of the robot dog's chassis according to the following vector deviation formula: ; Where e(t) represents the current kinematic tracking error; x(t) represents the ideal absolute coordinates that the robot dog must reach at time t based on the upper-level navigation algorithm (such as the optimal local trajectory calculated by combining the improved DWA dynamic window with the ant colony). x(t) represents the current actual coordinate state of the chassis.

[0041] In a real-world inspection scenario, suppose the robot dog is moving along a cable laid on a rugged hillside, and the upper-level planning algorithm gives the expected position it should reach in the next second. However, due to the slippery terrain, the chassis experienced slight lateral slippage (i.e., a physical disturbance f(t)) during forward movement, causing the actual current position x(t) calculated by the chassis odometer to not perfectly fall on the preset trajectory. At this time, the vector difference e(t) calculated by the above formula can intuitively and accurately quantify this physical displacement deviation caused by purely mechanical and terrain factors (including lateral error due to deviation from the cable and longitudinal lag error due to failure to keep up in time).

[0042] In this embodiment, by strictly defining the dimension of the underlying state vector and explicitly extracting the unknown disturbance term f(t) representing terrain slippage and external nonlinear interference from the robot dog's kinematic model, the current kinematic tracking error e(t) in the continuous time domain is calculated by subtracting the desired coordinates from the current coordinates. This processing mechanism can accurately quantify the degree of instantaneous trajectory deviation of the robot dog caused by harsh terrain from a purely physical and mechanical perspective within extremely high-frequency control cycles. It not only transforms the influence of complex terrain into a clear indicator that can be algebraically calculated, but also lays an extremely solid and logically sound mathematical benchmark system for subsequent heterogeneous coordinate system fusion with high-precision visual sensor data and for providing an absolute deviation target point with disturbance estimation for the nonlinear sliding mode controller.

[0043] Step S3: Based on the airborne camera of the robot dog, obtain the current visual pose deviation of the robot dog relative to the cable, transform the current visual pose deviation to the robot dog coordinate system through coordinate system transformation, and then perform weighted fusion with the current kinematic tracking error to obtain the current comprehensive tracking error; In a preferred embodiment, the step of transforming the current visual pose deviation to the robot dog coordinate system and then weighting and fusing it with the kinematic tracking error to obtain the current comprehensive tracking error specifically includes: obtaining the pre-calibrated lateral offset and installation yaw angle of the airborne camera, and constructing a calibration transformation matrix based on the lateral offset of the airborne camera and the installation yaw angle. Using the calibration transformation matrix, the current visual pose deviation defined in the coordinate system of the airborne camera is converted into the transformed visual pose deviation defined in the coordinate system of the robot dog; wherein, the current visual pose deviation specifically includes the lateral offset of the cable centerline in the image and the heading deviation. The current kinematic tracking error is multiplied by a given fusion weight coefficient to obtain the first weighting term; The converted visual pose deviation is weighted by the complementary weight coefficients corresponding to the fusion weight coefficients to obtain the second weighting term. The first weighted term and the second weighted term are summed to obtain the current comprehensive tracking error.

[0044] Preferably, in the process of transforming the visual parameters in the camera coordinate system to the robot dog coordinate system using the calibration transformation matrix, considering that the optical center of the airborne binocular vision system often cannot be completely aligned with the physical centroid of the robot dog chassis or the odometry reference center during physical installation, this embodiment constructs a rigid body coordinate transformation mathematical model to eliminate this physical installation error. The specific transformation formula is expressed as follows: ; In the formula, This represents the current visual pose deviation (i.e., the raw visual output parameter) extracted directly from the airborne vision system and defined in the camera's local coordinate system. Specifically, it is a column vector containing position and orientation. ; This indicates the lateral offset of the cable centerline in the currently captured image (i.e., the left-right physical deviation of the robot dog's head relative to the target cable). The geometric angle between the centerline of the cable and the robot's current forward direction (i.e., the direction of the camera's optical axis) (i.e., the yaw error of the robot dog relative to the cable). This indicates the physical deviation of the pre-calibrated offline installation position of the airborne camera relative to the axis of the robot dog's center of mass in lateral translation. This indicates the physical deviation of the pre-calibrated airborne camera optical axis from the forward axis of the robot dog's mounting yaw angle. To utilize the aforementioned installation yaw angle The core calibration transformation matrix is ​​constructed; This indicates that after the aforementioned matrix rotation compensation and vector translation compensation, the final transformed visual pose deviation column vector is uniformly mapped to the robot dog's global centroid coordinate system. Through this formula, the system mathematically eliminates parallax spurious offsets caused by eccentric camera mounting, ensuring consistency between the visually perceived deviation data and the actual chassis deviation on a spatial physical reference.

[0045] Furthermore, after obtaining visual and kinematic feedback under the same coordinate system reference, this embodiment calculates the current integrated tracking error by constructing a complementary weighted summation mechanism. The mathematical expression of this fusion mechanism is as follows: ; In the formula, This represents the current integrated tracking error that is finally output to the underlying controller; e(t) represents the current kinematic tracking error extracted in step S2 and calculated based on internal sensors such as the chassis odometer. This represents the fusion weighting coefficient, which is either manually set or adaptively adjusted by the system, and its value range is strictly limited to [specific range]. between, This refers to the first weighted term, which is dominated by chassis internal parameters; (1- This refers to the complementary weight coefficients that have a normalized correspondence with the fusion weight coefficients in the probability space. The product term in the formula (1- ) This is the second weighting term, which is dominated by visual extrinsic parameters.

[0046] In practical field cable inspection applications, this fusion logic has strong engineering anti-interference physical significance: the robot dog's chassis odometer calculation (kinematic error) has extremely high data smoothness and update frequency in a very short time, but when the robot dog walks on muddy ground for a long time and experiences slight slippage, the odometer will generate irreversible low-frequency cumulative drift error; conversely, the airborne vision system (visual pose deviation) always uses the actual physical entity of the cable as the absolute reference and will not generate long-term cumulative drift, but it is very prone to generating high-frequency visual measurement jump noise when encountering strong light, backlight, or when the cable is briefly blocked by leaves. At this time, by reasonably configuring the fusion weight coefficient ℏ (for example, taking ℏ=0.3, that is, giving 70% confidence to the reference visual data), when the pure kinematic model becomes blindly confident due to slippage, the high-weighted visual feedback will immediately forcefully increase the comprehensive error through the second weighting term, thereby waking up the control system to correct the deviation.

[0047] In this embodiment, a precise two-dimensional transformation matrix is ​​constructed by rigorously extracting the translational and rotational rigid body installation deviations of the camera and the machine's center of mass. A complementary weighted fusion architecture is introduced within a unified physical coordinate system to address the chassis kinematic blind push error and visual observation deviation, breaking down the heterogeneous barriers between optical and mechanical sensors in terms of reference standards and data sampling characteristics. This multi-source fusion mechanism not only mathematically eliminates the inherent blind spots of individual sensors, fundamentally eliminating the persistent problem of odometer drift caused by long-term operation of the robot dog in complex and uneven terrain, but also extracts a comprehensive error target point for the entire inspection system with extremely high fidelity, zero long-term drift, and smooth noise resistance. This provides a precise front-end closed-loop data source for subsequently driving the underlying sliding mode controller to perform highly robust anti-disturbance attitude correction.

[0048] Step S4: Input the current integrated tracking error and the current desired motion speed into the sliding mode controller based on proportional-integral ratio. The sliding mode controller constructs a sliding surface with the current integrated tracking error and substitutes the sliding surface into the saturation function to reduce chattering to solve the sliding mode control law. The underlying motion control quantity including linear velocity and angular velocity is calculated based on the sliding mode control law. In a preferred embodiment, a sliding surface is constructed using the integrated tracking error, and the sliding surface is substituted into a saturation function that reduces chattering to solve for the sliding mode control law, specifically including: The sliding mode control law is calculated using the following formula: ; In the formula, This includes the underlying motion control variables, which contain linear velocity and angular velocity. The inverse matrix of the control input coefficient matrix is ​​preset in the chassis kinematic model of the robot dog. For the current desired speed of motion, To track error variables, This is an estimate of the unknown disturbance. This refers to the proportional coefficient of the proportional adjustment term in the sliding mode controller. The integral coefficient of the integral adjustment term in the sliding mode controller is... As an additional input including the saturation function, t represents the current control time; in the actual closed-loop solution of the sliding mode controller, the tracking error variable Extract the value of the current comprehensive tracking error and perform calculations.

[0049] Preferably, the additional input containing the saturation function is determined by the following: ; In the formula, η is a given sliding surface constant, t is time, and s is the sliding surface constructed based on the current integrated tracking error, and satisfies the following conditions: ; To prevent interference with observation errors and satisfy the following conditions , The boundary layer thickness of the sliding surface. The coefficients of the convergence term, Let be the saturation function for reducing chattering.

[0050] Preferably, in the actual computational solution process of the sliding mode controller constructed based on proportional-integral methods, its mathematical core lies in first artificially pre-setting a virtual track (i.e., a sliding surface) in the system state space that can cause the error to decay exponentially to zero. In this embodiment, the rigorous mathematical expression and physical meaning of constructing the first-order linear sliding surface s(t) are as follows: ; In the formula, s(t) represents the currently constructed sliding surface function value; when the system is controlled such that s(t) ≡ 0, it means that the robot dog's state has completely fallen onto the preset sliding manifold surface; e(t) represents the tracking error variable of the control system. As mentioned earlier, in order to break down the barriers of multi-source heterogeneous sensors, this variable is strictly extracted from the current comprehensive tracking error fused in step S3 when it is actually assigned a value. ; The value represents the rate of change (i.e., the derivative) of the integrated tracking error over time, which physically characterizes the relative speed trend of the robot dog's current deviation from the target trajectory; η is a given sliding surface constant (and η>0), which mathematically determines the exponential decay rate of the error converging to zero; the larger the value of η, the faster the error is cleared to zero once the system reaches the sliding surface.

[0051] Furthermore, to ensure that the robot dog's actual motion state always approaches the aforementioned sliding surface s(t)=0 and to completely resist external terrain disturbances, this embodiment designs a system that includes a basic equivalent control law and robust additional inputs. The complete control closed loop. To clearly demonstrate the decoupling logic of the underlying computing power, the additional input... The complete calculation formula can be rigorously expressed as: ; In the aforementioned additional inputs and overall sliding mode control law (i.e., the aforementioned...) The deep cybernetics interpretation of each parameter and operational term in the formula is as follows: The specific command vector that is ultimately output to the chassis controller strictly contains the expected linear velocity V and angular velocity Ω. The inverse kinematics solution matrix for the chassis (i.e., the inverse matrix of the preset control input coefficient matrix) is used to map and restore the abstract sliding mode control torque to the specific wheel-end or joint desired velocity vector that the physical chassis can directly execute. (corresponding to the aforementioned parameter x(t)) represents the desired operating speed, which serves as a feedforward term for the control system to provide the basic driving force under error-free conditions; This constitutes a unique proportional-integral (PI) control loop; where the proportional term... It is responsible for providing instantaneous linear recovery thrust based on the current error magnitude; integral term Responsible for historical accumulated errors By performing integral compensation in the time dimension, this design overcomes the shortcomings of traditional pure sliding mode control in eliminating the steady-state error of the system, ensuring that the robot dog can still closely follow the center line of the cable after a long period of inspection. This is an estimate of the unknown disturbance; The mathematical constraint representing the interference observation error is the actual physical disturbance. f(t) and estimated value (The residuals between them). Due to the extreme complexity of the field environment, it is impossible for the observer to predict all disturbances 100% accurately, therefore This represents those sudden, unpredictable, and intense mechanical shocks; κ is the linear convergence term in the exponential convergence law, and κ is the positive convergence coefficient. When the system state is far from the sliding surface, this term can provide a huge pull proportional to the degree of deviation, forcing the robot dog's state to be quickly pulled back to the vicinity of the sliding surface. This is the most crucial anti-bounce robust compensation term in this embodiment; δ is the set boundary layer thickness; sat(s) is the saturation function for reducing bounce, mathematically defined as: when |s|>δ, sat(s) = sgn(s) (i.e., performing brute-force, full-amplitude switching control); when |s| ≤ δ, sat(s) = s / δ (i.e., performing smooth, linear transition control). In actual high-risk field cable inspection physical scenarios, this control architecture plays a crucial role in the underlying takeover: suppose a robot dog is crossing a moss-covered slope, and one of its paws suddenly slips severely. At this moment, the underlying physical system will experience an extremely large residual unknown disturbance. If traditional PID control or a conventional open-loop control mechanism is used, the system will experience significant yaw due to response lag. This lag will then lead to high-frequency oscillation commands (full left and full right rudder) due to severe error compensation, ultimately causing mechanical joint gear failure or the robot to tip over completely. However, with the control architecture of this embodiment, when slippage occurs, a comprehensive tracking error is generated. Then, the system immediately calculates that the sliding surface s(t) ≠ 0; at this time, by The resulting approach law term instantly triggers a powerful nonlinear compensation command. Because it uses the saturation function *sat(s)* instead of the traditional sign function *sgn(s)*, the control system establishes a buffer boundary layer of thickness δ on both sides of the sliding surface. When the robot dog's pose is forcibly pulled back and approaches the standard cable route, the output change of the control command automatically transforms into a smooth linear ramp transition. This mechanism, which provides strong anti-disturbance pull at the far end and smooth buffer filtering at the near end, achieves a balance between nonlinear disturbance rejection and low-level command debouncing.

[0052] In this embodiment, by strictly defining a linear sliding mode surface based on fusion error and time derivative, and deeply integrating the advantages of proportional-integral (PI) control in eliminating steady-state static error and sliding mode control (SMC) in resisting nonlinear disturbances, a saturation function with boundary layer thickness is used to replace the hard-switching logic that is prone to inducing mechanical chattering in traditional sliding mode control. This decouples the destructive force of complex terrain on the kinematic balance of the machine. This ensures that the robot dog can not only accurately identify cable lines, but also perfectly convert the expected linear angular velocity output by the upper navigation algorithm into the physical trajectory of the chassis with a highly disturbance-resistant and robust gait, thereby guaranteeing the high reliability and seismic tracking accuracy of the entire power automation inspection system.

[0053] Step S5: Based on the underlying motion control variables, control the robot dog's movement in the next moment.

[0054] Preferably, in the specific physical execution process of controlling the robot dog's movement at the next moment based on the underlying motion control quantity, since the robot dog is a high-degree-of-freedom biomimetic redundant mechanical system, the underlying motion control quantity output by the upper sliding mode controller (i.e., the combination of the three-dimensional spatial linear velocity and attitude angular velocity of the chassis center of mass) needs to be converted at the hardware level through an electromechanical decoupling and inverse kinematics mapping mechanism. Specifically, the gait scheduling module at the bottom of the system will decompose the received center of mass control quantity into the expected displacement trajectory and expected movement velocity of each foot in Cartesian space according to the current gait phase of the robot dog, and use the inverse kinematics model to accurately map the expected movement velocity of the foot into the expected rotational angular velocity and expected joint angle commands of the hip joint, thigh joint, and calf joint of the robot dog.

[0055] After receiving the control commands at the physical joint level, these commands are sent at high frequency to the underlying servo drivers distributed across the robot dog's joints. The joint servo drivers utilize a built-in closed-loop control algorithm to calculate and output the desired electromagnetic torque, thereby driving the physical motors to rotate. Thanks to the sliding mode controller's thorough filtering of high-frequency chattering components in the control commands using a saturation function in the preceding steps, the desired joint angular velocity commands received by the underlying servo drivers exhibit extremely smooth and continuous physical characteristics. This effectively avoids the risks of mechanical gear impact, rigid wear, and motor overheating and burnout caused by receiving abrupt or oscillating sudden commands at the hardware execution level.

[0056] Once all the joint motors work together to deliver the aforementioned torque, the robot dog's body completes its spatial displacement within the current control cycle in the real physical space, smoothly transitioning to the next operational state. As the physical position shifts, the robot dog's built-in chassis wheel-leg odometers and onboard vision system simultaneously resample the surrounding terrain and cable line targets, capturing new actual operating conditions and visual feedback data. At this point, the entire control system seamlessly jumps back to the initial data acquisition steps, initiating a new round of infinitely rolling closed-loop control processes—including error calculation, multi-source fusion, sliding mode solving, and reverse drive—within ultra-high-frequency control cycles, until the automated inspection task of the entire high-voltage cable is fully completed.

[0057] In actual field inspection deployment scenarios, the system architecture constructed by this invention exhibits extremely high fault tolerance and synergy. The visual sensing module provides a stable target reference for the system, while the planning algorithm layer completes dynamic obstacle avoidance and path deduction while ensuring a safe line of sight. At the same time, the underlying sliding mode control mechanism and electromechanical calculation module construct a robust physical anti-interference defense line. This defense line effectively absorbs and mitigates sudden mechanical slippage impacts caused by unstructured terrain such as foot slippage or wet mud, transforming complex underlying disturbances into smooth and flexible compensation torque, ensuring the continuity of the machine's movement and the stability of its posture.

[0058] Through the complete technical solution disclosed in this embodiment, which spans from macroscopic navigation to microscopic driving, this invention achieves a deep technical reconstruction in three dimensions: path planning layer, multi-source perception layer, and underlying execution layer. At the planning layer, by introducing a priori knowledge of cable geometry distribution and combining it with an ant colony algorithm that penalizes tortuosity, as well as a dynamic window method with hard constraints on deviation depth, the systemic flaw in traditional algorithms—which easily leads to the robot dog deviating from the inspection corridor due to excessive obstacle avoidance or blind searching—is completely plugged. At the perception layer, by performing complementary weighted fusion of visual pose deviation and chassis kinematic error under a unified physical coordinate system, the mileage drift deviation accumulated by a single sensor over long-term operation is eliminated from the data source. At the execution layer, by constructing a sliding mode control law that includes proportional-integral elements and boundary layer anti-shake characteristics, and projecting the solution to the physical drive end of the motor, the control system is directly endowed with strong nonlinear physical disturbance rejection capability and steady-state tracking accuracy. This solution overcomes the engineering and technical bottlenecks of traditional robot navigation algorithms, which are weak in resisting interference and prone to deviating from the target track in uneven terrain. It significantly improves the continuous and stable operation capability and closed-loop control reliability of power automation inspection equipment such as quadruped robot dogs in extremely harsh environments.

[0059] like Figure 2As shown, another embodiment of the present invention also provides a quadruped robot dog sliding mode tracking control device for cable inspection, including: a state acquisition module, an error calculation module, a multi-source fusion module, a sliding mode solution module, and a chassis drive module; the state acquisition module is used to acquire the current desired cable inspection path, the current desired movement speed, and the current running state of the robot dog; the error calculation module is used to calculate the current kinematic tracking error of the robot dog chassis based on the current desired cable inspection path and the current running state of the robot dog; A multi-source fusion module is used to acquire the current visual pose deviation of the robot dog relative to the cable based on the onboard camera of the robot dog. After transforming the current visual pose deviation to the robot dog coordinate system, the module performs weighted fusion with the current kinematic tracking error to obtain the current comprehensive tracking error. A sliding mode solving module is used to input the current comprehensive tracking error and the current desired motion velocity into a sliding mode controller based on proportional-integral motion. The sliding mode controller constructs a sliding surface with the current comprehensive tracking error and substitutes the sliding surface into a saturation function to reduce chattering to solve the sliding mode control law. The underlying motion control quantity, including linear velocity and angular velocity, is calculated based on the sliding mode control law. A chassis drive module is used to control the robot dog's motion at the next moment based on the underlying motion control quantity.

[0060] It is understood that the above-described device embodiments correspond to the method embodiments of the present invention, and can realize the four-legged robot dog sliding mode tracking control method for cable inspection provided by any of the above-described method embodiments of the present invention.

[0061] It should be noted that the device embodiments described above are merely illustrative, and some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can specifically be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0062] Based on the above embodiment of a quadruped robot dog sliding mode tracking control method for cable inspection, another embodiment of the present invention provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements any one of the quadruped robot dog sliding mode tracking control methods for cable inspection of the present invention.

[0063] For example, in this embodiment, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the terminal device.

[0064] The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0065] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting all parts of the terminal device via various interfaces and lines.

[0066] Based on the above-described method embodiments, another embodiment of the present invention provides a computer-readable storage medium, including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to execute the sliding mode tracking control method for a quadruped robot dog for cable inspection described in any of the above-described method embodiments of the present invention.

[0067] The modules / units integrated in the device / terminal equipment, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0068] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A sliding mode tracking control method for a quadruped robot dog for cable inspection, characterized in that, include: Obtain the current desired cable inspection path, current desired movement speed, and current operating status of the robot dog; Based on the current expected path of cable inspection and the current operating status of the robot dog, calculate the current kinematic tracking error of the robot dog chassis; based on the robot dog's onboard camera, obtain the current visual pose deviation of the robot dog relative to the cable, transform the current visual pose deviation to the robot dog coordinate system, and then perform weighted fusion with the current kinematic tracking error to obtain the current comprehensive tracking error; The current integrated tracking error and the current desired motion speed are input into a sliding mode controller based on proportional-integral ratio to generate a sliding surface. The sliding surface is then substituted into a saturation function to reduce chattering to solve for the sliding mode control law. Based on the sliding mode control law, the underlying motion control quantity, which includes linear velocity and angular velocity, is calculated. Based on the underlying motion control quantity, the motion of the robot dog at the next moment is controlled.

2. The quadruped robot dog sliding mode tracking control method for cable inspection as described in claim 1, characterized in that, The current methods for determining the desired cable inspection path include: If the current time is the initial time, then the static obstacle map of the environment and the known cable line routing information are obtained; based on the static obstacle map and the cable line routing information, the optimal global path that fits the cable line is generated by the ant colony algorithm, and the optimal global path is used as the expected path for the current cable inspection; wherein, the state transition probability heuristic function of the ant colony algorithm includes a path tortuosity penalty term. If the current moment is not the initial moment, the dynamic obstacle information of the surrounding environment is obtained through the airborne camera, and a feasible velocity space including the linear velocity and angular velocity of the robot dog is constructed based on the robot dog's current running speed, motor acceleration limit, and the safe braking distance between the robot dog and each dynamic obstacle calculated based on the dynamic obstacle information. Speed ​​sampling is performed within the feasible speed space. The current running state of the robot dog and the various speed combinations obtained by sampling are extrapolated within a preset time window to predict and generate local simulated trajectories under each speed combination. An evaluation function that includes a cable deviation penalty term is used to score the cost value of each local simulated trajectory, and the optimal local trajectory with the highest cost value score is selected as the expected path for the current cable inspection.

3. The quadruped robot dog sliding mode tracking control method for cable inspection as described in claim 2, characterized in that, An evaluation function incorporating a cable deviation penalty term is used to score the cost of each local simulated trajectory, specifically including: For each local simulated trajectory, the following evaluation factors are extracted: heading angle evaluation factor representing the alignment between the trajectory end and the target point, distance evaluation factor representing the closest distance between the trajectory and surrounding obstacles, and speed evaluation factor representing the magnitude of the movement speed. Calculate the vertical distance from the trajectory point on the local simulated trajectory to the optimal global path, and determine the ratio of the vertical distance to the maximum allowable deviation distance as the cable deviation penalty term; The basic evaluation value is obtained by multiplying the heading angle evaluation factor, the distance evaluation factor, and the speed evaluation factor by their respective weighting coefficients and then summing them. Calculate the product of the cable deviation penalty term and the corresponding weighting coefficient to obtain the weighted cable deviation penalty term; The cost score of the local simulated trajectory is calculated based on the difference between the basic evaluation value and the weighted cable deviation penalty term.

4. The quadruped robot dog sliding mode tracking control method for cable inspection as described in claim 3, characterized in that, The methods for determining the current desired speed of motion include: If the current moment is the initial moment, then the current desired speed is the preset speed; If the current time is not the initial time, the speed combination corresponding to the current cable inspection expected path is taken as the current expected movement speed.

5. The sliding mode tracking control method for a quadruped robot dog for cable inspection as described in claim 1, characterized in that, After transforming the current visual pose deviation to the robot dog coordinate system, it is weighted and fused with the current kinematic tracking error to obtain the current comprehensive tracking error. Specifically, it includes: obtaining the pre-calibrated lateral offset and installation yaw angle of the airborne camera, and constructing a calibration transformation matrix based on the lateral offset of the airborne camera and the installation yaw angle. Using the calibration transformation matrix, the current visual pose deviation defined in the coordinate system of the airborne camera is converted into the transformed visual pose deviation defined in the coordinate system of the robot dog; wherein, the current visual pose deviation specifically includes the lateral offset of the cable centerline in the image and the heading deviation. The current kinematic tracking error is multiplied by a given fusion weight coefficient to obtain the first weighting term; The converted visual pose deviation is weighted by the complementary weight coefficients corresponding to the fusion weight coefficients to obtain the second weighting term. The first weighted term and the second weighted term are summed to obtain the current integrated tracking error.

6. The sliding mode tracking control method for a quadruped robot dog for cable inspection as described in claim 1, characterized in that, The current integrated tracking error and the current desired motion speed are input into a sliding mode controller constructed based on proportional-integral ratio to generate a sliding surface. The sliding surface is then substituted into a saturation function that reduces chattering to solve for the sliding mode control law, specifically including: The sliding mode control law is calculated using the following formula: ； In the formula, This includes the underlying motion control variables, which contain linear velocity and angular velocity. The inverse matrix of the control input coefficient matrix is ​​preset in the chassis kinematic model of the robot dog. For the current desired speed of motion, To track error variables, This is an estimate of the unknown disturbance. This refers to the proportional coefficient of the proportional adjustment term in the sliding mode controller. The integral coefficient of the integral adjustment term in the sliding mode controller is... As an additional input including the saturation function, t represents the current control time; in the actual closed-loop solution of the sliding mode controller, the tracking error variable The value is the current comprehensive tracking error.

7. The quadruped robot dog sliding mode tracking control method for cable inspection as described in claim 6, characterized in that, The additional inputs containing the saturation function are determined as follows: ； In the formula, η is a given sliding surface constant, t is time, and s is the sliding surface constructed based on the current integrated tracking error, and satisfies the following conditions: ; To prevent interference with observation errors and satisfy the following conditions , The boundary layer thickness of the sliding surface is given. The coefficients of the convergence term, Let be the saturation function for reducing chattering.

8. A quadruped robot dog sliding mode tracking control device for cable inspection, characterized in that, include: The system includes a status acquisition module, an error calculation module, a multi-source fusion module, a sliding mode solution module, and a chassis drive module. The status acquisition module is used to acquire the current desired cable inspection path, the current desired movement speed, and the current operating status of the robot dog; The error calculation module is used to calculate the current kinematic tracking error of the robot dog chassis based on the expected path of the current cable inspection and the current operating status of the robot dog. The multi-source fusion module is used to obtain the current visual pose deviation of the robot dog relative to the cable based on the airborne camera of the robot dog. After transforming the current visual pose deviation to the robot dog coordinate system, it is weighted and fused with the current kinematic tracking error to obtain the current comprehensive tracking error. The sliding mode solver module is used to input the current integrated tracking error and the current desired motion speed into the sliding mode controller built on proportional-integral (PI) to generate a sliding surface. The sliding surface is then substituted into the saturation function that reduces chattering to solve the sliding mode control law. Based on the sliding mode control law, the underlying motion control quantities, including linear velocity and angular velocity, are calculated. The chassis drive module is used to control the robot dog's movement in the next moment based on the underlying motion control parameters.

9. A terminal device, characterized in that, The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the sliding mode tracking control method for a quadruped robot dog for cable inspection as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, include: A stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the sliding mode tracking control method for a quadrupedal robot dog for cable inspection as described in any one of claims 1-7.

Images