Robot dog inspection path intelligent planning and dynamic adjusting method, system and device and medium

Abstract

The invention discloses a robot dog inspection path intelligent planning and dynamic adjustment method, system and device and a medium, and belongs to the technical field of robot path planning, and the method comprises the steps: calculating a task emergency degree index according to a weight coefficient of an inspection point and a time constraint, and determining a task priority; constructing a hierarchical electronic map containing the terrain difficulty coefficient and the moving cost; performing global path planning through a comprehensive cost function by adopting an A star algorithm; environment information is collected in real time through multiple sensors, and the dynamic obstacle layer is updated; performing real-time trajectory optimization by adopting a local path planning mode; selecting local path correction or global path re-planning according to the path execution state score; energy needed for completing the remaining tasks is predicted, and a charging path is planned when necessary. According to the invention, multi-target adaptive optimization of path planning is realized, a coordination mechanism of global planning and local adjustment is established, and the execution efficiency and reliability of inspection tasks are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of robot path planning, in particular to a method, system, device and medium for intelligent planning and dynamic adjustment of a robot dog inspection path. BACKGROUND

[0002] As a typical representative of quadruped robots, robot dogs have strong terrain adaptability and flexible movement, and are widely used in power inspection, industrial park inspection, hazardous environment detection and other scenarios. The existing robot dog inspection system has the following deficiencies in path planning: Most existing global path planning methods use a single optimization objective, such as the shortest path or the shortest time, without considering multiple factors such as task urgency, energy consumption and terrain difficulty, resulting in low efficiency of the planned path in actual execution. For example, for high-priority emergency inspection tasks, the system still plans the path according to the energy optimization strategy, delaying the inspection time of critical equipment.

[0003] The existing local path adjustment and global path planning are independent of each other. When the robot dog deviates from the global path due to obstacles, the system lacks an effective deviation evaluation mechanism and cannot determine whether to take local correction or global re-planning, resulting in frequent unnecessary re-planning or continuous path deviation, affecting the inspection efficiency.

[0004] Most existing energy management strategies are passive warnings. When the power is insufficient, it only prompts to return to charging, and cannot dynamically predict the energy demand according to the remaining task amount, terrain difficulty and actual energy consumption, resulting in the robot dog stopping due to power depletion during inspection and failing to complete the inspection task. SUMMARY

[0005] In view of the above problems, the present application provides a method, system, device and medium for intelligent planning and dynamic adjustment of a robot dog inspection path.

[0006] Therefore, the technical problems solved by the present application are: how to realize multi-objective adaptive optimization of the robot dog inspection path, how to establish a coordination mechanism between global planning and local adjustment, and how to realize active prediction and charging decision of energy to improve the execution efficiency and reliability of the inspection task. Specifically, the problems of dynamically adjusting the path optimization objective in different inspection tasks of different urgency, the problem of reasonably selecting local correction or global re-planning when the path deviates, and the problem of predicting the energy demand of the remaining task according to the terrain difficulty and actual energy consumption need to be solved.

[0007] To solve the above technical problems, the present application provides the following technical solution: a method for intelligent planning and dynamic adjustment of a robot dog inspection path, comprising, Obtain inspection task instructions, which include a list of inspection point coordinates, weight coefficients, and time constraints. Calculate the task urgency index based on the weight coefficients and time constraints of the inspection points, and determine the task priority based on the task urgency index. The inspection area is divided into a grid, each grid is assigned terrain attribute information, the terrain difficulty coefficient of each grid and the movement cost between adjacent grids are calculated, a layered electronic map is constructed and a dynamic obstacle layer is maintained. Based on the hierarchical electronic map and the list of inspection point coordinates, the A* algorithm is used to perform global path planning through a comprehensive cost function to generate a global inspection path from the starting point through each inspection point to the end point. During the inspection task performed by the robot dog, environmental information is collected in real time through lidar, vision sensors and inertial measurement unit to identify obstacles and update the dynamic obstacle layer, and to obtain the robot dog's current position; Based on real-time environmental information and global inspection path, a local path planning method is used to optimize the real-time trajectory under speed constraints, select the optimal speed command, and calculate the lateral deviation of the robot dog's current position from the global path. The path execution status score is calculated based on lateral deviation, path obstruction degree and task progress. When the path execution status score is lower than the threshold, local path correction or global path replanning is selected based on the distance from the current position to the next inspection point. The robot dog's remaining battery power is monitored in real time. Based on the energy consumption statistics of the route already traveled, the energy required to complete the remaining tasks is predicted. When the predicted energy exceeds the safety threshold of the remaining battery power, a route to the charging station is planned. The speed command is converted into joint angle control commands for the robot dog's four legs, driving the robot dog to move along the planned path.

[0008] As a preferred embodiment of the intelligent planning and dynamic adjustment method for robot dog inspection paths described in this invention, the task urgency index includes: Obtain the weight coefficient for each inspection point, where the weight coefficient reflects the importance of the inspection point; Obtain the latest arrival time and task reception time for each inspection point, and determine the time constraints. The urgency index of a task is determined by combining the weighting coefficient of each inspection point with the time constraints. The weighting coefficient and the time constraints together affect the urgency of the task, and the inspection point with the tighter the time constraint contributes more to the urgency of the task. Task priority is determined based on the urgency index. Tasks with high urgency are classified as high-priority tasks, while those with low urgency are classified as routine tasks.

[0009] The beneficial effects of this preferred technical solution are as follows: by comprehensively considering the weighting coefficient of each inspection point and the time constraint, the urgency index of the task is determined. Both the weighting coefficient and the time constraint jointly influence the urgency of the task; the tighter the time constraint, the greater the contribution of the inspection point to the urgency, thus achieving rapid response to urgent inspection tasks. Task priority is determined based on the urgency index, and different path planning strategies are used for high-priority tasks and routine tasks, improving the system's ability to handle tasks of different priorities and ensuring timely inspection of important inspection points.

[0010] As a preferred embodiment of the intelligent planning and dynamic adjustment method for robot dog inspection paths described in this invention, the calculation of the terrain difficulty coefficient of each grid and the movement cost between adjacent grids includes: Obtain the slope information of the grid, where the slope information represents the degree of inclination of the terrain; Obtain the surface type of the grid, which includes hardened surfaces, grass, gravel, and mud. Obtain obstacle markers from the grid, where each marker indicates the traversability of an obstacle. The terrain difficulty coefficient of the grid is determined based on slope information, surface type, and obstacle markings; The movement cost between adjacent grids is determined based on the terrain difficulty coefficient of adjacent grids and the distance between grids.

[0011] As a preferred embodiment of the intelligent planning and dynamic adjustment method for robot dog inspection paths described in this invention, the comprehensive cost function includes: Calculate the actual cumulative cost from the starting point to the current node, where the actual cumulative cost takes into account both path length and energy consumption; The weights of path length and energy consumption in the actual cumulative cost are determined based on task priority. High-priority tasks focus on path length, while regular tasks balance path length and energy consumption. Calculate the heuristic estimated cost from the current node to the target point, whereby the heuristic estimated cost takes into account the distance between nodes and the difficulty of the terrain along the way; The combined cost of the current node is obtained by combining the actual cumulative cost and the heuristically estimated cost. Select the node with the lowest overall cost and expand until the target point is reached.

[0012] As a preferred embodiment of the intelligent planning and dynamic adjustment method for robot dog inspection paths described in this invention, the local path planning method includes: The executable velocity search space is determined based on the robot dog's current speed and acceleration constraints; Generate multiple candidate velocities within the velocity search space; For each candidate velocity, predict the robot dog's trajectory in the future prediction time domain; Check the collision situation between the predicted trajectory and obstacles, and eliminate candidate speeds that pose a collision risk; Evaluate the ability of feasible candidate velocities to approach targets, safe distances from obstacles, and speed efficiency; The candidate speed with the best overall evaluation is selected as the control command.

[0013] The beneficial effects of this preferred technical solution are as follows: By determining the executable speed search space based on the robot dog's current speed and acceleration constraints, generating multiple candidate speeds within the speed search space, and predicting the trajectory of each candidate speed, candidate speeds with collision risks are eliminated, ensuring the safety of local path planning. The feasible candidate speeds are evaluated for their ability to approach the target, safe distance from obstacles, and speed efficiency. The candidate speed with the best comprehensive evaluation is selected as the control command, achieving real-time local path planning and obstacle avoidance capabilities, and improving the robot dog's movement flexibility in complex dynamic environments.

[0014] As a preferred embodiment of the intelligent planning and dynamic adjustment method for robot dog inspection paths described in this invention, the path execution status score includes: The degree of path deviation is assessed based on the lateral deviation of the robot dog's current position from the global path; Assess the degree of path obstruction based on the time the robot dog spends at its current location; Assess the progress of the task based on the number of inspection points visited and the total number of inspection points. The path execution status score is determined by combining lateral deviation, path obstruction degree, and task progress. When the path execution status score is lower than the threshold, determine the distance from the current position to the next inspection point; If the distance is short and the deviation is the main problem, perform local path correction and replan the path segment from the current location to the next inspection point; If the distance is far or a large area of ​​obstacles is detected, the newly detected obstacle information is updated to the dynamic obstacle layer, and global path replanning is performed to generate a complete path with the current position as the new starting point.

[0015] As a preferred embodiment of the intelligent planning and dynamic adjustment method for robot dog inspection paths described in this invention, the energy required to predict and complete the remaining tasks includes: Energy consumption per unit distance is calculated based on the energy consumed and the distance traveled. Predict the energy required to complete the remaining tasks based on the length of the remaining path and the difficulty of the terrain along the way; Compare the remaining power with the predicted energy requirement to determine if the energy is sufficient; If the remaining battery power is insufficient to complete the remaining tasks and to maintain a safety margin, locate the nearest charging station. Determine if the remaining battery power is sufficient to reach the charging station; If there is sufficient remaining power, plan the route to the charging station and pause the current inspection task. Continue the task after charging. If the remaining battery power is insufficient to reach a charging station, an alarm will be issued and the machine will be shut down in the nearest safe location.

[0016] This invention provides a system for intelligent planning and dynamic adjustment of inspection paths for robotic dogs.

[0017] To solve the above technical problems, the present invention provides the following technical solution: a robot dog inspection path intelligent planning and dynamic adjustment system, comprising: a task parsing module, used to receive inspection task instructions, calculate the task urgency index according to the weight coefficient of the inspection point and the time constraint, and determine the task priority; The electronic map construction module is used to divide the inspection area into a grid, calculate the terrain difficulty coefficient and movement cost of the grid, construct a layered electronic map, and maintain a dynamic obstacle layer. The global path planning module is used to perform global path planning using the A* algorithm through a comprehensive cost function, and generate global inspection paths. The real-time environmental perception module is used to collect environmental information in real time through lidar, vision sensors and inertial measurement unit, identify obstacles and update the dynamic obstacle layer, and obtain the current position of the robot dog; The local path adjustment module is used to perform real-time trajectory optimization under speed constraints using local path planning, select the optimal speed command, and calculate the lateral deviation. The path evaluation and replanning module is used to calculate the path execution status score. When the path execution status score is lower than the threshold, it selects local path correction or global path replanning based on the distance from the current position to the next inspection point. The energy management module is used to monitor the remaining power in real time, predict the energy required to complete the remaining tasks, and plan the route to the charging station when the predicted energy exceeds the safety threshold of the remaining power. The execution control module is used to convert speed commands into joint angle control commands for the robot dog's four legs, driving the robot dog to move along the planned path.

[0018] The present invention provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the intelligent planning and dynamic adjustment method for a robot dog inspection path.

[0019] The present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the aforementioned method for intelligent planning and dynamic adjustment of inspection paths for a robot dog.

[0020] The beneficial effects of this invention are as follows: This invention quantifies task priority through a task urgency index and dynamically adjusts the weights of path length and energy consumption in the comprehensive cost function, achieving multi-objective adaptive optimization of path planning and improving its adaptability to different task requirements. This invention comprehensively evaluates lateral deviation, path obstruction degree, and task progress through path execution status scoring, selecting local path correction or global path replanning based on the distance from the current position to the next inspection point, establishing a coordination mechanism between global planning and local adjustment, and avoiding frequent unnecessary replanning. This invention predicts the energy required to complete the remaining tasks by statistically analyzing energy consumption per unit distance and combining the remaining path length and terrain difficulty, achieving proactive energy management and preventing the robot from shutting down due to battery depletion during inspections. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 The above is a flowchart of an intelligent planning and dynamic adjustment method for robot dog inspection paths provided in one embodiment of the present invention. Detailed Implementation

[0023] To make the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0024] Example 1, referring to Figure 1 This is one embodiment of the present invention, which provides a method for intelligent planning and dynamic adjustment of inspection paths for robot dogs, including: Step 1: Obtain the inspection task instruction, which includes a list of inspection point coordinates, weight coefficients, and time constraints. Calculate the task urgency index based on the weight coefficients and time constraints of the inspection points, and determine the task priority based on the task urgency index. Step 2: Divide the inspection area into a grid, assign terrain attribute information to each grid, calculate the terrain difficulty coefficient of each grid and the movement cost between adjacent grids, construct a layered electronic map and maintain a dynamic obstacle layer; Step 3: Based on the layered electronic map and the list of inspection point coordinates, the A* algorithm is used to perform global path planning through a comprehensive cost function to generate a global inspection path from the starting point through each inspection point to the end point; Step 4: During the inspection task performed by the robot dog, environmental information is collected in real time through LiDAR, vision sensors and inertial measurement unit to identify obstacles and update the dynamic obstacle layer, and to obtain the current position of the robot dog; Step 5: Based on real-time environmental information and the global inspection path, a local path planning method is used to optimize the real-time trajectory under speed constraints, select the optimal speed command, and calculate the lateral deviation of the robot dog's current position from the global path. Step 6: Calculate the path execution status score based on lateral deviation, path obstruction degree, and task progress. When the path execution status score is lower than the threshold, select local path correction or global path replanning based on the distance from the current position to the next inspection point. Step 7: Monitor the robot dog's remaining battery power in real time, predict the energy required to complete the remaining tasks based on the energy consumption statistics of the route already traveled, and plan a route to the charging station when the predicted energy exceeds the safety threshold of the remaining battery power. Step 8: Convert the speed command into joint angle control commands for the robot dog's four legs, and drive the robot dog to move along the planned path.

[0025] This method is applicable to scenarios such as power line inspection, industrial park inspection, and hazardous environment detection. Through multi-objective adaptive optimization and a global-local coordination mechanism, it improves the execution efficiency and reliability of inspection tasks. The method first calculates the task urgency index and determines the task priority based on the inspection task instructions. Then, it constructs a layered electronic map containing terrain difficulty information. Based on the task priority, it performs global path planning using a comprehensive cost function. During execution, it uses multiple sensors to perceive the environment in real time and optimizes the trajectory using local path planning. Based on the path execution status score, it selects between local correction and global replanning. Simultaneously, it monitors the energy status and plans charging paths when necessary. Finally, it converts speed commands into joint angle control commands to drive the robot dog's movement.

[0026] A multi-objective optimization model was established, comprehensively considering path length, energy consumption, terrain difficulty, and time constraints. A dynamic weight adjustment mechanism adaptively balances various indicators based on task priority. A hierarchical path planning architecture was designed, with the upper layer providing macroscopic navigation direction through global planning, and the lower layer handling dynamic obstacles and real-time optimization through local adjustments. A feedback coordination mechanism is established between the two layers through a path evaluation module, triggering global replanning when local deviations exceed a threshold. Real-time environmental information is collected through multi-sensor fusion to update the dynamic obstacle layer, and local trajectory optimization is performed using a dynamic window method, enabling the robot dog to adapt to the complex and ever-changing outdoor environment of the power distribution network.

[0027] Example 2, an embodiment of the present invention, provides a method for intelligent planning and dynamic adjustment of inspection paths for robot dogs based on the previous embodiment, including: In step one, the urgency index of the task is calculated based on the weighting coefficients of the inspection points and the time constraints, including the following steps A1-A4: A1: Obtain the weight coefficient of each inspection point, which reflects the importance of the inspection point; A2: Obtain the latest arrival time and task reception time for each inspection point, and determine the time constraints; A3: The task urgency index is determined by comprehensively considering the weight coefficient and time constraints of each inspection point. The weight coefficient and time constraints jointly affect the task urgency, and the inspection point with the tighter the time constraint contributes more to the task urgency. A4: Task priority is determined based on the urgency index. Tasks with high urgency are classified as high-priority tasks, while those with low urgency are classified as routine tasks.

[0028] In this embodiment of the application, in A3, the task urgency index is obtained by: acquiring the weight coefficient and time constraint of each inspection point; calculating the time difference between the latest arrival time and the task reception time for each inspection point; applying a time decay function to obtain the time decay value based on the time difference, with a larger time decay value for a smaller time difference; multiplying the weight coefficient of each inspection point by the corresponding time decay value to obtain the weighted decay value of that inspection point; averaging the weighted decay values ​​of all inspection points to obtain the task urgency index; comparing the task urgency index with a high-priority threshold and a regular task threshold; determining a high-priority task when the task urgency index is greater than the high-priority threshold, and a regular task when the task urgency index is less than or equal to the regular task threshold; increasing the weight of path length and decreasing the weight of energy consumption for high-priority tasks in subsequent global path planning; and making the weights of path length and energy consumption similar for regular tasks in subsequent global path planning.

[0029] In an optional implementation, in A3, the task urgency index can be obtained by: calculating the time difference between the current time and the latest arrival time of each inspection point, taking the reciprocal of the time difference to obtain the time constraint urgency, multiplying the weight coefficient of each inspection point by the corresponding time constraint urgency to obtain the weighted urgency, and summing and normalizing the weighted urgency of all inspection points to obtain the task urgency index.

[0030] In another optional implementation, in A3, the task urgency index can also be obtained by: dividing each inspection point into three levels—high-weight, medium-weight, and low-weight—based on the weight coefficient values; calculating the proportion of inspection points at each level and the average time constraint urgency; multiplying the proportion of each level by the average time constraint urgency and assigning the highest importance coefficient to high-weight points, the medium importance coefficient to medium-weight points, and the lowest importance coefficient to low-weight points; and summing these to obtain the task urgency index.

[0031] In step two, the terrain difficulty coefficient for each grid and the movement cost between adjacent grids are calculated, including the following steps B1-B5: B1: Obtain the slope information of the grid, where the slope information represents the degree of inclination of the terrain; B2: Obtain the surface type of the grid, which includes hardened surfaces, grass, gravel, and mud. B3: Obtain obstacle markers from the grid, where the obstacle markers indicate the traversability of the obstacles; B4: Determine the terrain difficulty coefficient of the grid based on slope information, surface type, and obstacle markings; B5: Determine the movement cost between adjacent grids based on the terrain difficulty coefficient and the distance between grids.

[0032] In this embodiment of the application, in B5, the movement cost between adjacent grids is determined by: obtaining the grid's slope information, surface type, and obstacle markers; comparing the slope information with the maximum passable slope to obtain a normalized slope value; determining the surface type coefficient based on the surface type (the surface type coefficient is smallest for hardened surfaces and largest for muddy surfaces); determining the obstacle coefficient based on the obstacle markers (the obstacle coefficient is zero when there are no obstacles and largest when there are large impassable obstacles); and summing the normalized slope value, surface type coefficient, and obstacle coefficient after assigning weights to them to obtain the grid's terrain difficulty coefficient. The movement cost between adjacent grids is determined by: obtaining the basic distance between two adjacent grids (the basic distance is determined based on the relative position between grids; the basic distance between adjacent grids (up, down, left, and right) is equal to the grid's side length; and the basic distance between diagonally adjacent grids is equal to the square root of the grid's side length); obtaining the terrain difficulty coefficients of the starting grid and the target grid; and combining the basic distance with the terrain difficulty coefficients of the starting grid and the target grid to calculate the movement cost (the higher the terrain difficulty coefficient, the higher the movement cost).

[0033] In an optional implementation, in B5, the movement cost between adjacent grids can be determined by: obtaining the base distance between two adjacent grids and the terrain difficulty coefficient of the target grid, and multiplying the base distance by the terrain difficulty coefficient of the target grid to obtain the movement cost.

[0034] In another alternative implementation, in B5, the movement cost between adjacent grids can also be determined by: obtaining the base distance between two adjacent grids, the slope information of the starting grid and the target grid, judging whether the slope exceeds the comfortable climbing angle based on the robot dog's climbing ability, introducing a terrain adaptation penalty coefficient when the slope exceeds the comfortable climbing angle, and combining the base distance with the terrain difficulty coefficients of the starting grid and the target grid and then multiplying it by the terrain adaptation penalty coefficient to obtain the movement cost.

[0035] In step three, based on the layered electronic map and the list of inspection point coordinates, the A* algorithm is used to perform global path planning through a comprehensive cost function, generating a global inspection path from the starting point through each inspection point to the endpoint, including the following steps C1-C5: C1: Calculate the actual cumulative cost from the starting point to the current node, which takes into account both path length and energy consumption; C2: Determine the weight of path length and energy consumption in the actual cumulative cost based on task priority. High-priority tasks focus on path length, while regular tasks balance path length and energy consumption. C3: Calculate the heuristic estimated cost from the current node to the target point, where the heuristic estimated cost takes into account the distance between nodes and the difficulty of the terrain along the way; C4: Combine the actual cumulative cost and the heuristically estimated cost to obtain the comprehensive cost of the current node; C5: Select the node with the lowest overall cost and expand until the target point is reached.

[0036] In this embodiment, in step three, the comprehensive cost function is calculated by: calculating the actual cumulative cost of the current node, which includes the actual cumulative cost of the parent node, the current movement cost, and the current movement energy consumption; determining the weight coefficients of the movement cost and the movement energy consumption based on task priority, where the weight coefficient of the movement cost is greater than that of the movement energy consumption for high-priority tasks and similar for regular tasks; multiplying the movement cost and movement energy consumption by their respective weight coefficients and summing the results, and adding the sum to the actual cumulative cost of the parent node to obtain the actual cumulative cost of the current node; calculating the heuristic estimated cost from the current node to the target point, which is calculated based on the straight-line distance between the current node and the target point and the average value of the terrain difficulty coefficient along the route; combining the straight-line distance with the average value of the terrain difficulty to obtain the heuristic estimated cost; and adding the actual cumulative cost to the heuristic estimated cost to obtain the comprehensive cost of the current node.

[0037] In an optional implementation, in step three, the comprehensive cost function can be obtained by adding a path smoothness cost to the actual cumulative cost and the heuristic estimated cost. The path smoothness cost is calculated based on the angle between the direction of the line connecting the current node and the parent node and the direction of the line connecting the parent node and the grandparent node. The larger the angle, the higher the path smoothness cost. The comprehensive cost is obtained by adding the actual cumulative cost, the heuristic estimated cost and the path smoothness cost together.

[0038] In another optional implementation, in step three, the comprehensive cost function can also be obtained by: adding a path safety cost to the actual cumulative cost and the heuristic estimated cost. The path safety cost is calculated based on the distance between the current node and the nearest obstacle. When the distance is less than the safety buffer distance, the path safety cost increases. The comprehensive cost is obtained by adding the actual cumulative cost, the heuristic estimated cost, and the path safety cost together.

[0039] In step five, based on real-time environmental information and the global inspection path, a local path planning method is used to optimize the real-time trajectory under speed constraints, select the optimal speed command, and calculate the lateral deviation of the robot dog's current position from the global path, including the following steps D1-D6: D1: Determine the executable velocity search space based on the robot dog's current speed and acceleration constraints; D2: Generate multiple candidate velocities within the velocity search space; D3: For each candidate velocity, predict the robot dog's trajectory in the future prediction time domain; D4: Check the collision situation between the predicted trajectory and obstacles, and eliminate candidate speeds that pose a collision risk; D5: Evaluate the ability of feasible candidate velocities to approach the target, their safe distance from obstacles, and their speed efficiency; D6: Select the candidate speed with the best overall evaluation as the control command.

[0040] In this embodiment, step five involves the local path planning method as follows: A speed search space is determined based on the robot dog's current linear velocity, current angular velocity, maximum linear acceleration, maximum angular acceleration, and control cycle. The linear velocity range of the speed search space is the sum of the current linear velocity minus the product of the maximum linear acceleration and the control cycle to the sum of the current linear velocity and the maximum linear acceleration and the control cycle. The angular velocity range is the sum of the current angular velocity minus the product of the maximum angular acceleration and the control cycle to the sum of the current angular velocity and the maximum angular acceleration and the control cycle. The speed search space is then discretely sampled to generate multiple candidate velocities. For each candidate... The robot dog's trajectory is predicted within the prediction time domain. The distance between each predicted position point on the predicted trajectory and obstacles is checked. If the distance is less than the safe distance, the candidate speed is excluded. For feasible candidate speeds, heading evaluation, distance evaluation, and speed evaluation are calculated. The heading evaluation is calculated based on the consistency of the direction of the end point of the predicted trajectory with the next path point of the global path. The distance evaluation is calculated based on the distance between the predicted trajectory and the nearest obstacle. The speed evaluation is calculated based on the ratio of the candidate speed to the maximum speed. The three evaluations are weighted and summed to obtain the comprehensive evaluation of the candidate speed. The candidate speed with the highest comprehensive evaluation is selected as the control command.

[0041] In an optional implementation, in step five, the local path planning method can be achieved by: adding an acceleration smoothness evaluation to the comprehensive evaluation of candidate speeds. The acceleration smoothness evaluation is calculated based on the difference between the candidate speed and the current speed. The larger the difference, the lower the acceleration smoothness evaluation. The comprehensive evaluation is obtained by assigning weights to the heading evaluation, distance evaluation, speed evaluation, and acceleration smoothness evaluation respectively and then summing them.

[0042] In another optional implementation, in step five, the local path planning method can also be achieved by: adding an energy consumption evaluation to the comprehensive evaluation of candidate speeds. The energy consumption evaluation is calculated based on the predicted energy consumption corresponding to the candidate speed. The higher the predicted energy consumption, the lower the energy consumption evaluation. The comprehensive evaluation is obtained by assigning weights to the heading evaluation, distance evaluation, speed evaluation, and energy consumption evaluation respectively and then summing them.

[0043] In step six, a path execution status score is calculated based on lateral deviation, path obstruction level, and task progress. When the path execution status score is below a threshold, local path correction or global path replanning is selected based on the distance from the current position to the next inspection point, including the following steps E1-E7: E1: Assess the degree of path deviation based on the lateral deviation of the robot dog's current position from the global path; E2: Assess the degree of path obstruction based on the robot dog's dwell time at the current location; E3: Assess the progress of the task based on the number of inspection points visited and the total number of inspection points; E4: Determine the path execution status score by comprehensively considering lateral deviation, path obstruction degree, and task progress; E5: When the path execution status score is lower than the threshold, determine the distance from the current position to the next inspection point; E6: If the distance is short and the deviation is the main problem, perform local path correction and replan the path segment from the current location to the next inspection point; E7: If the distance is far or a large area of ​​obstacles is detected, the newly detected obstacle information will be updated to the dynamic obstacle layer, and a global path replanning will be performed to generate a complete path with the current position as the new starting point.

[0044] In step seven, the robot dog's remaining battery power is monitored in real time. Based on the energy consumption statistics of the already traveled path, the energy required to complete the remaining tasks is predicted. When the predicted energy exceeds the safety threshold of the remaining battery power, a route to the charging station is planned, including the following steps F1-F7: F1: Calculates energy consumption per unit distance based on consumed energy and distance traveled; F2: Predict the energy required to complete the remaining tasks based on the length of the remaining path and the difficulty of the terrain along the way; F3: Compare the remaining power with the predicted energy requirement to determine if the energy is sufficient; F4: When the remaining battery power is insufficient to complete the remaining tasks and to retain a safety margin, locate the nearest charging station; F5: Determine if the remaining battery power is sufficient to reach a charging station; F6: If the remaining battery power is sufficient, plan the route to the charging station and pause the current inspection task. Continue execution after charging. F7: If the remaining battery power is insufficient to reach a charging station, an alarm will be issued and the machine will find a safe location to shut down.

[0045] Example 3, an embodiment of the present invention, provides a method for intelligent planning and dynamic adjustment of inspection paths for a robot dog based on the previous embodiment, including: In step one, a distribution network inspection task instruction is received. This instruction includes the inspection area identifier, a list of inspection point coordinates, priority information, and time constraints. The task parsing module first performs structured processing on the input information and extracts key parameters. Assuming the inspection task contains n inspection points, the information of the i-th inspection point is represented as follows: In the formula and Here are the planar coordinates of the inspection points, in meters. This is the weighting coefficient for the inspection point, reflecting its importance. The value ranges from 0.1 to 1.0, with important equipment inspection points having a weight close to 1.0, and general inspection points having a weight of around 0.5. The latest arrival time required for this point is specified in timestamp format. The system calculates the task urgency index U based on the inspection point weights and time constraints. (1) In the formula The moment the task is received. The time decay coefficient is set to 0.01 min⁻¹ in this method, ensuring that tasks with tighter time constraints have a higher urgency index. When U > 0.7, the task is considered high-priority and should prioritize time optimization. For routine tasks, energy consumption optimization can be given more consideration. The calculated urgency index U will be used to determine the weighting coefficients for multi-objective optimization in the third step of global path planning. The task parsing module also performs preliminary sorting of the inspection point sequence, and uses a strategy combining proximity principle and time window constraints to generate inspection point visit sequence suggestions, which serve as input for global path planning.

[0046] In step two, the electronic map module integrates data from the power distribution network GIS system, satellite remote sensing imagery, and historical inspection records to construct a layered electronic map that includes terrain information, obstacle distribution, and accessibility assessment. The map uses a rasterized representation method, dividing the inspection area into... The grid is designed so that each grid cell has a side length of 2 meters, matching the typical size of a robot dog. The attribute information of grid cell (i,j) includes elevation. ,slope Surface type and obstacle markers The system calculates the terrain difficulty coefficient for each grid cell. : (2) In the formula The maximum slope that the robot dog can traverse is set to 35 degrees. The coefficients for different land surface types are: 0.1 for paved surfaces, 0.3 for grasslands, 0.5 for gravel surfaces, and 0.8 for muddy surfaces. This is the obstacle coefficient: 0 for no obstacles, 0.5 for small, traversable obstacles, and 1.0 for large, impassable obstacles. Difficulty coefficient. The value of ranges from 0 to 1, with larger values ​​indicating higher passage difficulty. The system further calculates the movement cost between adjacent grids. The basic cost of moving from grid (i,j) to the adjacent grid (p,q) is the Euclidean distance between the center points of the two grids, denoted as . The distance between adjacent units (top, bottom, left, and right) is 2 meters, and the distance between diagonally adjacent units is 2.83 meters. This accounts for the actual movement cost after considering terrain difficulty. The calculation is as follows: (3) In the formula and These represent the terrain difficulty coefficients for the starting and target grids, respectively. This movement cost comprehensively reflects the impact of path length and terrain; areas with higher difficulty coefficients have significantly increased movement costs, guiding the path planning algorithm to select easier routes. The electronic map module also maintains a dynamic obstacle layer, initially empty, which is updated in real-time by the environmental perception module during inspection execution. The completed layered electronic map and the movement cost for each grid are shown. It is passed to the global path planning module in the third step.

[0047] In step three, the global path planning module generates the initial inspection path based on the electronic map constructed in step two and the inspection point sequence parsed in step one, using an improved A* algorithm. The traditional A* algorithm's evaluation function only considers path length; this method extends it to a multi-objective optimization form. The comprehensive cost function F(n) from the starting point S to the target point G, passing through the current node n, is defined as follows: (4) In the formula, G(n) represents the actual cumulative cost from the starting point to node n, and H(n) represents the heuristically estimated cost from node n to the target point. The actual cumulative cost G(n) includes not only the path length but also considers energy consumption and terrain factors. Its calculation method is as follows: (5) In the formula Let n be the parent node of node n. The movement cost is calculated in the second step. For mobile energy consumption, based on the kinematic model of the robot dog, it is estimated that walking 2 meters on flat ground consumes approximately 0.08 watt-hours. Energy consumption increases by 30% for every 10-degree increase in slope. Therefore... , Grid slope. Weighting coefficients. and The priority level U is dynamically determined based on the task urgency calculated in the first step. For high-priority tasks, let... =0.8 emphasizes path length. =0.2, for routine tasks, let... =0.5, =0.5 for balance consideration. The heuristic estimation cost H(n) uses a modified Euclidean distance that takes terrain into account: (6) In the formula , The coordinates of the current node. , The coordinates of the target point, This represents the average grid difficulty coefficient along the straight path from the current node to the target point, calculated by analyzing the grids traversed along the path. The mean is calculated. The A* algorithm maintains open and closed lists. Each time, it selects the node with the smallest F(n) value from the open list for expansion, adds its adjacent walkable nodes to the open list and updates the cost, and adds the expanded node to the closed list. When the target point is added to the closed list, a complete path from the starting point to the target point is generated by backtracking the parent node pointer. For tasks with multiple inspection points, the system plans the path segments between adjacent inspection points sequentially according to the access sequence determined in the first step, and finally connects them to form a complete global inspection path. The global path is stored as a sequence of path points. Each path point contains information such as coordinates, cumulative distance, and estimated energy consumption. This information is passed to the environmental perception module in the fourth step and the local path adjustment module in the fifth step as a navigation reference.

[0048] In step four, the robot dog is equipped with a LiDAR, a binocular vision camera, and an IMU (Inertial Measurement Unit) to continuously collect environmental information during its inspection mission. The LiDAR scans a 360-degree surrounding area at a frequency of 10 Hz, achieving a ranging accuracy of 2 cm and an effective detection range of 30 meters. The binocular camera acquires color images and depth information of a 180-degree field of view at a frequency of 30 frames per second, used to identify obstacle types and ground conditions. The IMU measures the robot dog's acceleration and angular velocity at a frequency of 100 Hz for attitude estimation and localization. The environmental perception module first filters the LiDAR data, removing noise points and measurements outside the effective range. The system establishes a local coordinate system at the robot dog's current position and transforms the LiDAR point cloud data into this coordinate system. For each LiDAR measurement point, its coordinates in the local coordinate system are... Using the ranging value r and the scanning angle The calculations were performed. The system performs cluster analysis on the point cloud to identify independent obstacle objects. A distance-based clustering method is used to cluster objects whose spatial distance is less than a threshold. Laser points are grouped into the same obstacle, with a threshold of 0.5 meters. For each obstacle cluster, the center coordinates of its envelope rectangle are calculated. and size Visual camera data is used to assist in determining the nature of obstacles. Image semantic segmentation identifies categories such as vegetation, rocks, equipment, and people, determining whether obstacles are traversable or require detour. The system registers the identified obstacle information with the electronic map from the second step, updating the dynamic obstacle layer. Position estimation is performed using IMU data and wheeled odometers, employing extended Kalman filtering to fuse multi-source positioning information to obtain the robot dog's precise current location. and heading angle The positioning accuracy is controlled within 0.2 meters. The environmental perception module outputs perception results every 0.1 seconds, including the current location, a list of surrounding obstacles, and local map update information. This data is transmitted in real time to the fifth step local path adjustment module and the sixth step path evaluation module.

[0049] In step five, the local path adjustment module receives the real-time perception data from step four and the global path information from step three, performing real-time trajectory optimization and obstacle avoidance during the robot dog's movement. A dynamic window method is used as the core algorithm for local planning. This method, under the constraints of the robot dog's current speed and acceleration, searches the executable speed space in the near future, evaluates the trajectory safety and target proximity corresponding to each candidate speed, and selects the optimal speed command to control the robot dog's movement. The robot dog's motion state is expressed using linear velocity v and angular velocity. The description, considering its dynamic constraints, defines the velocity search space as follows: (7) In the formula and Let's consider the robot dog's linear velocity and angular velocity at the current moment. and These are the maximum linear acceleration and angular acceleration, determined based on the robot dog's performance parameters. = 1.0m / s 2 , = 60° / s 2 , The control period is set to 0.1 seconds. The velocity space also needs to satisfy safety constraints, meaning the robot dog must stop before a collision at this speed. Discrete sampling is performed on the velocity space to generate a set of candidate velocities. For each candidate velocity... Predicting robot dogs in the future time domain The motion trajectory within the range is predicted over a time domain of 2 seconds. Assuming uniform motion at this speed, the trajectory consists of a series of predicted position points. The system checks whether the predicted trajectory collides with any obstacle; if any predicted point is less than the safe distance from an obstacle... If the candidate speed is not feasible, then the safety distance is set to the robot dog's radius plus a safety margin of 0.3 meters. For feasible candidate speeds, their evaluation functions are calculated. : (8) In the formula The heading evaluation function represents the consistency of the direction between the end point of the predicted trajectory and the next path point of the global path. It is calculated as the cosine of the heading angle difference, and the value ranges from -1 to 1. The larger the value, the closer the directions are. This is a distance evaluation function, representing the distance between the predicted trajectory and the nearest obstacle. It is normalized to between 0 and 1, with larger values ​​indicating greater safety as the distance increases. The velocity evaluation function is equal to the linear velocity v divided by the maximum velocity. This encourages the robot dog to travel at higher speeds to improve efficiency. Weighting coefficients. The values ​​are set to 0.5, 0.3, and 0.2 respectively, prioritizing proximity to the target direction while considering safety and speed. The system selects the candidate speed with the highest evaluation function value as the control command for the current moment and sends it to the execution control module to drive the robot dog's movement. The local path adjustment module executes the above process cyclically at a period of 0.1 seconds to achieve real-time response to the dynamic environment. The deviation between the adjusted actual walking trajectory and the global path is measured by lateral deviation. Quantization is defined as the vertical distance from the robot dog's current position to the nearest point on the global path. When Exceeding the threshold At this point, the sixth step of the path evaluation and replanning process is triggered.

[0050] In step six, the path evaluation module continuously monitors the robot dog's path execution. When it detects deviations or encounters insurmountable obstacles, it triggers a path replanning mechanism. Evaluation metrics include lateral deviation, path obstruction level, and task progress. Lateral deviation... This is calculated by the local adjustment module in step five. The degree of path obstruction is evaluated through the continuous attempt time; if the robot dog stays near a certain location for longer than the allotted time... If still unable to proceed, the path is considered blocked, threshold. Set to 30 seconds. Task progress is represented by the ratio of completed inspection points to the total number of inspection points. System-defined path execution status score. : (9) In the formula The maximum permissible lateral deviation threshold is set at 5 meters. The time spent near the current location. This represents the number of inspection points visited. This represents the total number of inspection points. (Score) The value ranges from 0 to 1, with higher values ​​indicating better path execution. When the distance is less than 0.3, the path execution is deemed abnormal and intervention is required. The system further determines whether to perform local path correction or global path replanning. The straight-line distance from the current location to the next inspection point is calculated. ,like Since lateral deviation is the primary issue, a local correction solution is chosen: only the path segment from the current location to the next inspection point is replanned, while the rest of the path remains unchanged. If a large-scale obstacle blockage is detected, a global replanning scheme is selected. Using the current location as the new starting point and unvisited inspection points as the new target sequence, the global path planning algorithm from step three is invoked to regenerate the complete path. During replanning, the system updates the dynamic obstacle layer of the electronic map with information on newly detected obstacles, ensuring the new path avoids known obstacles. After replanning is complete, the new path replaces the original path as the navigation reference, and the robot continues its inspection task. The path evaluation module also records the location, reason, and new path information for each replanning step for offline analysis and map updates. The path evaluation results and replanning decision information are fed back to the energy management module in step seven for adjusting the energy budget and charging strategy.

[0051] In step seven, the energy management module collects battery voltage, current, and remaining power information in real time through the robot dog's battery management system. The system calculates the current remaining power. Total capacity percentage : (10) In the formula The total battery capacity of the robot dog is typically 2000 watt-hours. The system establishes an energy consumption prediction model based on actual energy consumption statistics of the traveled routes. The average energy consumption per unit distance is defined. Energy already consumed Divide by the distance already traveled Considering the terrain features of the future path, the energy consumption requirement of the remaining path is estimated. The remaining path length from the current location to the destination is extracted from the global path in step three or the replanned path in step six. and the average difficulty level of the terrain along the way Predict the energy required to complete the remaining tasks. : (11) In the formula, the coefficient 0.5 represents the degree of impact of terrain difficulty on energy consumption; for every 0.1 increase in the difficulty coefficient, energy consumption increases by approximately 5%. To ensure safety, the system requires a 20% safety margin in remaining power, defining an energy sufficiency index. : (12) when When the remaining battery level is less than 1, the system determines that the remaining battery is insufficient to complete the task and returns, initiating a charging decision process. The system searches for the nearest charging station on the electronic map and calculates the path length to the charging station. and required energy Determine if the current battery level is sufficient to reach the charging station: (13) If the condition is met, the system calls the third-step planning algorithm to generate a path to the charging station and pauses the current inspection task. If the condition is not met, it means that the battery is insufficient to reach any charging station. The system issues an emergency alarm and finds a safe location nearby to shut down and await manual assistance. After successfully reaching the charging station, the robot dog automatically docks with the charging interface to charge. Once the SOC is greater than 80%, the system replans the path from the charging station back to the unfinished inspection point and continues to execute the task. The decision information from the energy management module is fed back to the sixth-step path evaluation module, which considers energy consumption constraints during path replanning and prioritizes the path with lower energy consumption. Energy consumption data is also passed to the eighth-step execution control module to optimize the robot dog's movement speed and gait, reducing energy consumption.

[0052] In step eight, the execution control module receives the speed command output from the local path adjustment module in step five, converts it into joint angle control commands for the robot dog's four legs, and drives the robot dog to move along the planned path. The robot dog adopts a quadrupedal walking structure, with each leg containing three degrees of freedom (hip, knee, and ankle joints), for a total of 12 degrees of freedom. The system calculates the landing positions of the four legs based on the target linear velocity v and angular velocity. A static stable gait is adopted to ensure that at least three legs are on the ground for support at any given time. For forward movement, the four legs lift and swing in the order of right front-left rear-left front-right rear, with each swing's stride length... With target velocity v and gait period Relatedly, the gait cycle adjusts according to the difficulty of the terrain, especially on flat ground. Seconds, complex terrain Seconds. For turning motion, the inner foot shortens its stride length, while the outer foot lengthens, achieving differential turning. Inverse kinematics calculations convert the foot placement coordinates into target angles for each joint. The system uses a PID controller to drive the joint motors to track the target angles, with proportional, integral, and derivative coefficients. During movement, the IMU provides real-time feedback on the robot's attitude angle. The system adjusts the support force distribution of each leg based on attitude deviations to maintain the robot's horizontal stability. When the detected tilt angle exceeds 15 degrees, the system reduces its speed and adjusts its gait to enhance stability. The execution control module, based on instructions from the seventh-step energy management module, adopts a fast gait to improve efficiency when battery power is sufficient and switches to an energy-saving gait to reduce energy consumption when battery power is low. The energy-saving gait is achieved by reducing stride length and leg lift height, reducing energy consumption by approximately 25%. During walking, foot contact force sensors detect ground support reaction forces. The system determines ground hardness based on the magnitude of the reaction force and automatically adjusts leg stiffness parameters to adapt to different terrains. The execution control module feeds back the robot's actual position, speed, and attitude information to the fourth-step environmental perception module for positioning updates and state estimation, forming a perception-planning-execution control flow.

[0053] Example 4 is an embodiment of the present invention, which provides a method for intelligent planning and dynamic adjustment of inspection paths for robot dogs. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments.

[0054] This simulation example is based on a 10 kV distribution line inspection scenario of a power supply company. The distribution line is approximately 15 kilometers long, including main lines and branch lines, with 35 key inspection points along the line, including equipment such as transformers, switchgear, and towers. The line traverses various terrains, including mountains, hills, and plains, with an elevation difference of up to 200 meters and slopes exceeding 25 degrees in some sections. A 3D terrain model of the distribution network was built on a joint simulation platform using MATLAB and the ROS robot operating system, including surface elevation, vegetation distribution, and obstacle location information. The terrain data is derived from a digital elevation model obtained from actual surveys, with a grid resolution of 2 meters. A virtual robot dog model was deployed in the simulation environment, with parameters referenced from the Boston Dynamics Spot robot dog: a body length of 1 meter, a width of 0.5 meters, a maximum walking speed of 1.6 meters per second, a battery capacity of 2000 watt-hours, and an energy consumption of approximately 60 watt-hours per kilometer for cruising on flat ground. Five typical inspection tasks were set up: Task 1 was a routine inspection of the entire line, including all 35 inspection points; Task 2 was a key equipment inspection, including 12 high-weight inspection points; Task 3 was a fault diagnosis, including 8 time-constrained emergency inspection points; Task 4 was performed in complex mountainous terrain; and Task 5 was performed in a dynamic obstacle environment. Each task was run using the patented method and three comparative methods: the traditional Dijkstra's shortest path method, the simple A* algorithm, and the artificial potential field method. Simulations recorded path length, inspection time, energy consumption, number of path adjustments, and task completion rate. To verify the adaptability to dynamic environments, 10 temporary obstacles were randomly added during task execution, and the response performance of each method was observed. The simulation was run 20 times, and the average value was taken to eliminate the influence of random factors. The results are shown in Tables 1, 2, and 3.

[0055] Table 1 Comparison of path planning performance under different task scenarios

[0056] Table 2 Comparison of Dynamic Environmental Adaptability

[0057] Table 3 Energy Management and Battery Life Performance

[0058] As shown in Table 1, this patented method, with a path length similar to Dijkstra's shortest path method, reduces inspection time by an average of approximately 12% and energy consumption by approximately 15%. Compared to the artificial potential field method, the task completion rate increases from an average of 86.9% to 100%, and the overall score improves by approximately 30 points. In emergency task scenarios, this method dynamically adjusts weight coefficients, prioritizing time optimization, achieving a minimum inspection time of 36 minutes, demonstrating even greater advantages.

[0059] Table 2 illustrates the performance of each method in dynamic environments. Traditional Dijkstra's and A* algorithms, being static planning algorithms, lack the ability to cope with newly added obstacles, leading to multiple task interruptions. While the artificial potential field method can avoid obstacles in real time, frequent local adjustments result in tortuous paths and significantly increased distances. This patented method, through a hierarchical planning architecture, combining the navigation function of the global path with the flexibility of local adjustments, successfully avoids all 10 obstacles, with a path adjustment response time of only 3.2 seconds and minimal increase in extra path distance, fully demonstrating its advantages in adapting to dynamic environments.

[0060] Table 3 compares the energy consumption performance of different methods. This patented method, by considering the energy consumption model of terrain difficulty and real-time energy consumption monitoring, selects the most energy-efficient path, achieving an energy efficiency index of 17.8 m / Wh, approximately 20% higher than traditional methods. In ultra-long-distance inspection scenarios, other methods require 2 to 3 charging cycles, while this method, through optimized path and motion control, only requires 1 charging cycle, reducing the total time by approximately 20 minutes, fully demonstrating the effectiveness of energy consumption management.

[0061] Example 5 is an embodiment of the present invention. This embodiment provides a robot dog inspection path intelligent planning and dynamic adjustment system, including: a task parsing module, used to receive inspection task instructions, calculate the task urgency index according to the weight coefficient of the inspection point and the time constraint, and determine the task priority; The electronic map construction module is used to divide the inspection area into a grid, calculate the terrain difficulty coefficient and movement cost of the grid, construct a layered electronic map, and maintain a dynamic obstacle layer. The global path planning module is used to perform global path planning using the A* algorithm through a comprehensive cost function, and generate global inspection paths. The real-time environmental perception module is used to collect environmental information in real time through lidar, vision sensors and inertial measurement unit, identify obstacles and update the dynamic obstacle layer, and obtain the current position of the robot dog; The local path adjustment module is used to perform real-time trajectory optimization under speed constraints using local path planning, select the optimal speed command, and calculate the lateral deviation. The path evaluation and replanning module is used to calculate the path execution status score. When the path execution status score is lower than the threshold, it selects local path correction or global path replanning based on the distance from the current position to the next inspection point. The energy management module is used to monitor the remaining power in real time, predict the energy required to complete the remaining tasks, and plan the route to the charging station when the predicted energy exceeds the safety threshold of the remaining power. The execution control module is used to convert speed commands into joint angle control commands for the robot dog's four legs, driving the robot dog to move along the planned path.

[0062] This embodiment also provides an electronic device applicable to a method for intelligent planning and dynamic adjustment of inspection paths for a robot dog, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the method for intelligent planning and dynamic adjustment of inspection paths for a robot dog as proposed in the above embodiment.

[0063] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements a method for intelligent planning and dynamic adjustment of inspection paths for a robot dog as proposed in the above embodiment.

[0064] The storage medium proposed in this embodiment and the method for intelligent planning and dynamic adjustment of inspection path for a robot dog proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0065] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0066] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for intelligent planning and dynamic adjustment of inspection paths for robotic dogs, characterized in that: The method comprises the following steps: acquiring a patrol task instruction, the patrol task instruction comprising a patrol point coordinate list, a weight coefficient and a time constraint, calculating a task urgency index according to the weight coefficient and the time constraint of the patrol point, and determining a task priority according to the task urgency index; dividing the patrol area into a grid network, assigning terrain attribute information to each grid, calculating a terrain difficulty coefficient of each grid and a moving cost between adjacent grids, constructing a hierarchical electronic map and maintaining a dynamic obstacle layer; based on the hierarchical electronic map and the patrol point coordinate list, using an A-star algorithm to perform global path planning through a comprehensive cost function, and generating a global patrol path from the starting point to the end point through each patrol point; in the process of executing the patrol task by the robot dog, collecting environmental information in real time through a laser radar, a visual sensor and an inertial measurement unit, identifying obstacles and updating the dynamic obstacle layer, and acquiring the current position of the robot dog; based on the real-time environmental information and the global patrol path, performing real-time trajectory optimization under the speed constraint in a local path planning mode, selecting an optimal speed instruction, and calculating a lateral deviation of the current position of the robot dog from the global path; calculating a path execution state score according to the lateral deviation, the path obstruction degree and the task progress, and when the path execution state score is lower than a threshold value, selecting local path correction or global path re-planning according to the distance from the current position to the next patrol point; monitoring the remaining power of the robot dog in real time, predicting the energy required to complete the remaining task according to the energy consumption statistics of the traveled path, and planning a path to a charging station when the predicted energy exceeds a safety threshold of the remaining power; converting the speed instruction into joint angle control instructions of the four legs of the robot dog, and driving the robot dog to move along the planned path.

2. The method of claim 1, wherein the method further comprises: determining a patrol path of the robot dog based on the environment information; and adjusting the patrol path of the robot dog based on the environment information. The task urgency index comprises: acquiring a weight coefficient of each patrol point, the weight coefficient reflecting the importance of the patrol point; acquiring the latest arrival time and the task receiving time of each patrol point, and determining a time constraint condition; comprehensively determining the task urgency index according to the weight coefficient and the time constraint condition of each patrol point, the weight coefficient and the time constraint jointly affecting the task urgency, and the more stringent the time constraint of the patrol point, the greater the contribution of the patrol point to the task urgency; determining the task priority according to the task urgency index, a high-priority task being determined when the task urgency is high, and a regular task being determined when the task urgency is low.

3. The method of claim 2, wherein the method further comprises: determining a patrol path of the robot dog based on the environment information; and adjusting the patrol path of the robot dog based on the environment information. The calculation of the terrain difficulty coefficient of each grid and the moving cost between adjacent grids comprises: acquiring slope information of the grid, the slope information indicating the inclination of the terrain; acquiring a ground surface type of the grid, the ground surface type comprising a hardened road surface, grassland, gravel ground and muddy ground; acquiring an obstacle identifier of the grid, the obstacle identifier indicating the crossability of the obstacle; determining the terrain difficulty coefficient of the grid according to the slope information, the ground surface type and the obstacle identifier; determining the moving cost between adjacent grids according to the terrain difficulty coefficients of the adjacent grids and the distance between the grids.

4. The method of claim 3, wherein the method further comprises: determining a patrol path of the robot dog based on the environment information; and adjusting the patrol path of the robot dog based on the environment information. The comprehensive cost function comprises: calculating an actual cumulative cost from the starting point to the current node, the actual cumulative cost comprehensively considering the path length and the energy consumption; According to the task priority, the weight of the path length and the energy consumption in the actual cumulative cost is determined, the high-priority task focuses on the path length, and the conventional task balances the path length and the energy consumption; A heuristic estimated cost from the current node to the target point is calculated, which considers the distance between nodes and the terrain difficulty along the way; The actual cumulative cost and the heuristic estimated cost are combined to obtain the comprehensive cost of the current node; The node with the minimum comprehensive cost is selected for expansion until the target point is reached.

5. The method of claim 4, wherein the method further comprises: determining a patrol path of the robot dog based on the environment information; and adjusting the patrol path of the robot dog based on the environment information. The local path planning mode includes: According to the current speed and acceleration constraints of the robot dog, a executable speed search space is determined; A plurality of candidate speeds are generated in the speed search space; For each candidate speed, the motion trajectory of the robot dog in the future prediction time domain is predicted; The collision between the predicted trajectory and the obstacle is checked, and the candidate speed with collision risk is excluded; For the feasible candidate speed, its approach to the target ability, safe distance from the obstacle and speed efficiency are evaluated; The candidate speed with the optimal comprehensive evaluation is selected as the control instruction.

6. The method of claim 5, wherein the method further comprises: determining a patrol path of the robot dog based on the environment information; and adjusting the patrol path of the robot dog based on the environment information. The path execution state scoring includes: According to the lateral deviation of the current position of the robot dog to the global path, the path deviation degree is evaluated; According to the stay time of the robot dog at the current position, the path blocking degree is evaluated; According to the number of visited inspection points and the total number of inspection points, the task completion progress is evaluated; The lateral deviation, path blocking degree and task progress are comprehensively considered to determine the path execution state score; When the path execution state score is lower than the threshold value, the distance from the current position to the next inspection point is determined; If the distance is short and the deviation is the main problem, local path correction is performed, and the path segment from the current position to the next inspection point is re-planned; If the distance is far or a large range of obstacles is detected, the newly detected obstacle information is updated to the dynamic obstacle layer, and the global path is re-planned, and the complete path is generated from the current position as a new starting point.

7. The method of claim 6, wherein the method further comprises: determining a patrol path of the robot dog based on the environment information; and adjusting the patrol path of the robot dog based on the environment information. The predicted energy required to complete the remaining task includes: According to the consumed energy and the traveled distance, the energy consumption per unit distance is calculated; According to the length of the remaining path and the terrain difficulty along the way, the energy required to complete the remaining task is predicted; The remaining battery capacity is compared with the predicted energy required to determine whether the energy is sufficient; When the remaining battery capacity is insufficient to complete the remaining task and reserve a safety margin, the location of the nearest charging station is found; It is determined whether the remaining battery capacity is sufficient to reach the charging station; If the remaining battery capacity is sufficient, a path to the charging station is planned and the current inspection task is suspended, and then the execution continues after charging; If the remaining battery capacity is insufficient to reach the charging station, an alarm is issued and the robot dog is parked at a safe location.

8. A system for intelligent planning and dynamic adjustment of a patrol path of a robot dog, applying the method for intelligent planning and dynamic adjustment of a patrol path of a robot dog according to any one of claims 1-7. It includes: A task analysis module is used to receive an inspection task instruction, calculate a task urgency index according to a weight coefficient and a time constraint of an inspection point, and determine a task priority; An electronic map construction module is used to divide an inspection area into a grid, calculate a terrain difficulty coefficient and a moving cost of the grid, construct a hierarchical electronic map and maintain a dynamic obstacle layer; A global path planning module is used to generate a global inspection path by using an A-star algorithm through a comprehensive cost function. An environmental real-time sensing module is configured to collect environmental information in real time through a laser radar, a visual sensor and an inertial measurement unit, identify obstacles and update a dynamic obstacle layer, and obtain a current position of the robot dog; A local path adjustment module is configured to perform real-time trajectory optimization under speed constraints in a local path planning manner, select an optimal speed instruction, and calculate a lateral deviation; A path evaluation and re-planning module is configured to calculate a path execution state score, and select local path correction or global path re-planning according to a distance from a current position to a next inspection point when the path execution state score is lower than a threshold value; An energy consumption management module is configured to monitor a remaining power in real time, predict energy required to complete a remaining task, and plan a path to a charging station when predicted energy exceeds a safety threshold of the remaining power; An execution control module is configured to convert the speed instruction into a joint angle control instruction of a quadruped of the robot dog, and drive the robot dog to move along the planned path. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8. The processor executes the computer program to implement the steps of the method for intelligent planning and dynamic adjustment of a robot dog inspection path according to any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method for intelligent planning and dynamic adjustment of a robot dog inspection path according to any one of claims 1 to 7.

Images