A method and system for planning a patrol path for a robot dog

Abstract

The application belongs to the technical field of path planning, and particularly relates to a robot dog inspection path planning method and system, which solves the problem of poor work endurance and inspection efficiency of the robot dog in the path planning method of the prior art, and comprises the following steps: S1, obtaining a three-dimensional grid map of an inspection environment; S2, generating an inspection point sequence; S3, sequentially planning each path segment based on an improved A* algorithm, wherein the actual cost from the starting point of the path segment to the current node is a scalar value which comprehensively considers the path length, the passing energy consumption and the information value of the device inspection; the heuristic cost from the current node to the end point of the path segment is proportional to the Euclidean distance from the current node to the end point of the path segment; S4, when an obstacle is detected, performing dynamic re-planning of a local path; and S5, setting optimal walking gait parameters for each terrain path segment. The path length, the passing energy consumption and the information value of the inspection point are comprehensively considered, and the work endurance and the inspection efficiency of the robot dog are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of path planning, and in particular relates to a method and system for planning a robot dog inspection path. Background Art

[0002] Due to their high maneuverability and environmental adaptability, quadruped robots such as robot dogs are used in automated inspection tasks in complex scenarios such as power, chemical, and mining. In the field of robot dog path planning, traditional global path planning algorithms such as the A* algorithm and the Dijkstra algorithm can plan a collision-free path from the starting point to the end point for the robot dog in a known static environment. However, the cost functions of these classic algorithms usually only consider the path length or time, ignoring the robot dog's own motion characteristics and the impact of complex terrain on energy consumption. For robot dogs that rely on battery life, the shortest path is often not the most energy-efficient or safest path. For example, a short path with a steep slope and rough ground will rapidly consume the robot dog's battery and even risk tipping over. This greatly limits its ability to operate for long periods of time and with high efficiency in real-world environments.

[0003] Furthermore, existing path planning methods mostly perform one-time global planning based on static environmental maps. However, in actual inspection scenarios, dynamic obstacles (such as pedestrians and temporary equipment) or unknown obstacles that are not marked on the map often appear. Traditional obstacle avoidance methods are often slow to respond, or the locally replanned paths are disconnected from the overall objectives. They also lack comprehensive consideration of factors such as energy consumption and safety, resulting in rigid and inefficient obstacle avoidance behavior. Furthermore, existing technologies often separate path planning from motion control. After the path planner generates a geometric path, the motion controller executes it using a fixed or simply adaptive gait, failing to proactively match the optimal movement gait to the terrain characteristics of different sections along the path. This "plan first, execute later" model fails to achieve coordinated optimization of the path and gait, resulting in poor stability and high energy consumption when the robot dog travels on complex roads. This failure to fully utilize the advantages of multi-gait motion affects the endurance and efficiency of the overall inspection mission. Summary of the Invention

[0004] The present invention provides a robot dog inspection path planning method and system to solve the technical problems of poor operation endurance and inspection efficiency of robot dogs in the path planning methods of the prior art.

[0005] To solve the above problems, the present invention provides a robot dog inspection path planning method using the following technical solutions: A robot dog inspection path planning method comprising the following steps:

[0006] S1, obtaining a three-dimensional grid map of the inspection environment, which is marked with multiple inspection points and static obstacles;

[0007] S2, uses a heuristic algorithm to determine the inspection order covering all inspection points and generates an inspection point sequence;

[0008] S3, taking two adjacent inspection points in the inspection point sequence as the starting point and end point of the path segment, plan each path segment in turn based on the improved A* algorithm, and splice them into a complete global inspection path; among them, from the starting point of the path segment to the current node The actual cost It is a combination of path length, travel energy consumption calculated based on terrain slope and roughness, and node The scalar value of the associated device inspection information value; from the current node Heuristic cost to the end of the path segment With the current node Proportional to the Euclidean distance to the end point of the path segment;

[0009] S4, while traveling along the global inspection path, when a dynamic or unknown obstacle that exceeds a preset threshold and is not recorded in the 3D grid map is detected, the current position is set as a temporary starting point, and the unobstructed node ahead on the global path is used as the temporary end point, and the improved A* algorithm is called to dynamically replan the local path;

[0010] S5, dividing the path into multiple terrain sections according to the average terrain slope and roughness of consecutive nodes on the path, and matching and setting the optimal walking gait parameters for each terrain section from a pre-established gait-energy consumption mapping model.

[0011] Compared with the existing technology, the present invention comprehensively considers the path length, the energy consumption determined by the terrain slope and roughness, and the information value of the inspection points when performing path planning, so as to plan a global path that is more balanced in terms of energy consumption, safety and mission goal achievement, thereby improving the robot dog's operating endurance and inspection efficiency. When encountering obstacles not recorded in the map, its local replanning process also follows the above-mentioned comprehensive optimization principles to ensure the efficiency and safety of the detour path. The present invention deeply couples the path planning results with the underlying gait control, and realizes the coordination of high-level planning and underlying execution by pre-setting the optimal walking gait that matches different terrain sections, reducing the actual energy consumption of the robot dog when traveling on complex roads, enhancing its traffic stability, and improving the overall performance of the inspection task.

[0012] As a further improvement, the heuristic algorithm is an ant colony optimization algorithm, which searches and determines an inspection sequence with the lowest comprehensive inspection cost in multiple iterations through a pheromone update mechanism and heuristic information; wherein the pheromone represents the empirical value of path selection, and the heuristic information comprehensively considers the estimated path cost between each inspection point.

[0013] Global optimization is achieved through the pheromone update mechanism, and the search efficiency can be accelerated through heuristic information. The inspection sequence with the lowest overall cost can be efficiently found in the inspection path planning.

[0014] As a further improvement, in the improved A* algorithm, from the starting point of the path segment to the current node The actual cost The calculation method is: ;

[0015] in, From the starting point of the path segment along the 3D grid map to the current node The cumulative path length The normalized value of is a node Energy consumption cost of transportation based on terrain The normalized value of node The slope value of the terrain is proportional to the roughness value; Information value cost The normalized value of node The importance of the associated equipment is inversely proportional to its historical failure rate; 、 、 is the preset non-negative weight coefficient.

[0016] As a further improvement, The calculation formula is: ;

[0017] in, For nodes The slope value of the terrain; For nodes The roughness value of the terrain.

[0018] As a further improvement, The calculation formula is: ;

[0019] in, For nodes The importance of the associated equipment, For nodes Historical failure rates of associated equipment, is the stability constant.

[0020] For regional nodes that are closer to key inspection equipment and have higher failure rates, their information value cost is lower. A negative cost value is given to them as a reward to guide the path through.

[0021] As a further improvement, the heuristic cost For the current node The normalized Euclidean distance to the end point of this path segment.

[0022] As a further improvement, when traveling along the global inspection path, if the sensor detects an obstacle that is not recorded in the three-dimensional grid map and the number of voxels or volume occupied by the obstacle in the local map exceeds a preset threshold, dynamic replanning is triggered; during replanning, the current position is used as the temporary starting point, and a reachable node after the obstacle on the global path is used as the temporary end point.

[0023] On the basis of the static global optimal path, the local correction capability for the dynamic unknown environment is added, which not only retains the advantage of global optimization, but also makes up for the defect of insufficient adaptability to unknown obstacles in the existing technology.

[0024] As a further improvement, in S5, the gait-energy consumption mapping model pre-stores multiple gaits and their energy consumption and stability data under different terrain conditions; the optimal walking gait parameters are matched from the gait-energy consumption mapping model based on the characteristics of the terrain section, with the lowest energy consumption or the highest stability as the optimization goal.

[0025] As a further improvement, in S5 , the division of the terrain sections is based on whether the average terrain slope and roughness values ​​of a plurality of consecutive nodes on the path fall within a preset range.

[0026] The preset threshold is used to filter out insignificant small obstacles or sensor noise to avoid unnecessary frequent replanning.

[0027] The present invention also provides a robot dog inspection path planning system, including a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, any one of the above-described robot dog inspection path planning methods is implemented.

[0028] The beneficial effect is: compared with the existing technology, the present invention comprehensively considers the path length, the energy consumption determined by the terrain slope and roughness, and the information value of the inspection points when performing path planning, so as to plan a global path that is more balanced in terms of energy consumption, safety and mission goal achievement, thereby improving the robot dog's operating endurance and inspection efficiency. When encountering obstacles not recorded in the map, its local replanning process also follows the above-mentioned comprehensive optimization principle to ensure the efficiency and safety of the detour path. The present invention deeply couples the path planning results with the underlying gait control, and realizes the coordination of high-level planning and underlying execution by pre-setting the optimal walking gait that matches different terrain sections, reducing the actual energy consumption of the robot dog when traveling on complex roads, enhancing its traffic stability, and improving the overall performance of the inspection task. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 A flowchart of the inspection path planning method for a robot dog;

[0030] Figure 2 This is the structural block diagram of the robot dog inspection path planning system. DETAILED DESCRIPTION

[0031] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Those skilled in the art should know that the embodiments described below are part of the embodiments of the present disclosure, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of the present invention.

[0032] Embodiments of the robot dog inspection path planning method provided by the present invention:

[0033] like Figure 1 As shown, the robot dog inspection path planning method includes the following steps:

[0034] S1, obtaining a three-dimensional grid map of the inspection environment, which is marked with multiple inspection points and static obstacles.

[0035] The inspection environment is scanned using lidar or depth cameras, and a mapping algorithm such as OctoMap is used to generate a voxelized 3D grid map of the environment. Once the voxelized 3D grid map is constructed, the operator manually clicks on the host computer software interface to mark the locations of key equipment as inspection points and demarcate areas such as walls and large fixed equipment as permanent static obstacles.

[0036] S2, uses a heuristic algorithm to determine the inspection order covering all inspection points and generates an inspection point sequence.

[0037] In this embodiment, the heuristic algorithm for sequence planning of each inspection point is the ant colony optimization algorithm, which searches and determines an inspection sequence with the lowest comprehensive inspection cost in multiple iterations through a pheromone update mechanism and heuristic information; among them, pheromones represent the empirical value of path selection, and heuristic information comprehensively considers the estimated path cost between inspection points.

[0038] Specifically, the checkpoint sequence planning problem can be compared to the shortest path problem for ants to find food. Each checkpoint is considered a city. Virtual ants start from the starting point and probabilistically select the next checkpoint to visit based on the pheromone concentration and heuristic information along the path between each checkpoint, until all points have been traversed. Heuristic information is typically the inverse of the estimated path cost between two points. For example, if the estimated path cost from point A to point B is 100, the heuristic information value is 0.01. The lower the cost, the higher the heuristic information.

[0039] At the beginning of the ACO iteration, the pheromone concentration on all paths is set to the same initial value, such as 1.0. After an iteration, when all ants have completed a full inspection, the system updates the pheromone concentration based on the total cost of each path traversed by the ant. Paths with lower costs receive more pheromone increments, while all paths evaporate at a certain rate. For example, after 100 iterations, the pheromone concentration on the path from point A to point C may accumulate to 8.5 due to repeated appearances in low-cost inspection sequences, while the pheromone concentration on the path to point D may decay to 0.3 due to its high cost. This leads subsequent ants to prefer the path from A to C, ultimately converging to a globally optimal or suboptimal inspection sequence.

[0040] In other embodiments, a genetic algorithm can be used to sequence the checkpoints. A full permutation of the checkpoints is considered a chromosome, and multiple random permutations are used to form an initial population. The estimated total energy consumption or total length of the global path under this sequence is used as the fitness function. Multiple generations of iterative optimization are performed using selection, crossover, and mutation operators to ultimately determine a checkpoint sequence with the lowest energy consumption or shortest path.

[0041] S3, taking two adjacent inspection points in the inspection point sequence as the starting point and end point of the path segment, plan each path segment in turn based on the improved A* algorithm, and splice them into a complete global inspection path; among them, from the starting point of the path segment to the current node The actual cost It is a combination of path length, travel energy consumption calculated based on terrain slope and roughness, and node The scalar value of the associated device inspection information value; from the current node Heuristic cost to the end of the path segment With the current node Proportional to the Euclidean distance to the end point of the path segment.

[0042] Specifically, the improved A* algorithm evaluates the actual cost of the node When , the first item is from the starting point to the node The normalized value of the cumulative path length represents the path length; the second item is the normalized value of the traffic energy consumption, which is based on the node The terrain slope is calculated based on the height difference between the node and the previous node. The terrain roughness is calculated by the degree of discreteness of the point cloud within the grid, and then the corresponding energy consumption value is calculated through a pre-calibrated energy consumption model function. The third item is the information value cost. The closer the regional node is to the key inspection equipment, the lower its information value cost is. A negative cost value is assigned to it as a reward to guide the path through. For key inspection equipment with a higher historical failure rate, the lower its information value cost is. A negative cost value is assigned to it as a reward to guide the path through. Heuristic cost The current node The normalized Euclidean distance to the end point of this path segment.

[0043] In the improved A* algorithm, from the starting point of the path segment to the current node The actual cost The calculation formula is: ;

[0044] in, From the starting point of the path segment along the grid map to the current node The cumulative path length The normalized value of .

[0045] is a node Energy consumption cost of transportation based on terrain The normalized value of node The slope of the terrain is directly proportional to the roughness. The more complex the terrain, the higher the cost. The roughness reflects the unevenness of the terrain surface and is positively correlated with the coefficient of friction. The rougher the surface (such as gravel), the greater and more unstable the friction coefficient, resulting in greater resistance and energy consumption for the robot dog. The smoother the surface (such as concrete), the smaller and more stable the friction coefficient, resulting in less resistance and energy consumption for the robot dog. By node The slope value and node of the terrain The roughness value of the terrain is multiplied by the roughness value of the terrain, that is: ;

[0046] in, For nodes The slope value of the terrain, For nodes The roughness value of the terrain.

[0047] Information value cost The normalized value of information value cost Based on the node The importance of the associated equipment, the historical failure rate calculation, and the node The importance of the associated equipment is inversely proportional to the historical failure rate, and the calculation formula is: ;

[0048] in, For nodes The importance of the associated equipment, For nodes Historical failure rates of associated equipment, is the stability constant.

[0049] 、 、 is the preset non-negative weight coefficient.

[0050] The total cost from the start point to the end point of the path segment for: .

[0051] Actual cost The goal is to achieve comprehensive optimization of path planning rather than just pursuing the shortest distance. 、 、 The setting reflects the strategy preference under different tasks. For example, when the battery is sufficient and the task time is loose, you can set is 0.6, is 0.2, is 0.2, with the path length as the main consideration. When the power is tight, it may be The weight of is increased to 0.5, giving priority to the gentle path with lower energy consumption, even if it slightly increases the total distance.

[0052] Suppose the robot dog faces two choices when planning a path, node A and node B. The path length to node A is Energy consumption cost of 20 meters and flat terrain 10, there is no important equipment information value nearby is 50. The path length to node B 22 meters, but requires climbing energy consumption The value of the information is 40 and is close to a key valve that requires urgent inspection. is 5. If the weight is set to is equal to 0.4, is equal to 0.4, is equal to 0.2, then The value is less than Under this weight, the algorithm will give priority to path A. However, if the weight of information value is Increase to 0.5, then The calculated value of will be lower, and the algorithm will choose path B to prioritize high-priority tasks.

[0053] In this embodiment, an exponential function may be used for normalization.

[0054] S4, when traveling along the global inspection path, when a dynamic or unknown obstacle that is not recorded in the three-dimensional grid map and exceeds the preset threshold is detected, the current position is set as the temporary starting point, and the unblocked node ahead on the global path is used as the temporary end point, and the improved A* algorithm is called to dynamically replan the local path.

[0055] While the robot dog is moving, the laser radar it carries continuously scans at a high frequency, comparing the real-time point cloud data with the original three-dimensional grid map. If an area is found to be occupied by point clouds for several consecutive frames, and the area is marked as idle in the map, it is identified as a dynamic or unknown obstacle. When the total number of voxels occupied by the obstacle exceeds a preset threshold, for example, 50, re-planning is triggered. The system sets the current precise coordinates of the robot dog as a temporary starting point, and searches forward along the original global path to find the first node with no new obstacles between it and the current position as a temporary end point, and calls the above-mentioned improved A* algorithm again to plan a detour path. In other embodiments, the preset threshold can be the volume of the obstacle.

[0056] The preset threshold is used to filter out insignificant small obstacles or sensor noise to avoid unnecessary frequent replanning. For example, the preset threshold can be set to a volume of 0.5m 3 , or the number of voxels reaches 500. Only when the amount of detected obstacles exceeds this preset threshold is it considered to pose a substantial obstruction to the path.

[0057] Suppose a robot dog is on patrol when a temporarily stacked toolbox suddenly appears on the path ahead. This toolbox occupies 620 voxels in the local map, exceeding the preset threshold. The current movement command is immediately suspended, and dynamic replanning is initiated. Using the robot dog's current coordinates as the starting point for replanning, the robot then searches forward along the original global path, skipping the area occupied by the toolbox and finding a node 10 meters behind the toolbox as the temporary endpoint. The improved A* algorithm calculates a new path from the current location, bypassing the toolbox to reach the temporary endpoint. The robot dog then follows this new path, bypassing the toolbox and rejoining the original global inspection path.

[0058] S5, dividing the path into multiple terrain sections according to the average terrain slope and roughness of consecutive nodes on the path, and matching and setting the optimal walking gait parameters for each terrain section from a pre-established gait-energy consumption mapping model.

[0059] Terrain segments are determined based on whether the average terrain slope and roughness values ​​at multiple consecutive nodes along the path fall within pre-set ranges. To avoid frequent gait changes due to subtle changes in the terrain, which can lead to inconsistent movements and excessive energy consumption, gait decisions are not made at every node. Instead, the robot analyzes the path ahead over a certain distance, for example, the 25 nodes along the next 5 meters. The average terrain slope and roughness values ​​for these 25 nodes are calculated. This smoothing process reflects the overall characteristics of the path segment, rather than local, instantaneous changes.

[0060] To do this, we can traverse all nodes on the path and use a sliding window, for example, a window of ten nodes, to calculate the average slope and average roughness for each node within the window. As the window slides across the path, if the calculated averages undergo a significant step change, such as a slope change of more than 8°, a new terrain segment is identified at that location.

[0061] The preset interval refers to the preset terrain classification standard. For example, a flat road section is defined as a road section with an average slope less than 5 degrees and an average roughness value less than 0.2; while a rugged road section is defined as a road section with an average slope greater than 20 degrees or an average roughness value greater than 0.6.

[0062] For each terrain section, the system uses its average slope and roughness as an index to query a pre-established gait-energy consumption mapping table. This table records energy consumption data for different gaits, such as trotting, stepping, and slow walking, under different terrain conditions. The system then selects the gait with the lowest energy consumption or the highest stability and sends the corresponding parameters, such as step height and frequency, to the underlying motion controller for execution.

[0063] The gait-energy consumption mapping model pre-stores multiple gaits and their energy consumption and stability data under different terrain conditions; the optimal walking gait parameters are matched from the model based on the characteristics of the terrain section, with the lowest energy consumption or the highest stability as the optimization goal.

[0064] For a quadruped robot dog, the model might include multiple gaits, such as a high-speed trot, a standard walk, and a low-profile crawl. Each gait corresponds to a set of performance metrics for a specific terrain. Terrain conditions can be quantified using parameters such as slope, roughness, and friction coefficient. For example, the model might record that on a flat surface with a slope of 5° and a roughness of 0.1, a high-speed trot consumes 120 joules per meter and a stability score of 0.7; while a low-profile crawl consumes 300 joules per meter and a stability score of 0.95.

[0065] When the robot dog is about to enter a new terrain section, it senses or obtains the terrain characteristic parameters of that section from a map. Assuming the current mission objective prioritizes energy conservation, the quadruped robot dog detects a long, flat corridor ahead whose terrain parameters match the applicable conditions for an energy-efficient gait in the model. It queries the model to find the gait with the lowest energy consumption, a high-frequency trot, and adjusts its motion controller to adopt this gait. Conversely, if the road ahead is a steep slope or a gravelly area, even if energy conservation is the primary goal, the optimization objective will be switched to maximum stability based on a preset safety threshold. A low-profile crawling gait will be selected from the model to ensure safe passage.

[0066] For example, as the robot dog moves, it continuously calculates the average terrain parameters of the road ahead. If it calculates that the average slope of the road ahead is 3 degrees and the average roughness is 0.15, it will classify it as a flat section and match it with the most energy-efficient fast walking gait in the database. If the system detects that the average slope of the road ahead reaches 25 degrees, even if the roughness is not high, it will be classified as a rough section. The robot dog will slow down in advance and switch to the slower climbing gait with more support.

[0067] Embodiments of the robot dog inspection path planning system provided by the present invention:

[0068] like Figure 2 As shown, the robot dog inspection path planning system includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned robot dog inspection path planning method is implemented.

[0069] The robot dog inspection path planning system also includes other components familiar to those skilled in the art, such as a communication interface. Their settings and functions are known in the art and will not be described in detail here.

[0070] In the present invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, the computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store the required information and can be accessed by an application, module, or both. Any such computer storage medium can be part of, accessible to, or connectable to a device. Any application or module described in the present invention can be implemented using computer-readable / executable instructions stored or otherwise maintained by such a computer-readable medium.

[0071] In addition, in the description of this specification, “a plurality of” means at least two, for example, two, three or more, etc., unless otherwise clearly and specifically defined.

Claims

1. A robot dog inspection path planning method, characterized in that: The following steps are involved: S1, obtain a three-dimensional grid map of the inspection environment, which is marked with multiple inspection points and static obstacles; S2, use a heuristic algorithm to determine the inspection order covering all inspection points and generate an inspection point sequence; S3, use two adjacent inspection points in the inspection point sequence as the starting and ending points of the path segment, and plan each path segment in turn based on the improved A* algorithm to splice it into a complete global inspection path; among them, from the starting point of the path segment to the current node The actual cost It is a combination of path length, travel energy consumption calculated based on terrain slope and roughness, and node The scalar value of the associated device inspection information value; from the current node Heuristic cost to the end of the path segment With the current node The Euclidean distance to the end point of the path segment is proportional; S4, when traveling along the global inspection path, when a dynamic or unknown obstacle that is not recorded in the three-dimensional grid map and exceeds a preset threshold is detected, the current position is set as a temporary starting point, and the unobstructed node ahead on the global path is used as the temporary end point. The improved A* algorithm is called to dynamically replan the local path; S5, the path is divided into multiple terrain segments based on the average terrain slope and roughness of consecutive nodes, and the optimal walking gait parameters are matched and set for each terrain segment from the pre-established gait-energy consumption mapping model; In the improved A* algorithm, from the starting point of the path segment to the current node The actual cost The calculation method is: ; From the starting point of the path segment along the 3D grid map to the current node The cumulative path length The normalized value of is a node Energy consumption cost of transportation based on terrain The normalized value of node The slope value of the terrain is proportional to the roughness value; Information value cost The normalized value of node The importance of the associated equipment is inversely proportional to its historical failure rate; 、 、 is a preset non-negative weight coefficient; 、 、 The settings reflect the strategic preferences under different tasks. When the battery is sufficient and the task time is loose, is 0.6, is 0.2, is 0.2, taking the path length into consideration; when the power is tight, The weight of is increased to 0.5, and the smooth path with lower energy consumption is selected; The calculation formula is: ; For nodes The importance of the associated equipment, For nodes Historical failure rates of associated equipment, is the stability constant; The calculation formula is: ; in, For nodes The slope value of the terrain; For nodes The roughness value of the terrain.

2. The robot dog inspection path planning method according to claim 1, characterized in that: The heuristic algorithm is an ant colony optimization algorithm, which searches and determines an inspection sequence with the lowest comprehensive inspection cost in multiple iterations through a pheromone update mechanism and heuristic information; wherein the pheromone represents the empirical value of path selection, and the heuristic information comprehensively considers the estimated path cost between each inspection point.

3. The robot dog inspection path planning method according to claim 1, characterized in that: Heuristic Cost For the current node The normalized Euclidean distance to the end point of this path segment.

4. The robot dog inspection path planning method according to any one of claims 1 to 3, characterized in that: When traveling along the global inspection path, if the sensor detects an obstacle that is not recorded in the 3D grid map and the number of voxels or volume occupied by the obstacle in the local map exceeds a preset threshold, dynamic replanning is triggered; during replanning, the current position is used as the temporary starting point, and a reachable node after the obstacle on the global path is used as the temporary end point.

5. The robot dog inspection path planning method according to any one of claims 1 to 3, characterized in that: In S5, the gait-energy consumption mapping model pre-stores multiple gaits and their energy consumption and stability data under different terrain conditions; the optimal walking gait parameters are matched from the gait-energy consumption mapping model based on the characteristics of the terrain section, with the lowest energy consumption or the highest stability as the optimization goal.

6. The robot dog inspection path planning method according to any one of claims 1 to 3, characterized in that: In S5 , the terrain road segment is divided based on whether the average terrain slope and roughness values ​​of a plurality of consecutive nodes on the path fall within a preset range.

7. A robot dog inspection path planning system, characterized in that: The invention comprises a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the robot dog inspection path planning method according to any one of claims 1 to 6 is implemented.

Images