A path planning method and system for a patrol robot

Abstract

The application relates to the technical field of robot path planning, and particularly provides a kind of inspection robot path planning method and system, method includes steps: obtaining environment model and motion model;The preset starting position is selected as the initial starting point, and the node adjacent to the starting point is used as a candidate node;S3, for each candidate node, obtain the cumulative value of the comprehensive transition cost and the heuristic cost, calculate the comprehensive transition cost, then the minimum value of the sum result of the cumulative value of the comprehensive transition cost, the comprehensive transition cost and the heuristic cost is used as the estimated cost;S4, analyze whether the candidate node with the minimum estimated cost is the preset target position, if yes, generate an inspection path according to all selected starting points and the candidate node with the minimum estimated cost, if not, select the candidate node with the minimum estimated cost as a new starting point, the node adjacent to the starting point is used as a candidate node, and return to S3;The method can realize the collaborative optimization of robot motion stability and environmental safety.

Description

Technical Field

[0001] This application relates to the field of robot path planning technology. Specifically, this invention belongs to the field of intelligent sensing and control equipment in the national strategic emerging industries. It relates to a path planning method and system for inspection robots. Background Technology

[0002] With the rapid advancement of smart grid construction, the demand for automation and intelligence in substation equipment inspection is increasing daily. Quadruped robots, with their superior ability to traverse unstructured terrain and their flexible movement, have become an important development direction for substation inspection robots. Compared to traditional wheeled or tracked robots, quadruped robots can autonomously traverse complex terrains such as stairs, cable trenches, gravel roads, and lawns, greatly expanding the inspection coverage. However, substations are extreme industrial environments characterized by high voltage, strong electromagnetic interference, and complex spatial layouts, placing extremely high demands on the safety and reliability of inspection operations. Applying quadruped robots to routine autonomous substation inspections currently faces severe technical challenges. The core contradiction lies in the difficulty of using existing path planning methods to coordinate and optimize the robot's motion safety with the operational safety of the substation environment.

[0003] Existing mobile robot path planning techniques, such as the A* algorithm, Dijkstra's algorithm, and their variants, are mostly based on searching in two-dimensional or three-dimensional geometric space. The core of these methods is finding the shortest or collision-free geometric path from the starting point to the ending point, often simplifying the robot into a point mass or a rigid body with a fixed shape. This simplification completely ignores the essential characteristics of a quadruped robot as a multi-degree-of-freedom dynamic equilibrium system. In the actual terrain of a substation, different gaits exhibit significant differences in stability on the same terrain, and the passability and risks of the same gait vary drastically across different terrains. A geometrically shortest path might require the robot to run on stairs or jog quickly on soft grass, which can easily lead to slipping, instability, or even overturning, directly threatening the robot's safety and task completion. Existing methods, lacking quantitative modeling of the coupling relationship between gait, terrain, and stability, cannot proactively avoid such movement risks during the planning phase.

[0004] Meanwhile, path planning in substations is far more complex than simple obstacle avoidance. Current electrical safety regulations not only stipulate minimum electrical safety distances from energized equipment but also include numerous detailed behavioral guidelines. For example, specific meters must be observed from a specific angle directly in front, and robot speed and posture must be restricted in electromagnetically sensitive areas. Furthermore, equipment status is dynamically changing; the risk level of the surrounding area can increase in real time for a locally overheated disconnect switch or an abnormally noisy transformer. Traditional path planning cost maps typically only handle static, geometric obstacles and cannot transform these complex, non-geometric, dynamically evolving safety constraints into calculable cost factors. This can lead to robots potentially traversing a collision-free path while violating critical safety procedures or failing to avoid sudden risks, potentially causing serious equipment or safety incidents.

[0005] More importantly, the two major challenges mentioned above—robot motion stability and substation environmental safety—are not independent but deeply intertwined and frequently conflicting. For example, to urgently avoid a sudden discharge risk, a robot might need to make a sharp turn onto a rugged, gravel path, significantly increasing its probability of instability. Conversely, to achieve absolute stability on a steep staircase, a robot might be forced to adopt a slow gait, leading to prolonged exposure time in areas with strong electromagnetic fields. Existing multi-objective path planning methods typically use a linear weighted approach with fixed weights to add different costs, and these weights are often set empirically and cannot be dynamically adjusted. This rigid trade-off mechanism cannot adapt to the complex scenarios of substations, where risk conditions change rapidly and motion requirements vary from place to place. It often results in planned paths that perform reasonably well in a single dimension but have fatal flaws at the global level, failing to meet the stringent requirements of absolute safety and flawless operation for substation inspections.

[0006] Therefore, existing technologies suffer from three fundamental deficiencies: First, at the robot model level, there is a lack of a stability cost function capable of accurately quantifying the multi-gait motion characteristics of quadruped robots and their interaction with complex terrain; second, at the environmental modeling level, there is a lack of the ability to real-time, digitally integrate and express the dynamic risk field and structured safety procedures of substations; and third, at the decision optimization level, there is a lack of an optimization mechanism capable of intelligently and dynamically coordinating the two often conflicting objectives of motion stability and environmental safety. These deficiencies severely restrict the practical application and effectiveness of quadruped robots in high-risk scenarios such as substations.

[0007] There is currently no effective technical solution to the above problems. Summary of the Invention

[0008] The purpose of this application is to provide a path planning method and system for inspection robots, which can achieve synergistic optimization of robot motion stability and environmental safety.

[0009] Firstly, this application provides a path planning method for an inspection robot, which includes the following steps: S1. Obtain the environmental model of the area to be inspected and the motion model of the inspection robot. The environmental model includes multiple nodes and their corresponding terrain attribute information and safety constraint information. The motion model includes the stability scores of different motion modes under different terrain attributes. S2. Select the preset starting position as the initial starting point, and select the nodes adjacent to the starting point as candidate nodes; S3. For each candidate node, obtain the cumulative value of the comprehensive transition cost from the preset starting position to the current starting point. Calculate the comprehensive transition cost of the candidate node under different motion modes based on the terrain attribute information, safety constraint information, and real-time risk information corresponding to the candidate node. Obtain the heuristic cost from the candidate node to the preset target position. Then, take the minimum value of the sum of the cumulative value of the comprehensive transition cost, the comprehensive transition cost, and the heuristic cost as the estimated cost of the candidate node. S4. Analyze whether the candidate node with the lowest estimated cost is the preset target location. If yes, generate an inspection path based on all selected starting points and the candidate node with the lowest estimated cost. If no, select the candidate node with the lowest estimated cost as the new starting point, select the node adjacent to the starting point as the candidate node, and return to step S3.

[0010] Optionally, step S3 includes: S31. For each candidate node, obtain the cumulative value of the comprehensive transition cost from the preset starting position to the current starting point; S32. For each candidate node, query the motion model based on the terrain attribute information corresponding to the candidate node to obtain the motion stability score of the candidate node under different motion modes. S33. For each candidate node, obtain the static rule violation cost of the candidate node under different motion modes based on the security constraint information corresponding to the candidate node. S34. For each candidate node, calculate the dynamic safety field cost of the candidate node under different motion modes based on the real-time risk information and preset risk threshold corresponding to the candidate node, as well as the static rule violation cost of the candidate node under different motion modes. S35. For each candidate node, calculate the comprehensive transition cost of the candidate node under different motion modes based on the motion stability score and dynamic safety field cost of the candidate node under different motion modes. S36. For each candidate node, obtain the Euclidean distance from the candidate node to the preset target location, obtain the preset static safety baseline value and optimal stability cost of each node along the path from the candidate node to the preset target location, and then calculate the heuristic cost from the candidate node to the preset target location based on the Euclidean distance, the average of all preset static safety baseline values, and the average of all optimal stability costs; the optimal stability cost is the maximum value among all stability scores of each node along the path from the candidate node to the preset target location under its corresponding terrain attribute information; S37. For each candidate node, the minimum value of the sum of the cumulative value of the comprehensive transition cost, the comprehensive transition cost, and the heuristic cost corresponding to the candidate node is taken as the estimated cost of the candidate node.

[0011] Optionally, the safety constraint information includes a preset minimum safe distance, a preset maximum dwell time on the main channel, a preset maximum included angle deviation, and a preset prohibited movement mode for each candidate node. Step S33 includes: S331. For each candidate node, obtain the actual minimum distance between the candidate node and the surrounding energized equipment, analyze whether the actual minimum distance is less than the preset minimum safe distance. If so, set the static rule violation cost corresponding to the candidate node to infinity and execute step S34. If not, execute step S332. S332. When the current candidate node is located on the preset main inspection channel and the calibration inspection time corresponding to the current candidate node is greater than the preset maximum dwell time of the main channel, calculate the cost of violating the occupancy rule based on the difference between the calibration inspection time and the preset maximum dwell time of the main channel. S333. Assign a mode rule violation cost greater than 0 to the movement mode that conforms to the preset prohibited movement mode corresponding to the current candidate node, and assign a mode rule violation cost of size 0 to the movement mode that conforms to the preset prohibited movement mode corresponding to the current candidate node. S334. Analyze whether the current candidate node is a meter observation point. If yes, proceed to step S336; otherwise, proceed to step S335. S335. For each motion mode corresponding to the current candidate node, calculate the static rule violation cost corresponding to the motion mode based on the occupancy rule violation cost and the mode rule violation cost corresponding to the motion mode, and then execute step S34. S336. Analyze whether the calibration angle deviation corresponding to the current candidate node is greater than the preset maximum angle deviation. If so, calculate the angle rule violation cost based on the difference between the absolute value of the calibration angle deviation and the preset maximum angle deviation. If not, set the angle rule violation cost to 0. The calibration angle deviation is the size of the angle between the observation axis of the inspection robot at the current inspection node and the meter normal, which is pre-calibrated. S337. For each motion mode corresponding to the current candidate node, calculate the static rule violation cost corresponding to the motion mode based on the occupancy rule violation cost, the angle rule violation cost, and the mode rule violation cost corresponding to the motion mode, and then execute step S34.

[0012] Optionally, the formula for calculating the cost of violating the occupancy rule is as follows: ; Where C1(q) represents the cost of violating the occupancy rule corresponding to candidate node q, and t stay This indicates the calibration inspection time corresponding to candidate node q; The formula for calculating the cost of violating the included angle rule is: ; Where C2(q) represents the cost of violating the included angle rule corresponding to candidate node q, and Δθ represents the calibration included angle deviation corresponding to candidate node q, θ max This indicates the preset maximum included angle deviation.

[0013] Optionally, the area to be inspected is a substation, and the preset minimum safe distance is determined based on the voltage level of the substation.

[0014] Optionally, the real-time risk information includes various types of real-time risk values, and the preset risk thresholds include various types of preset risk values. Step S34 includes: S341. For each candidate node, calculate the inherent strength of each type based on the real-time risk value and the preset risk value of each type. Then, calculate the dynamic risk field cost based on the maximum value of the inherent strength and the distance between the candidate node and the risk source corresponding to the maximum value of the inherent strength. S342. For each candidate node, calculate the dynamic safety field cost of the candidate node under different motion modes based on the dynamic risk field cost corresponding to the candidate node and the static rule violation cost of the candidate node under different motion modes. Step S35 includes: S351. For each candidate node, calculate the adaptive weight of the candidate node in different motion modes based on the motion stability score and dynamic safety field cost of the candidate node in different motion modes. S352. For each candidate node, calculate the comprehensive transition cost of the candidate node under different motion modes based on the motion stability score, dynamic safety field cost, and adaptive weight of the candidate node under different motion modes.

[0015] Optionally, the formula for calculating the cost of a dynamic risk field is: ; Among them, C d(q) represents the dynamic risk field cost of inspection node q, λ h d(q, h) represents the maximum value of the inherent strength corresponding to the inspection node q, k represents the preset distance attenuation coefficient, and d(q, h) represents the distance between the candidate node q and the risk source corresponding to the maximum value of its inherent strength. The formula for calculating adaptive weights is: ; Where, α jq Let α0 represent the adaptive weight of candidate node q in the j-th motion mode, β represent the preset base weight, and C represent the preset adjustment coefficient. sj (q) represents the dynamic safety field cost of candidate node q in the j-th motion mode, C gj (q) represents the motion stability score of candidate node q in the j-th motion mode, and ε represents the preset zero-prevention constant; The formula for calculating the overall transition cost is as follows: ; Among them, C j (q) represents the comprehensive transition cost of candidate node q in the j-th motion mode.

[0016] Optionally, the real-time risk information includes four types of real-time risk values: real-time temperature rise data, real-time discharge amplitude, real-time gas concentration, and background noise abrupt change index. The preset risk thresholds include three types of preset risk values: preset maximum temperature rise, preset maximum discharge amplitude, and preset maximum gas concentration. The fixed intensity corresponding to the background noise abrupt change index is the value after normalization of the background noise abrupt change index. The formula for calculating the fixed intensity corresponding to the real-time temperature rise data is as follows: ; Where, λ hT ΔT represents the fixed intensity corresponding to the real-time temperature rise data, and T represents the real-time temperature rise data. max Indicates the preset maximum temperature rise; The formula for calculating the fixed intensity corresponding to the real-time discharge amplitude is: ; Where, λ hA ΔA represents the fixed intensity corresponding to the real-time discharge amplitude, and A represents the real-time discharge amplitude. max Indicates the preset maximum discharge amplitude; The formula for calculating the fixed intensity corresponding to the real-time gas concentration is: ; Where, λ hC CSF represents the fixed intensity corresponding to the real-time gas concentration. maxThis indicates the preset maximum gas concentration.

[0017] Optionally, the inspection robot path planning method also includes the following steps: A1. When controlling the inspection robot to perform inspections according to the inspection path, the sensors on the inspection robot are used to collect real-time data of the path window within a preset distance in front of the current position in order to obtain real-time sensor data. A2. Based on the terrain attribute information and real-time sensor data of the area where the inspection robot is currently located, recalculate the comprehensive transition cost of each sampling node within the path window; A3. If the deviation between the recalculated comprehensive transition cost and the original comprehensive transition cost corresponding to the sampling node is greater than a preset threshold, take the current position of the inspection robot as the temporary starting point and the node on the inspection path located after the end of the path window as the temporary ending point, generate a locally optimized path that connects the temporary starting point and the temporary ending point at both ends and bypasses the current position of the inspection robot, and use the locally optimized path to replace the path in the inspection path that connects the temporary starting point and the temporary ending point, and then control the inspection robot to continue inspection according to the replaced inspection path.

[0018] Secondly, this application also provides a path planning system for an inspection robot, which is used to perform the steps in the path planning method for an inspection robot provided in the first aspect above. The system includes: The acquisition module is used to acquire the environmental model of the area to be inspected and the motion model of the inspection robot. The environmental model includes multiple nodes and their corresponding terrain attribute information and safety constraint information, and the motion model includes the stability scores of different motion modes under different terrain attributes. The selection module is used to select a preset starting position as the initial starting point and select the nodes adjacent to the starting point as candidate nodes. The calculation module is used to obtain the cumulative value of the comprehensive transition cost from the preset starting position to the current starting point for each candidate node, calculate the comprehensive transition cost of the candidate node under different motion modes based on the terrain attribute information, safety constraint information and real-time risk information corresponding to the candidate node, obtain the heuristic cost from the candidate node to the preset target position, and then take the minimum value of the sum of the cumulative value of the comprehensive transition cost, the comprehensive transition cost and the heuristic cost as the estimated cost of the candidate node. The analysis module is used to analyze whether the candidate node with the lowest estimated cost is the preset target location. If so, an inspection path is generated based on all selected starting points and the candidate node with the lowest estimated cost. If not, the candidate node with the lowest estimated cost is selected as the new starting point, the nodes adjacent to the starting point are selected as candidate nodes, and the calculation module is triggered.

[0019] As can be seen from the above, the inspection robot path planning method and system provided in this application achieves the synergistic optimization of robot motion stability and environmental safety by acquiring environmental models and motion models, and calculating comprehensive transition costs based on terrain attributes, safety constraints and real-time risk information. It has the advantages of being able to accurately quantify the stability of robot multi-gait motion characteristics and complex terrain interaction, realizing the digital integration of substation dynamic risk field and structured safety procedures, and dynamically coordinating motion stability and environmental safety, thereby improving the safety and reliability of inspection operations. Attached Figure Description

[0020] Figure 1 A flowchart of a path planning method for an inspection robot provided in an embodiment of this application.

[0021] Figure 2 This is a schematic diagram of the structure of an inspection robot path planning system provided in an embodiment of this application.

[0022] Attached reference numerals: 1. Acquisition module; 2. Selection module; 3. Calculation module; 4. Analysis module. Detailed Implementation

[0023] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0024] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0025] Firstly, such as Figure 1 As shown, this application provides a path planning method for an inspection robot, which includes the following steps: S1. Obtain the environmental model of the area to be inspected and the motion model of the inspection robot. The environmental model includes multiple nodes and their corresponding terrain attribute information and safety constraint information. The motion model includes the stability scores of different motion modes under different terrain attributes. S2. Select the preset starting position as the initial starting point, and select the nodes adjacent to the starting point as candidate nodes; S3. For each candidate node, obtain the cumulative value of the comprehensive transition cost from the preset starting position to the current starting point. Calculate the comprehensive transition cost of the candidate node under different motion modes based on the terrain attribute information, safety constraint information, and real-time risk information corresponding to the candidate node. Obtain the heuristic cost from the candidate node to the preset target position. Then, take the minimum value of the sum of the cumulative value of the comprehensive transition cost, the comprehensive transition cost, and the heuristic cost as the estimated cost of the candidate node. S4. Analyze whether the candidate node with the lowest estimated cost is the preset target location. If yes, generate an inspection path based on all selected starting points and the candidate node with the lowest estimated cost. If no, select the candidate node with the lowest estimated cost as the new starting point, select the node adjacent to the starting point as the candidate node, and return to step S3.

[0026] For ease of understanding, some key terms in this embodiment are explained below. The area to be inspected in this embodiment refers to the specific geographical area where the inspection robot needs to perform its inspection tasks, such as equipment areas or passageways within a substation. This area is typically complex and includes various equipment, obstacles, and terrain variations. The environment model in this embodiment is a digital representation of the area to be inspected, designed to provide necessary spatial and attribute information for path planning. This environment model is preferably a pre-constructed high-precision two-dimensional grid semantic map. The construction of this map is existing technology, and its specific process will not be discussed in detail here. The nodes in this embodiment represent discrete locations where the inspection robot may stop or pass through. For example, they can be individual grid points in a high-precision two-dimensional grid semantic map. Each node is accompanied by terrain attribute information and safety constraint information. The terrain attribute information describes the ground features at the node's location, such as flat ground, stairs, ramps, grass, or gravel roads. This information is crucial for evaluating the stability of the inspection robot's movement pattern at that location. The safety constraint information includes safety regulations and restrictions related to the node, such as minimum safe distances from energized equipment, time limits for staying in specific areas, and prohibited movement patterns. The motion model of the inspection robot in this embodiment describes the motion characteristics and performance of the inspection robot under different motion modes. A motion mode refers to different gaits or movement methods that the robot can adopt, such as walking, jogging, crawling, or slow walking. Different motion modes are suitable for different terrains, and the same motion mode has different stability scores under different terrains (i.e., the stability score of the same motion mode is different under different terrain attributes). The stability score is an indicator that quantifies the stability of the robot when using a specific motion mode under specific terrain attributes. For example, a higher score indicates greater stability, while a lower score indicates greater instability or a risk of instability. This score is obtained by repeatedly conducting a passability test on simulated terrain in a standard test area using a specific motion mode, collecting indicators such as the variance of the robot's attitude angle, the number of foot slippages, and the success rate, and performing normalized fusion processing to obtain the robot's stability score under the current motion mode and current terrain attribute information. The preset starting position in this embodiment is the initial position point where the inspection robot begins its inspection task, and the preset target position in this embodiment is the final position point that the inspection robot needs to reach, which can be the endpoint of the inspection task. In this embodiment, candidate nodes refer to nodes adjacent to the current starting point that may be selected as the next path point during the path planning process. The cumulative comprehensive transition cost in this embodiment represents the sum of the comprehensive transition costs of all nodes traversed on the path from the preset starting position to the current starting point; it reflects the total cost incurred from the starting point to the current point.The comprehensive transition cost in this embodiment represents the cost incurred from the current starting point to a candidate node. Its calculation comprehensively considers the candidate node's terrain attributes, safety constraints, real-time risk information, and the robot's performance under different motion modes. The heuristic cost in this embodiment is an metric for evaluating the estimated cost from the current candidate node to the preset target location. It is typically an optimistic estimate of future costs, used to guide the search direction and improve path planning efficiency. The estimated cost in this embodiment is the minimum value of the sum of the cumulative comprehensive transition cost, the comprehensive transition cost, and the heuristic cost. It is used to evaluate the merits of each candidate node and guide the path search direction. The candidate node with the minimum estimated cost is considered the optimal next path point.

[0027] This embodiment provides a path planning method for inspection robots, aiming to solve the problem that existing path planning methods for inspection robots in complex substation environments struggle to coordinate the optimization of robot motion stability and environmental operational safety, resulting in planned paths that cannot effectively avoid motion risks and violate safety regulations. This method is implemented through the following steps: In step S1, the environmental model of the area to be inspected and the motion model of the inspection robot are acquired. The environmental model includes multiple nodes and their corresponding terrain attribute information and safety constraint information. The motion model includes stability scores for different motion modes under different terrain attributes. For example, the substation area can be scanned using devices such as LiDAR and vision sensors to construct a two-dimensional grid map, and the geographical coordinates, elevation, slope, and other terrain attribute information of each node can be extracted from it. At the same time, equipment areas, restricted areas, inspection channels, etc., within the substation can be manually labeled or identified using image recognition technology, and their corresponding safety constraint information (such as minimum safe distance) can be stored in the environmental model. For the motion model, the inspection robot's motion mode can be tested on different terrains (such as flat ground, gravel roads, and stairs), recording data such as posture changes and foot force under different motion modes (such as pacing, trotting, walking, and running). Based on this data, the stability score of each gait under different terrains can be calculated, and these scores can be stored in a lookup table for subsequent querying.

[0028] In step S2, a preset starting position is selected as the initial starting point, and nodes adjacent to the starting point are selected as candidate nodes. For example, the inspection robot can use the charging pile location as the preset starting position, which is set as the initial starting point. Then, the system identifies all reachable nodes directly connected to this starting point and adds these nodes to the candidate node list.

[0029] In step S3, for each candidate node, the cumulative value of the comprehensive transition cost from the preset starting position to the current starting point is obtained. Based on the terrain attribute information, safety constraint information, and real-time risk information corresponding to the candidate node, the comprehensive transition cost of the candidate node under different movement modes is calculated. The heuristic cost from the candidate node to the preset target position is also obtained. Then, the minimum value of the sum of the cumulative comprehensive transition cost, the comprehensive transition cost, and the heuristic cost is used as the estimated cost of the candidate node. For example, for a candidate node, the system first queries the sum of the comprehensive transition costs of all selected nodes on the path from the preset starting position to the current starting point to obtain the cumulative comprehensive transition cost. Next, based on the candidate node's terrain attributes (e.g., gravel road), safety constraints (e.g., proximity to electrical equipment), and real-time risk information (e.g., local temperature rise anomalies), the comprehensive transition cost of the inspection robot passing through the candidate node under different movement modes (e.g., pacing, jogging) is calculated. Simultaneously, the Euclidean distance from the candidate node to the preset target position (e.g., charging pile location) is calculated as the heuristic cost. Finally, these three costs are added together to obtain the estimated cost of the candidate node.

[0030] In step S4, it is analyzed whether the candidate node with the lowest estimated cost is the preset target location. If so, an inspection path is generated based on all selected starting points and the candidate node with the lowest estimated cost. If not, the candidate node with the lowest estimated cost is selected as the new starting point, and the nodes adjacent to the starting point are selected as candidate nodes, and the process returns to step S3. For example, during the iteration process, if the candidate node with the lowest estimated cost happens to be the preset target location, it means that an optimal path from the starting point to the target point has been found. At this time, the system will backtrack all selected starting points and the final candidate node with the lowest estimated cost to generate a complete inspection path. If the candidate node with the lowest estimated cost is not the target location, the node is selected as the new starting point, and the search for its adjacent candidate nodes continues, repeating step S3 until the candidate node with the lowest estimated cost is the preset target location. It should be understood that if the candidate node with the lowest estimated cost is not the preset target location and all the nodes adjacent to the candidate node with the lowest estimated cost are already selected nodes or unavailable nodes (nodes located in the restricted area), it means that there is no feasible path from the preset starting position to the preset target location.

[0031] The following example provides a more detailed explanation of the above technical solution: Suppose that in a substation area, an inspection robot needs to inspect from a charging pile location (preset starting position) to a distant transformer meter observation point (preset target position). The substation area includes a level cement road, a section of gravel road, a cable trench that needs to be crossed, and some high-voltage live equipment.

[0032] First, in step S1, the system acquires the environmental model of the substation area and the motion model of the inspection robot. In the environmental model, areas such as cement roads, gravel roads, and cable trenches are marked as having different terrain attributes, and the area around high-voltage live equipment is marked as having safety constraint information (e.g., minimum safe distance). In the motion model, the stability scores of the inspection robot's movement modes such as "pacing," "trotting," "slipping," and "running" under different terrains (such as cement roads, gravel roads, and cable trenches) are pre-calibrated and stored. For example, the "running" gait has a higher stability score on cement roads but a lower stability score on gravel roads.

[0033] Next, in step S2, the location of the charging pile is selected as the initial starting point, and several nodes adjacent to the charging pile and located outside the restricted area are identified as candidate nodes.

[0034] Then, the iterative process of step S3 begins. The system evaluates each candidate node. For example, consider candidate node A located on a gravel road. The system first obtains the cumulative value of the comprehensive transition cost from the charging pile to the current starting point. Then, based on the terrain attributes (gravel road), safety constraint information (e.g., proximity to high-voltage equipment), and real-time risk information (e.g., nearby partial discharge signal) of node A, the system calculates the comprehensive transition cost of the inspection robot passing through node A in different motion modes. Since the safety constraint information and real-time risk information of the same node are the same, the comprehensive transition cost of the same node in different motion modes is negatively correlated with the stability score of the motion mode (the higher the stability score, the smaller the comprehensive transition cost). For example, node A is on a gravel road, the stability score of the pacing mode is higher, and the stability score of the trotting mode is lower. Therefore, the comprehensive transition cost of node A in the pacing mode is greater than the stability score of node A in the trotting mode. At the same time, the system calculates the heuristic cost (e.g., Euclidean distance) from node A to the transformer meter observation point. Finally, the cumulative value of the comprehensive transition cost, the comprehensive transition cost, and the heuristic cost are added together to obtain the estimated cost of node A. This process is repeated for all candidate nodes.

[0035] In step S4, the system compares the estimated costs of all candidate nodes and selects the node with the lowest estimated cost. If this node is a transformer meter observation point, path planning is complete, and an inspection path is generated. Otherwise, the node with the lowest estimated cost is used as the new starting point, and its adjacent nodes are used as new candidate nodes, returning to step S3 to continue the iteration. This iterative process continues until the target location is found.

[0036] Through the above process, the method of this embodiment can generate an optimal inspection path. For example, on gravel roads, the system will choose the pacing mode because it has a high stability score on gravel roads and can effectively reduce the risk of robot instability. In areas near high-voltage electrical equipment, even the path with the shortest geometric distance will have its comprehensive transition cost set to infinite if it violates the minimum safe distance, thus being avoided by the planning algorithm. When there is a real-time risk (such as partial discharge), the dynamic safety field cost in this area will increase significantly, causing the planned path to actively avoid high-risk areas.

[0037] The technical solution of this embodiment has significant technical contributions compared to traditional path planning methods. Traditional methods often only focus on the shortest geometric distance or collision-free path, simplifying the robot to a point mass and ignoring the stability issues of quadruped robots' multi-gait motion characteristics and interactions with complex terrain. For example, in the example above, if a traditional method is used, a geometrically shortest path may be planned, but this path may require the robot to adopt an unstable "running" gait on gravel roads or cross the safe distance area of ​​high-voltage equipment, thereby causing the robot to become unstable or violate safety regulations.

[0038] This embodiment acquires the environmental and motion models in step S1, enabling quantitative modeling of terrain attributes, safety constraints, and motion pattern stability scores, providing a comprehensive data foundation for subsequent cost calculations. This allows the planning process to proactively consider the robot's motion safety. In step S3, by calculating the comprehensive transition cost, terrain attributes, safety constraints, and real-time risk information are deeply integrated, allowing the cost function to comprehensively reflect the comprehensive risks faced by the robot. This contrasts sharply with traditional methods that only handle static geometric obstacles, effectively addressing the dynamically changing risk situation and complex safety procedures of substations. Furthermore, by using the minimum value of the sum of the comprehensive transition cost, the comprehensive transition cost, and the heuristic cost as the estimated cost, this embodiment improves search efficiency while ensuring path optimality. This path planning mechanism, which comprehensively considers motion stability and environmental safety, can intelligently and dynamically coordinate the two often contradictory goals of motion stability and environmental safety, thereby generating an inspection path that meets both motion stability requirements and substation safety procedures. This effectively solves the fundamental defects in existing technologies and enhances the practical application and effectiveness of inspection robots in high-risk and complex scenarios.

[0039] In some preferred embodiments, step S3 includes: S31. For each candidate node, obtain the cumulative value of the comprehensive transition cost from the preset starting position to the current starting point; S32. For each candidate node, query the motion model based on the terrain attribute information corresponding to the candidate node to obtain the motion stability score of the candidate node under different motion modes. S33. For each candidate node, obtain the static rule violation cost of the candidate node under different motion modes based on the security constraint information corresponding to the candidate node. S34. For each candidate node, calculate the dynamic safety field cost of the candidate node under different motion modes based on the real-time risk information and preset risk threshold corresponding to the candidate node, as well as the static rule violation cost of the candidate node under different motion modes. S35. For each candidate node, calculate the comprehensive transition cost of the candidate node under different motion modes based on the motion stability score and dynamic safety field cost of the candidate node under different motion modes. S36. For each candidate node, obtain the Euclidean distance from the candidate node to the preset target location, obtain the preset static safety baseline value and optimal stability cost of each node along the path from the candidate node to the preset target location, and then calculate the heuristic cost from the candidate node to the preset target location based on the Euclidean distance, the average of all preset static safety baseline values, and the average of all optimal stability costs; the optimal stability cost is the maximum value among all stability scores of each node along the path from the candidate node to the preset target location under its corresponding terrain attribute information; S37. For each candidate node, the minimum value of the sum of the cumulative value of the comprehensive transition cost, the comprehensive transition cost, and the heuristic cost corresponding to the candidate node is taken as the estimated cost of the candidate node.

[0040] In step S31, the cumulative value of the comprehensive transition cost represents the path cost accumulated from the starting point to the current node. It is a key indicator for evaluating the quality of the path and ensures the continuity and global optimality of the path planning. This cumulative value can be used as the g(n) function in the A* algorithm, which is the actual path cost from the starting node to the current node, or it can be updated by maintaining an open list and a closed list, and updating the cumulative value of the comprehensive transition cost of the parent node each time a node is expanded.

[0041] In step S32, for each candidate node, the motion model is queried based on the terrain attribute information corresponding to that candidate node to obtain the motion stability score of the candidate node under different motion modes. The motion stability score quantifies the stability of the inspection robot when using a specific motion mode on a specific terrain, enabling path planning to take into account the robot's own motion capabilities and actively avoid using inappropriate gaits on unstable terrain. The motion model can be a pre-established database or lookup table that stores stability scores under different combinations of terrain types (e.g., flat concrete, gravel, stairs) and different motion modes (e.g., pacing, trotting, walking, running). Specifically, the formula for calculating the motion stability score can be: Motion stability score = 1 − stability score extracted from the motion model according to the specific motion mode.

[0042] In step S33, for each candidate node, the static rule violation cost of the candidate node under different motion modes is obtained based on the safety constraint information corresponding to the candidate node. The static rule violation cost transforms the fixed safety procedures (e.g., minimum safe distance, prohibited areas) in high-risk environments such as substations into quantifiable costs, ensuring compliance during the inspection robot's movement. The safety constraint information can be stored as specific area markers on a map (e.g., prohibited areas, speed-limited areas) and attributes associated with the node (e.g., distance to energized equipment, whether it is a meter observation point).

[0043] In step S34, for each candidate node, the dynamic safety field cost of the candidate node under different motion modes is calculated based on the real-time risk information, preset risk threshold, and static rule violation cost of the candidate node under different motion modes. The dynamic safety field cost reflects the comprehensive safety of the node after considering real-time risks and static rules, enabling path planning to respond in real time to dynamically changing risks in the environment (e.g., local overheating, abnormal discharge). Real-time risk information can be acquired in real time by sensors (e.g., infrared thermal imagers, partial discharge detectors, gas sensors) and compared with preset thresholds to generate risk intensity.

[0044] In step S35, for each candidate node, the comprehensive transition cost of the candidate node under different motion modes is calculated based on its motion stability score and dynamic safety field cost. The comprehensive transition cost is a combined reflection of the motion stability score and the dynamic safety field cost, achieving a dynamic trade-off between the inspection robot's motion performance and environmental safety, avoiding potential risks associated with single-dimensional optimization. This comprehensive transition cost can be achieved by combining the motion stability score and the dynamic safety field cost through a weighted summation, where the weights can be adjusted according to actual needs.

[0045] In step S36, for each candidate node, the heuristic cost from that candidate node to the preset target position is obtained. The heuristic cost provides guiding information for the path search algorithm, guiding the search process to converge more efficiently to the target position while considering the expected safety and stability along the path. This heuristic cost can be based on geometric distance metrics such as Euclidean distance or Manhattan distance. Alternatively, it can be calculated by combining the estimated safety cost and motion difficulty cost, for example, by averaging the preset static safety baseline values ​​and the optimal stability cost. Specifically, the formula for calculating the heuristic cost is as follows: Where h(n) represents the heuristic cost of candidate node n, and d e (n,G) represents the Euclidean distance from candidate node n to the preset target position G. This represents the average value of the preset static security baseline values ​​for each node along the path from the candidate node to the preset target location G. This represents the average of the sum of optimal stability costs for all nodes along the path from candidate node n to the preset target location. wd, ws, and wg represent the preset heuristic weight coefficients. The preset static safety baseline value is a value manually pre-set in the environmental model. For example, because the risk is high near transformers, the preset static safety baseline value for nodes near transformers is set to 0.8; because the risk is low on open roads, the preset static safety baseline value for nodes on open roads is set to 0.1.

[0046] In step S37, for each candidate node, the minimum value of the sum of the cumulative comprehensive transition cost, the comprehensive transition cost, and the heuristic cost is taken as the estimated cost of that candidate node. The estimated cost is an important indicator used in path planning algorithms to evaluate node priority. It comprehensively considers the cumulative cost from the preset starting position to the current node (the sum of the cumulative comprehensive transition cost and the comprehensive transition cost) and the estimated cost from the current node to the preset target position. The formula for calculating the estimated cost is: Where f(n) represents the estimated cost of candidate node n, g(n) represents the cumulative cost from the preset starting position to candidate node n (the sum of the cumulative value of the comprehensive transition cost corresponding to candidate node n and the comprehensive transition cost), and h(n) represents the heuristic cost.

[0047] The proposed solution constructs a quantitative evaluation system from environmental perception to motion decision-making through the detailed cost calculation steps described above, achieving refined evaluation of inspection paths. By obtaining the cumulative value of the comprehensive transition cost from the starting position to the current starting point, a historical cost benchmark is provided for path search, ensuring the continuity and global optimality of path planning. By querying the motion model based on terrain attributes to obtain motion stability scores, the robot's motion capability is directly linked to specific terrain features, enabling the planning process to proactively avoid terrain areas prone to instability. By obtaining static rule violation costs based on safety constraint information, hard constraints such as power safety regulations are transformed into calculable cost factors, ensuring compliance during the robot's movement. By combining real-time risk information, preset risk thresholds, and static rule violation costs to calculate dynamic safety field costs, real-time response to dynamic risks within the substation is achieved, enhancing the flexibility and safety of path planning. By calculating comprehensive transition costs by integrating motion stability scores and dynamic safety field costs, a dynamic trade-off between motion performance and environmental safety is achieved, avoiding potential risks associated with single-dimensional optimization. By introducing Euclidean distance, the average value of the static safety baseline, and the average value of the optimal stability cost to calculate the heuristic cost, more guiding heuristic information is provided to the search algorithm, enabling the search process to converge to the target position more efficiently while taking into account the safety and stability expectations along the path. Finally, by summing the above costs and taking the minimum value as the estimated cost, a multi-objective collaborative optimization evaluation function is constructed, so that the final generated path can achieve an optimal balance between satisfying safety procedures, ensuring motion stability, and avoiding real-time risks.

[0048] In some preferred embodiments, the safety constraint information includes a preset minimum safety distance, a preset maximum dwell time on the main channel, a preset maximum included angle deviation, and a preset prohibited movement mode corresponding to each candidate node. Step S33 includes: S331. For each candidate node, obtain the actual minimum distance between the candidate node and the surrounding energized equipment, analyze whether the actual minimum distance is less than the preset minimum safe distance. If so, set the static rule violation cost corresponding to the candidate node to infinity and execute step S34. If not, execute step S332. S332. When the current candidate node is located on the preset main inspection channel and the calibration inspection time corresponding to the current candidate node is greater than the preset maximum dwell time of the main channel, calculate the cost of violating the occupancy rule based on the difference between the calibration inspection time and the preset maximum dwell time of the main channel. S333. Assign a mode rule violation cost greater than 0 to the movement mode that conforms to the preset prohibited movement mode corresponding to the current candidate node, and assign a mode rule violation cost of size 0 to the movement mode that conforms to the preset prohibited movement mode corresponding to the current candidate node. S334. Analyze whether the current candidate node is a meter observation point. If yes, proceed to step S336; otherwise, proceed to step S335. S335. For each motion mode corresponding to the current candidate node, calculate the static rule violation cost corresponding to the motion mode based on the occupancy rule violation cost and the mode rule violation cost corresponding to the motion mode, and then execute step S34. S336. Analyze whether the calibration angle deviation corresponding to the current candidate node is greater than the preset maximum angle deviation. If so, calculate the angle rule violation cost based on the difference between the absolute value of the calibration angle deviation and the preset maximum angle deviation. If not, set the angle rule violation cost to 0. The calibration angle deviation is the size of the angle between the observation axis of the inspection robot at the current inspection node and the meter normal, which is pre-calibrated. S337. For each motion mode corresponding to the current candidate node, calculate the static rule violation cost corresponding to the motion mode based on the occupancy rule violation cost, the angle rule violation cost, and the mode rule violation cost corresponding to the motion mode, and then execute step S34.

[0049] The preset minimum safe distance in this embodiment can be set according to the voltage level of the substation equipment. For example, it can be set to 0.7 meters for 10kV equipment and 3.0 meters for 220kV equipment. The preset maximum dwell time in the main channel in this embodiment can be set to 30 seconds to avoid the robot occupying the critical channel for a long time. The preset maximum included angle deviation in this embodiment can be set to 15 degrees to ensure observation accuracy. The preset prohibited movement mode corresponding to each candidate node can specify that certain gaits (such as running) are not allowed based on terrain type (such as slippery ground) or environmental conditions (such as areas with strong electromagnetic interference).

[0050] In step S331, for each candidate node, the system can obtain the actual minimum distance between the candidate node and surrounding electrical equipment. This distance can be measured in real time using the lidar, vision sensor, or ultrasonic sensor mounted on the robot, or the nearest electrical equipment to the inspection robot can be identified from a pre-established environmental model based on the current position of the inspection robot, and the actual minimum distance can be calculated based on the position of the electrical equipment and the current position of the inspection robot. Subsequently, the system will analyze whether the actual minimum distance is less than the preset minimum safe distance. If it is less, it indicates that the candidate node is in a danger zone, and the corresponding static rule violation cost will be set to infinite, thereby forcibly avoiding the node in path planning.

[0051] In step S332, when the current candidate node is identified as being located on the preset main inspection channel, and its corresponding calibration inspection time (i.e., the time required for the robot to perform the inspection task at this node) is greater than the preset maximum dwell time of the main channel, the system will calculate the cost of violating the occupancy rule based on the difference between the calibration inspection time and the preset maximum dwell time of the main channel. This cost is intended to punish the robot for occupying the key channel for a long time in order to maintain the smooth flow of the channel and the inspection efficiency.

[0052] In step S333, for each candidate node, the system can assign a pattern rule violation cost greater than 0 to movement modes that conform to its preset prohibited movement modes. For example, if a node's terrain attribute is slippery ground and the "running" gait is listed as a prohibited movement mode, then the "running" gait will receive a high pattern rule violation cost at that node. Conversely, for movement modes that do not conform to the prohibited modes, their pattern rule violation cost can be set to 0, indicating that the movement mode is allowed at that node.

[0053] In step S334, the system can analyze whether the current candidate node is a preset meter observation point. This can be achieved by analyzing whether the position of the current candidate node is the same as the position of the preset meter observation point in the environment model. For candidate nodes that are not meter observation points, for each corresponding motion mode, the system can calculate the static rule violation cost of the motion mode based on the previously calculated occupancy rule violation cost and the mode rule violation cost corresponding to the motion mode. This is usually achieved by summing to comprehensively reflect the degree of rule violation of the node in a specific motion mode. When the current candidate node is confirmed as a meter observation point, the system will analyze whether the calibration angle deviation corresponding to the candidate node is greater than the preset maximum angle deviation. The calibration angle deviation refers to the angle between the observation axis of the pre-calibrated inspection robot at the current inspection node and the meter normal. If the deviation is greater than the preset maximum angle deviation, the angle rule violation cost is calculated based on the difference between the absolute value of the calibration angle deviation and the preset maximum angle deviation. If it is not greater, the angle rule violation cost is set to 0. This cost is used to make the system more inclined to use nodes that can ensure the robot maintains the correct posture and angle when the observation table is being used (candidate nodes whose calibration angle deviation is less than or equal to the preset maximum angle deviation) to ensure the accuracy of the observation data.

[0054] In step S337, for each candidate node of a meter observation point, the system can calculate the static rule violation cost of that movement mode based on the previously calculated occupancy rule violation cost, angle rule violation cost, and mode rule violation cost corresponding to that movement mode. This calculation can be performed by direct summation to comprehensively evaluate the rule violations of the meter observation point under a specific movement mode.

[0055] The inspection robot path planning method of this application, when calculating the comprehensive transition cost of candidate nodes, introduces multi-dimensional safety constraint information to transform the operating specifications of the power industry into quantifiable cost factors, achieving a deep integration of path planning and safety procedures. Specifically, in step S33, for each candidate node, an electrical safety distance check is first performed. The system obtains the actual minimum distance between the candidate node and surrounding live equipment and compares it with the preset minimum safety distance. If the safety distance requirement is not met, the static rule violation cost of the node is immediately set to infinity, thereby forcibly avoiding all unsafe areas in subsequent path searches and ensuring the electrical safety of robot operations. If the electrical safety distance requirement is met, operating procedures and motion mode restrictions are further considered. The system determines whether the current candidate node is located on the preset main inspection channel and compares it with its calibrated inspection time and the preset maximum dwell time on the main channel. If the robot dwells on this node for too long (calibrated inspection time is greater than the preset maximum dwell time on the main channel), the corresponding occupancy rule violation cost is calculated, so that the system is more inclined to select nodes that will not block the critical channel for a long time, thereby avoiding affecting other operations or causing congestion as much as possible. Simultaneously, based on the preset prohibited movement modes corresponding to the candidate nodes, unsuitable movement modes are assigned a mode rule violation cost greater than 0. This guides the robot to choose a more stable and safer movement mode under the current terrain or environmental conditions, avoiding the risk of instability caused by using inappropriate movement modes. Furthermore, for meter observation points requiring data collection, the system also considers observation quality. By analyzing whether the current candidate node is a meter observation point, if so, the system further calculates the calibration angle deviation between the inspection robot's observation axis and the meter normal when observing at that node. If this deviation exceeds the preset maximum angle deviation, the corresponding angle rule violation cost is calculated, causing the system to favor selecting nodes that allow the robot to collect data in a compliant posture, thus ensuring the effectiveness and data quality of the inspection task. Finally, the above various rule violation costs are combined to calculate the static rule violation cost of the candidate node under different movement modes. This static rule violation cost is then incorporated into the calculation of the dynamic safety field cost, further affecting the calculation of the comprehensive transition cost. In this way, the proposed solution transforms complex, non-geometric, and dynamically evolving safety constraints into computable cost factors, enabling the path planning algorithm to simultaneously consider electrical safety, operational efficiency, motion mode restrictions, and observation quality. This ensures that the planned inspection path is not only geometrically feasible but also strictly adheres to electrical safety operation specifications during actual execution, effectively improving the quality of inspection tasks. This addresses the fundamental deficiency of traditional methods in coordinating the safety of robot motion with the operational safety of the substation environment in high-risk scenarios such as substations.

[0056] In some preferred embodiments, the formula for calculating the cost of violating the occupancy rule is as follows: ; Where C1(q) represents the cost of violating the occupancy rule corresponding to candidate node q, and t stay This indicates the calibration inspection time corresponding to candidate node q; The formula for calculating the cost of violating the included angle rule is: ; Where C2(q) represents the cost of violating the included angle rule corresponding to candidate node q, and Δθ represents the calibration included angle deviation corresponding to candidate node q, θ max This indicates the preset maximum included angle deviation (preferably 15°).

[0057] In some preferred embodiments, the area to be inspected is a substation, and the preset minimum safety distance is determined based on the substation's voltage level. Limiting the inspection area to a substation means that this solution is designed for this specific industrial environment. Substations typically contain high-voltage electrical equipment, complex terrain, electromagnetic interference, and other special environmental factors, placing extremely high demands on the safety and reliability of inspection operations. For example, the area to be inspected could be a 220kV substation or a 110kV substation. The preset minimum safety distance refers to the minimum physical distance that the inspection robot must maintain from live equipment during the inspection process to ensure the robot's safety. This distance is no longer a single fixed value but is dynamically adjusted according to the voltage level of different equipment within the substation. Voltage level is a key indicator for measuring the degree of danger of electrical equipment; the higher the voltage, the more stringent the safety distance requirements. The distance can be determined using a pre-established mapping table of voltage levels and minimum safe distances. This table contains different voltage levels and their corresponding minimum safe distance values. For example, the preset minimum safe distance is 0.7 meters for a 10kV transformer substation; 1.0 meter for a 35kV substation; 1.5 meters for a 110kV substation; 3.0 meters for a 220kV substation; and 5.0 meters for a 500kV substation. When the robot enters a specific voltage level area, it consults this table to obtain the corresponding safe distance. Alternatively, it can be dynamically calculated using a function or algorithm, for example, by using an empirical formula or interpolation algorithm based on the voltage level value to derive the corresponding minimum safe distance.

[0058] The operating principle of this scheme is as follows: First, the execution environment of the inspection task is explicitly limited to a substation. This allows subsequent safety constraints and path planning to fully consider the complex environmental factors unique to substations, such as high voltage and strong electromagnetic fields. Second, during path planning, the system no longer uses a single fixed minimum safety distance, but dynamically adjusts the minimum safety distance that the inspection robot needs to maintain based on the voltage level of the substation. This means that when the inspection robot is inspecting high-voltage equipment areas, the system will automatically set a larger minimum safety distance to ensure electrical safety; while when it is inspecting low-voltage or non-sensitive areas, a relatively smaller minimum safety distance will be set, thereby improving the efficiency and flexibility of path planning while ensuring safety. This mechanism of dynamically adjusting the minimum safety distance based on voltage level is closely integrated with the steps of obtaining safety constraint information and calculating the static rule violation cost in the aforementioned inspection robot path planning method. In step S331, the system obtains the actual minimum distance between the candidate node and surrounding energized equipment and analyzes whether this distance is less than the preset minimum safety distance. Through this scheme, this preset minimum safety distance is no longer a fixed value, but is determined in real time based on the voltage level of the area where the current candidate node is located. Therefore, when the actual minimum spacing is less than the minimum safe distance determined based on the voltage level, the static rule violation cost of the candidate node will be set to infinity, effectively preventing the robot from entering the dangerous area. This synergy enables path planning to achieve refined safety management, providing stronger safety guarantees in high-voltage areas while offering more optional nodes in low-voltage areas.

[0059] In some preferred embodiments, the real-time risk information includes multiple types of real-time risk values, the preset risk threshold includes multiple types of preset risk values, and step S34 includes: S341. For each candidate node, calculate the inherent strength of each type based on the real-time risk value and the preset risk value of each type. Then, calculate the dynamic risk field cost based on the maximum value of the inherent strength and the distance between the candidate node and the risk source corresponding to the maximum value of the inherent strength. S342. For each candidate node, calculate the dynamic safety field cost of the candidate node under different motion modes based on the dynamic risk field cost corresponding to the candidate node and the static rule violation cost of the candidate node under different motion modes. Step S35 includes: S351. For each candidate node, calculate the adaptive weight of the candidate node in different motion modes based on the motion stability score and dynamic safety field cost of the candidate node in different motion modes. S352. For each candidate node, calculate the comprehensive transition cost of the candidate node under different motion modes based on the motion stability score, dynamic safety field cost, and adaptive weight of the candidate node under different motion modes.

[0060] In step S341, for each candidate node, the inherent strength of various types is calculated based on real-time risk values ​​and preset risk values. Inherent strength is an indicator that measures the potential hazard level of a specific risk source. Its calculation can employ a normalization function, such as truncating the ratio of the real-time risk value to a threshold or mapping it to the [0,1] interval. Subsequently, the dynamic risk field cost is calculated based on the maximum value of the inherent strength and the distance between the candidate node and the risk source corresponding to the maximum value of the inherent strength.

[0061] In step S342, for each candidate node, the dynamic safety field cost is calculated based on the dynamic risk field cost corresponding to the candidate node and the static rule violation cost of the candidate node under different motion modes. This step combines the real-time perceived dynamic risk with the preset static safety rules to form a comprehensive safety assessment index. The dynamic safety field cost can be calculated by directly adding the dynamic risk field cost to the static rule violation cost.

[0062] In step S351, for each candidate node, an adaptive weight is calculated based on its motion stability score and dynamic safety field cost under different motion modes. This step introduces an adaptive weighting mechanism to dynamically balance the considerations of motion stability and environmental safety during the path planning process of the inspection robot. The adaptive weights can be adjusted based on the ratio or difference between the motion stability score and the dynamic safety field cost. For example, when the dynamic safety field cost is high, the weight of the dynamic safety field cost is increased; when the motion stability score is high, the weight of the motion stability cost is increased.

[0063] In step S352, for each candidate node, the comprehensive transition cost of the candidate node under different motion modes is calculated based on its motion stability score, dynamic safety field cost, and adaptive weights. The comprehensive transition cost can be calculated using a weighted summation, where the adaptive weights are used to adjust the relative importance of the motion stability score and the dynamic safety field cost.

[0064] This solution refines and dynamically processes the calculation of the comprehensive transition cost in step S3 of the inspection robot's path planning, through the aforementioned steps. First, by using real-time risk information and preset risk thresholds, it can perceive potential dynamic threats in the environment in real time and quantify them as inherent strengths, then combine this with spatial distance to construct a dynamic risk field cost. This mechanism allows path planning to move beyond static environmental information and respond promptly to sudden and localized risks. Second, the dynamic risk field cost is integrated with the static rule violation cost to form a dynamic safety field cost, thereby unifying the real-time safety status of the environment and preset safety procedures into a calculable cost factor. More importantly, this solution introduces an adaptive weighting mechanism, which dynamically adjusts the weights based on the motion stability scores and dynamic safety field costs of candidate nodes under different motion modes. This means that when the dynamic safety field is high, the adaptive weights increase the importance of the dynamic safety field cost in calculating the comprehensive transition cost, causing the system to tend to select nodes with higher dynamic safety field costs (equivalent to safer paths); when the motion stability score is high, the adaptive weights increase the importance of the motion stability score in calculating the comprehensive transition cost, causing the system to tend to select more stable nodes and motion modes. Ultimately, based on motion stability scores, dynamic safety field costs, and dynamically calculated adaptive weights, the transition costs for each motion mode are comprehensively derived, ensuring that the robot can select the optimal path that conforms to safety regulations and maintains motion stability under different motion modes. This dynamic optimization mechanism, combined with the basic A* algorithm framework (such as S1-S4) and the calculation of motion stability scores and static rule violation costs in S3 (such as S32 and S33), enables the generation of inspection paths to not only consider the shortest path and static obstacle avoidance, but also incorporate in-depth consideration of the robot's motion characteristics and the dynamic environmental safety of the substation. This effectively improves the robot's autonomous decision-making ability and environmental adaptability in complex industrial environments, solving the problem that traditional methods struggle to coordinate the optimization of robot motion safety and substation environmental operational safety.

[0065] In some preferred embodiments, the formula for calculating the cost of the dynamic risk field is as follows: ; Among them, C d (q) represents the dynamic risk field cost of inspection node q, λ h d(q, h) represents the maximum value of the inherent strength corresponding to the inspection node q, k represents the preset distance attenuation coefficient (preferably 0.5), and d(q, h) represents the distance between the candidate node q and the risk source corresponding to the maximum value of its inherent strength. The formula for calculating adaptive weights is: ; Where, αjq Let α0 represent the adaptive weight of candidate node q in the j-th motion mode, β represent the preset base weight (preferably 0.8), and C represent the preset adjustment coefficient (preferably 1.2). sj (q) represents the dynamic safety field cost of candidate node q in the j-th motion mode, C gj (q) represents the motion stability score of candidate node q in the j-th motion mode, and ε represents the preset zero-prevention constant (preferably 0.0001). The formula for calculating the overall transition cost is as follows: ; Among them, C j (q) represents the comprehensive transition cost of candidate node q in the j-th motion mode.

[0066] In some preferred embodiments, real-time risk information includes four types of real-time risk values: real-time temperature rise data, real-time discharge amplitude, real-time gas concentration, and background noise abrupt change index. Preset risk thresholds include three types of preset risk values: preset maximum temperature rise, preset maximum discharge amplitude, and preset maximum gas concentration. The fixed intensity corresponding to the background noise abrupt change index is the value after normalization of the background noise abrupt change index. The formula for calculating the fixed intensity corresponding to the real-time temperature rise data is as follows: ; Where, λ hT ΔT represents the fixed intensity corresponding to the real-time temperature rise data, and T represents the real-time temperature rise data. max This indicates the preset maximum temperature rise (preferably 30°C). The formula for calculating the fixed intensity corresponding to the real-time discharge amplitude is: ; Where, λ hA ΔA represents the fixed intensity corresponding to the real-time discharge amplitude, and A represents the real-time discharge amplitude. max This indicates the preset maximum discharge amplitude (preferably 80dB). The formula for calculating the fixed intensity corresponding to the real-time gas concentration is: ; Where, λ hC CSF represents the fixed intensity corresponding to the real-time gas concentration. max This indicates the preset maximum gas concentration (preferably 1000 ppm). The background noise abrupt change index in this embodiment is obtained from the output of an existing acoustic sensor.

[0067] In some preferred embodiments, the inspection robot path planning method further includes the following steps: A1. When controlling the inspection robot to perform inspections according to the inspection path, the sensors on the inspection robot are used to collect real-time data of the path window within a preset distance in front of the current position in order to obtain real-time sensor data. A2. Based on the terrain attribute information and real-time sensor data of the area where the inspection robot is currently located, recalculate the comprehensive transition cost of each sampling node within the path window; A3. If the deviation between the recalculated comprehensive transition cost and the original comprehensive transition cost corresponding to the sampling node is greater than a preset threshold, take the current position of the inspection robot as the temporary starting point and the node on the inspection path located after the end of the path window as the temporary ending point, generate a locally optimized path that connects the temporary starting point and the temporary ending point at both ends and bypasses the current position of the inspection robot, and use the locally optimized path to replace the path in the inspection path that connects the temporary starting point and the temporary ending point, and then control the inspection robot to continue inspection according to the replaced inspection path.

[0068] In step A1, the sensors on the inspection robot refer to hardware devices used to perceive environmental information. These may include, but are not limited to: LiDAR, used to acquire accurate 3D point cloud data to construct local environmental maps and detect obstacles; depth cameras, used to acquire images and depth information to identify terrain features and potential risk sources; ultrasonic sensors, used for near-range obstacle detection; infrared sensors, used for temperature anomaly detection; gas concentration sensors, used to collect the concentration of leaked gases; and ultrasonic discharge detection components, used to collect the amplitude of discharge behavior. Through these sensors, the inspection robot can acquire detailed information about its surrounding environment in real time. The path window within a preset distance in front of the current position refers to a finite-length area extending from the current position of the inspection robot along the current inspection path. This preset distance can be set according to factors such as the inspection robot's movement speed, sensor sensing range, and frequency of environmental changes; for example, it can be set to 3 meters. Real-time data acquisition refers to the process by which sensors continuously acquire environmental information during the inspection robot's movement. This differs from a pre-established static environmental model; instead, it dynamically captures the instantaneous state of the environment. Real-time sensor data refers to the raw or pre-processed data acquired in real time by the aforementioned sensors, reflecting the current environmental state. These data can include 3D point cloud data of local terrain, image data, temperature data, gas concentration data, etc., which are the basis for subsequent path recalculation and decision-making.

[0069] In step A2, the terrain attribute information of the area where the inspection robot is currently located refers to the physical characteristics of the ground, such as the type, slope, and flatness, of the ground in the robot's current location and its surrounding area. This information can be obtained from a pre-established environmental model or updated or corrected using real-time sensor data. For example, terrain attributes could be flat concrete, gravel, grass, or stairs. Recalculating the comprehensive transition cost of each sampling node within the path window refers to re-evaluating the movement and safety costs of each potential path point (sampling node) within the path window during the inspection process, using the latest environmental perception data. The calculation method for the comprehensive transition cost is the same as in the initial path planning stage, but this time it incorporates real-time sensor data to update terrain attributes, safety constraints, and real-time risk information, thereby obtaining a cost assessment that more closely reflects the current actual environment.

[0070] In step A3, if the deviation between the recalculated comprehensive transition cost and the original comprehensive transition cost corresponding to the sampling node is greater than a preset threshold, a locally optimized path is generated, using the current position of the inspection robot as a temporary starting point and the node located after the end of the path window on the inspection path as a temporary ending point, connecting the temporary starting point and the temporary ending point respectively and bypassing the current position of the inspection robot. This locally optimized path replaces the path connecting the temporary starting point and the temporary ending point in the inspection path, and the inspection robot continues its inspection according to the replaced path. A deviation between the recalculated comprehensive transition cost and the original comprehensive transition cost corresponding to the sampling node being greater than a preset threshold is one of the conditions for triggering local path replanning. The preset threshold is a parameter used to determine whether environmental changes are significant enough to warrant replanning. For example, when the ratio of the deviation between the recalculated comprehensive cost and the original comprehensive transition cost to the original comprehensive transition cost exceeds 30%, local optimized path planning is triggered. Using the current position of the inspection robot as a temporary starting point means that when replanning is triggered, the actual current position of the inspection robot is used as the starting point for the new local path planning. Using a node after the end of the path window on the inspection path as a temporary endpoint means that when replanning is triggered, a node in the original inspection path that is after the end of the current path window is selected as the endpoint for local path planning. Generating a locally optimized path that connects the temporary start point and the temporary endpoint and bypasses the current position of the inspection robot means that after determining the temporary start point and the temporary endpoint, a path planning algorithm is used to search for an optimal path from the temporary start point to the temporary endpoint within a local area, bypassing the current position of the inspection robot. Replacing the path connecting the temporary start point and the temporary endpoint in the inspection path with the locally optimized path means replacing the portion of the original inspection path between the temporary start point and the temporary endpoint with the newly generated locally optimized path. Controlling the inspection robot to continue inspection according to the replaced inspection path means that after the path replacement is completed, the inspection robot will continue to execute the inspection task according to the updated inspection path.

[0071] This method involves the inspection robot collecting real-time data from a path window within a preset distance ahead of its current position using its onboard sensors while it is inspecting along a pre-planned path. This data, combined with terrain information of the robot's current location, is used to recalculate the comprehensive transition cost of each sampling node within the path window. This recalculation of the comprehensive transition cost reflects dynamic environmental changes, such as sudden obstacles, risk sources, or terrain changes, ensuring that the cost assessment always aligns with the actual environment. The system then compares the recalculated comprehensive transition cost with the original comprehensive transition cost corresponding to that sampling node. If this deviation exceeds a preset threshold, it indicates that the environmental changes have reached a point where path adjustment is necessary, triggering a local path replanning. During replanning, the robot's current position is set as a temporary starting point, while nodes on the original inspection path located after the end of the path window are selected as temporary endpoints. Based on these two temporary points, the system generates a new locally optimized path that fully considers the latest real-time environmental data and updated cost information while bypassing the robot's current position. Finally, this locally optimized path replaces the portion of the original inspection path between the temporary starting point and the temporary ending point, allowing the inspection robot to continue its inspection along the updated path. Through this mechanism, the proposed solution elevates the inspection robot's path planning method from static planning to dynamic adaptive planning. Building upon the foundation of basic path planning methods that comprehensively consider motion stability, environmental safety constraints, and real-time risk information, this solution further introduces real-time perception and local replanning capabilities. This means that even in unforeseen circumstances not fully anticipated during the initial planning phase, the robot can respond through real-time perception and rapid local adjustments, avoiding path failure or reduced safety due to environmental changes. This dynamic adjustment capability enables the inspection robot to not only plan an optimal path under initial conditions in the complex and ever-changing substation environment but also continuously optimize it during execution, ensuring the continuity, safety, and efficiency of the inspection task, thus effectively addressing the shortcomings of traditional methods in adaptability to dynamic environments.

[0072] Secondly, such as Figure 2 As shown, this application also provides a path planning system for an inspection robot, which is used to perform the steps in the path planning method for an inspection robot provided in the first aspect above. The system includes: Module 1 is used to acquire the environmental model of the area to be inspected and the motion model of the inspection robot. The environmental model includes multiple nodes and their corresponding terrain attribute information and safety constraint information, and the motion model includes the stability scores of different motion modes under different terrain attributes. Selection module 2 is used to select a preset starting position as the initial starting point and select the nodes adjacent to the starting point as candidate nodes; The calculation module 3 is used to obtain the cumulative value of the comprehensive transition cost from the preset starting position to the current starting point for each candidate node, calculate the comprehensive transition cost of the candidate node under different motion modes based on the terrain attribute information, safety constraint information and real-time risk information corresponding to the candidate node, obtain the heuristic cost from the candidate node to the preset target position, and then take the minimum value of the sum of the cumulative value of the comprehensive transition cost, the comprehensive transition cost and the heuristic cost as the estimated cost of the candidate node. Analysis module 4 is used to analyze whether the candidate node with the lowest estimated cost is the preset target location. If so, an inspection path is generated based on all selected starting points and the candidate node with the lowest estimated cost. If not, the candidate node with the lowest estimated cost is selected as the new starting point, the node adjacent to the starting point is selected as the candidate node, and the calculation module 3 is triggered.

[0073] The inspection robot path planning system provided in this embodiment includes an acquisition module 1, a selection module 2, a calculation module 3, and an analysis module 4. The inspection robot path planning system provided in this embodiment is used to execute the steps in the inspection robot path planning method provided in the first aspect above. The principle of the inspection robot path planning system provided in this embodiment is the same as the principle of the inspection robot path planning method provided in the first aspect above, and will not be repeated here.

[0074] As can be seen from the above, the inspection robot path planning method and system provided in this application achieves the synergistic optimization of robot motion stability and environmental safety by acquiring environmental models and motion models, and calculating comprehensive transition costs based on terrain attributes, safety constraints and real-time risk information. It has the advantages of being able to accurately quantify the stability of robot multi-gait motion characteristics and complex terrain interaction, realizing the digital integration of substation dynamic risk field and structured safety procedures, and dynamically coordinating motion stability and environmental safety, thereby improving the safety and reliability of inspection operations.

[0075] In the embodiments provided in this application, it should be understood that relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0076] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A path planning method for an inspection robot, characterized in that, The inspection robot path planning method includes the following steps: S1. Obtain the environmental model of the area to be inspected and the motion model of the inspection robot; the environmental model includes multiple nodes and their corresponding terrain attribute information and safety constraint information, and the motion model includes stability scores of different motion modes under different terrain attributes. S2. Select the preset starting position as the initial starting point, and select the nodes adjacent to the starting point as candidate nodes; S3. For each candidate node, obtain the cumulative value of the comprehensive transition cost from the preset starting position to the current starting point. Calculate the comprehensive transition cost of the candidate node under different motion modes based on the terrain attribute information, safety constraint information, and real-time risk information corresponding to the candidate node. Obtain the heuristic cost from the candidate node to the preset target position. Then, take the minimum value of the sum of the cumulative value of the comprehensive transition cost, the comprehensive transition cost, and the heuristic cost as the estimated cost of the candidate node. S4. Analyze whether the candidate node with the lowest estimated cost is the preset target location. If yes, generate an inspection path based on all selected starting points and the candidate node with the lowest estimated cost. If no, select the candidate node with the lowest estimated cost as the new starting point, select the node adjacent to the starting point as the candidate node, and return to step S3. Step S3 includes: S31. For each candidate node, obtain the cumulative value of the comprehensive transition cost from the preset starting position to the current starting point; S32. For each candidate node, query the motion model according to the terrain attribute information corresponding to the candidate node to obtain the motion stability score of the candidate node under different motion modes. S33. For each candidate node, obtain the static rule violation cost of the candidate node under different motion modes based on the security constraint information corresponding to the candidate node. S34. For each candidate node, calculate the dynamic safety field cost of the candidate node in different motion modes based on the real-time risk information and preset risk threshold corresponding to the candidate node, as well as the static rule violation cost of the candidate node in different motion modes. S35. For each candidate node, calculate the comprehensive transition cost of the candidate node under different motion modes based on the motion stability score and dynamic safety field cost of the candidate node under different motion modes. S36. For each candidate node, obtain the Euclidean distance from the candidate node to the preset target location, obtain the preset static safety baseline value and optimal stability cost of each node passed from the candidate node to the preset target location, and then calculate the heuristic cost from the candidate node to the preset target location based on the Euclidean distance, the average of all the preset static safety baseline values, and the average of all the optimal stability costs; the optimal stability cost is the maximum value among all the stability scores of each node passed from the candidate node to the preset target location under its corresponding terrain attribute information; S37. For each candidate node, the minimum value of the sum of the cumulative value of the comprehensive transition cost, the comprehensive transition cost, and the heuristic cost corresponding to the candidate node is taken as the estimated cost of the candidate node. The safety constraint information includes a preset minimum safety distance, a preset maximum dwell time on the main channel, a preset maximum included angle deviation, and a preset prohibited movement mode corresponding to each candidate node. Step S33 includes: S331. For each candidate node, obtain the actual minimum distance between the candidate node and the surrounding energized equipment, analyze whether the actual minimum distance is less than the preset minimum safe distance. If so, set the static rule violation cost corresponding to the candidate node to infinity and execute step S34. If not, execute step S332. S332. When the current candidate node is located on the preset main inspection channel and the calibration inspection time corresponding to the current candidate node is greater than the preset main channel maximum dwell time, calculate the occupancy rule violation cost based on the difference between the calibration inspection time and the preset main channel maximum dwell time. S333. Assign a mode rule violation cost greater than 0 to the movement mode that conforms to the preset prohibited movement mode corresponding to the current candidate node, and assign a mode rule violation cost of size 0 to the movement mode that conforms to the preset prohibited movement mode corresponding to the current candidate node. S334. Analyze whether the current candidate node is a meter observation point. If yes, proceed to step S336; otherwise, proceed to step S335. S335. For each motion mode corresponding to the current candidate node, calculate the static rule violation cost corresponding to the motion mode based on the occupancy rule violation cost and the mode rule violation cost corresponding to the motion mode, and then execute step S34. S336. Analyze whether the calibration angle deviation corresponding to the current candidate node is greater than the preset maximum angle deviation. If so, calculate the angle rule violation cost based on the difference between the absolute value of the calibration angle deviation and the preset maximum angle deviation. If not, set the angle rule violation cost to 0. The calibration angle deviation is the pre-calibrated angle between the observation axis of the inspection robot at the current inspection node and the meter normal. S337. For each motion mode corresponding to the current candidate node, calculate the static rule violation cost corresponding to the motion mode based on the occupancy rule violation cost, the angle rule violation cost, and the mode rule violation cost corresponding to the motion mode, and then execute step S34. The formula for calculating the cost of violating the occupancy rule is as follows: ； Where C1(q) represents the cost of violating the occupancy rule corresponding to candidate node q, and t stay This indicates the calibration inspection time corresponding to candidate node q; The formula for calculating the cost of violating the included angle rule is as follows: ； Where C2(q) represents the cost of violating the included angle rule corresponding to candidate node q, and Δθ represents the calibration included angle deviation corresponding to candidate node q, θ max Indicates the preset maximum included angle deviation; The real-time risk information includes various types of real-time risk values, and the preset risk threshold includes various types of preset risk values. Step S34 includes: S341. For each candidate node, calculate the inherent strength of each type based on the real-time risk value and the preset risk value, and then calculate the dynamic risk field cost based on the maximum value of the inherent strength and the distance between the candidate node and the risk source corresponding to the maximum value of the inherent strength. S342. For each candidate node, calculate the dynamic safety field cost of the candidate node in different motion modes based on the dynamic risk field cost corresponding to the candidate node and the static rule violation cost of the candidate node in different motion modes. Step S35 includes: S351. For each candidate node, calculate the adaptive weight of the candidate node in different motion modes based on the motion stability score and dynamic safety field cost of the candidate node in different motion modes. S352. For each candidate node, calculate the comprehensive transition cost of the candidate node under different motion modes based on the motion stability score, dynamic safety field cost, and adaptive weight of the candidate node under different motion modes. The formula for calculating the cost of the dynamic risk field is as follows: ； Among them, C d (q) represents the dynamic risk field cost of inspection node q, λ h d(q, h) represents the maximum value of the inherent strength corresponding to the inspection node q, k represents the preset distance attenuation coefficient, and d(q, h) represents the distance between the candidate node q and the risk source corresponding to the maximum value of its inherent strength. The formula for calculating the adaptive weight is: ； Where, α jq Let α0 represent the adaptive weight of candidate node q in the j-th motion mode, β represent the preset base weight, and C represent the preset adjustment coefficient. sj (q) represents the dynamic safety field cost of candidate node q in the j-th motion mode, C gj (q) represents the motion stability score of candidate node q in the j-th motion mode, and ε represents the preset zero-prevention constant; The formula for calculating the comprehensive transition cost is as follows: ； Among them, C j (q) represents the comprehensive transition cost of candidate node q in the j-th motion mode.

2. The inspection robot path planning method according to claim 1, characterized in that, The area to be inspected is a substation, and the preset minimum safe distance is determined based on the voltage level of the substation.

3. The inspection robot path planning method according to claim 1, characterized in that, The real-time risk information includes four types of real-time risk values: real-time temperature rise data, real-time discharge amplitude, real-time gas concentration, and background noise mutation index. The preset risk thresholds include three types of preset risk values: preset maximum temperature rise, preset maximum discharge amplitude, and preset maximum gas concentration. The fixed intensity corresponding to the background noise mutation index is the value after normalization of the background noise mutation index. The formula for calculating the fixed intensity corresponding to the real-time temperature rise data is as follows: ； Where, λ hT ΔT represents the fixed intensity corresponding to the real-time temperature rise data, and T represents the real-time temperature rise data. max Indicates the preset maximum temperature rise; The formula for calculating the fixed intensity corresponding to the real-time discharge amplitude is: ； Where, λ hA ΔA represents the fixed intensity corresponding to the real-time discharge amplitude, and A represents the real-time discharge amplitude. max Indicates the preset maximum discharge amplitude; The formula for calculating the fixed intensity corresponding to the real-time gas concentration is: ； Where, λ hC The constant intensity corresponding to the real-time gas concentration is represented by CSF, which represents the real-time gas concentration. max This indicates the preset maximum gas concentration.

4. The inspection robot path planning method according to claim 1, characterized in that, The inspection robot path planning method also includes the following steps: A1. When controlling the inspection robot to perform inspections according to the inspection path, the sensors on the inspection robot are used to collect real-time data of the path window within a preset distance in front of the current position to obtain real-time sensor data. A2. Based on the terrain attribute information of the area where the inspection robot is currently located and the real-time sensor data, recalculate the comprehensive transition cost of each sampling node within the path window; A3. If the deviation between the recalculated comprehensive transition cost and the original comprehensive transition cost corresponding to the sampling node is greater than a preset threshold, the current position of the inspection robot is taken as a temporary starting point, and the node located after the end of the path window on the inspection path is taken as a temporary ending point. A locally optimized path is generated with the two ends connecting the temporary starting point and the temporary ending point respectively and bypassing the current position of the inspection robot. The locally optimized path is used to replace the path in the inspection path that connects the temporary starting point and the temporary ending point. Then, the inspection robot is controlled to continue inspection according to the replaced inspection path.

5. A path planning system for an inspection robot, characterized in that, The inspection robot path planning system is used to execute the steps in the inspection robot path planning method of claim 1, and the inspection robot path planning system includes: The acquisition module is used to acquire the environmental model of the area to be inspected and the motion model of the inspection robot. The environmental model includes multiple nodes and their corresponding terrain attribute information and safety constraint information, and the motion model includes stability scores of different motion modes under different terrain attributes. The selection module is used to select a preset starting position as the initial starting point and select the nodes adjacent to the starting point as candidate nodes. The calculation module is used to obtain the cumulative value of the comprehensive transition cost from the preset starting position to the current starting point for each candidate node, calculate the comprehensive transition cost of the candidate node under different motion modes based on the terrain attribute information, safety constraint information and real-time risk information corresponding to the candidate node, obtain the heuristic cost from the candidate node to the preset target position, and then take the minimum value of the sum of the cumulative value of the comprehensive transition cost, the comprehensive transition cost and the heuristic cost as the estimated cost of the candidate node. The analysis module is used to analyze whether the candidate node with the lowest estimated cost is the preset target location. If so, an inspection path is generated based on all selected starting points and the candidate node with the lowest estimated cost. If not, the candidate node with the lowest estimated cost is selected as the new starting point, the node adjacent to the starting point is selected as the candidate node, and the calculation module is triggered.

Images