Cable tunnel robot path generation method based on environmental perception and dynamic planning

Abstract

The invention relates to the technical field of robots, and discloses a cable tunnel robot path generation method based on environment perception and dynamic planning, and the method comprises the steps: an environment perception module is used for collecting environment data in a cable tunnel, and uploading the original data to a target environment database; the path planning module is used for extracting environmental parameters of the same area and calculating a real-time path feasibility index of each node; the map modeling module is used for constructing a three-dimensional topological map of the tunnel and marking obstacle positions and passable areas; the path deviation calculation module is used for calculating a deviation threshold range of each path node; the path generation and dynamic updating module is used for dynamically planning a path and adjusting the path; information such as cable tunnel terrain and obstacles is obtained in real time through the environment sensing technology, the optimal path is rapidly generated in combination with a dynamic planning algorithm, the robot inspection efficiency is effectively improved, the collision risk is reduced, and it is guaranteed that cable operation and maintenance are safe and efficient.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of robots, and more particularly, to a cable tunnel robot path generation method based on environment perception and dynamic planning. BACKGROUND

[0002] Under the background of deep integration of digital economy and real economy, as an important infrastructure for urban power transmission, the intelligent operation and maintenance demand of cable tunnel is increasingly urgent. The cable tunnel robot has gradually become an important tool for cable inspection due to its advantages of high efficiency and safety, and the multi-dimensional information such as environment perception data and path planning data generated during the robot operation provides data support for realizing accurate path planning and ensuring safe operation of the cable.

[0003] At present, in the field of cable tunnel robot path planning, the existing technologies mainly include navigation methods based on fixed routes and traditional path planning algorithms. Among them, the navigation method based on fixed routes usually follows the pre-set inspection route to let the robot execute the task step by step, while the traditional path planning algorithm, such as Dijkstra algorithm and A* algorithm, calculates the path from the starting point to the ending point based on the pre-set map. In addition, some schemes also adjust the advancing route of the robot in real time through artificial remote control.

[0004] However, there are still some shortcomings in actual use. First, the navigation method based on fixed routes is too dependent on the pre-set path and cannot cope with the dynamic changes of the environment in the cable tunnel, such as sudden water accumulation, collapse, and temporary obstacles, which leads to the robot stopping or even colliding when encountering obstacles; second, although the traditional path planning algorithm can find the optimal path to a certain extent, it lacks real-time perception ability of the environment, is difficult to quickly adapt to the complex and changeable tunnel environment, has high computational complexity, and low path planning efficiency, which cannot meet the real-time requirement; third, the artificial remote control method not only increases the labor cost, but also has the problem of not responding in time, which is difficult to guarantee the safe operation of the robot in dangerous environment, resulting in low efficiency of cable inspection and failure to achieve the goal of intelligent and efficient operation and maintenance. SUMMARY

[0005] Therefore, the embodiment of the present application provides a cable tunnel robot path generation method based on environment perception and dynamic planning, which effectively solves the problems raised in the background art.

[0006] To achieve the above purpose, the present application provides the following technical scheme: S1: The environment perception module collects the environment data in the cable tunnel and uploads the original data to the target environment database; S2: The path planning module extracts the environment parameters of the same area from the target environment database according to the area range, and transmits them to the map modeling module; S3: The map modeling module constructs a three-dimensional topological map of the tunnel according to the extracted environment parameters, and marks the obstacle positions, passable areas and path nodes; S4: The real-time environment data acquisition module acquires the environment data of the current position of the robot in real time, and transmits them to the path planning module to calculate the real-time path feasibility index of each path node in the current environment; S5: The path deviation calculation module compares the real-time path feasibility index with the theoretical path benchmark, and calculates the deviation threshold range of each path node; S6: The path deviation index analysis module analyzes the path deviation index based on the deviation threshold range of multiple path nodes, and comprehensively analyzes the deviation degree of the current environment and the theoretical model; S7: The path generation and dynamic updating module calculates the optimal path again by triggering the dynamic planning algorithm if the deviation exceeds the threshold, or performs local adjustment based on the theoretical path if the deviation is small.

[0007] Technical effects and advantages of the present application: 1. The environment perception module and the real-time environment data acquisition module of the present application work together to comprehensively collect cable tunnel environment data, which is not limited to traditional fixed route navigation and simple path planning method. On the one hand, it can accurately identify dynamic obstacles and complex terrain changes in the tunnel, and on the other hand, it can improve the accuracy of quantitative evaluation of the deviation between the actual environment and the theoretical model through path deviation calculation and analysis, and ensure the reliability and safety of robot path planning; 2. The three-dimensional topological map constructed based on the map modeling module, combined with the dynamic planning algorithm for path generation and updating, can quickly generate an optimal detour path for sudden conditions in the cable tunnel. The dynamic adjustment mechanism is triggered by the path deviation index to ensure that the robot always travels along the best route in complex environments. Compared with traditional algorithms, the path planning efficiency is greatly improved, the invalid movement distance and energy consumption of the robot are reduced, the collision risk is reduced, and the execution efficiency and success rate of the cable inspection task are significantly improved; 3. The path deviation index analysis module of the present application comprehensively evaluates the environmental changes and dynamically adjusts the path planning strategy accordingly to realize high-precision real-time path optimization. It avoids the delay and equipment loss caused by using fixed paths or inefficient algorithms, effectively reduces the need for manual intervention, reduces operation and maintenance costs, and improves the adaptability of the cable tunnel robot to different complex environments, providing efficient and reliable technical support for the intelligent operation and maintenance of cable facilities. BRIEF DESCRIPTION OF DRAWINGS

[0008] Figure 1This is a schematic diagram of the overall structure of the present invention.

[0009] Figure 2 This is a flowchart illustrating the overall structure of the present invention.

[0010] Figure 3 A flowchart for constructing a three-dimensional topology map for this invention is provided. Detailed Implementation

[0011] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0012] As attached Figure 1 The cable tunnel robot path generation method shown includes: an environmental perception module, a target environment database, a path planning module, a map modeling module, a real-time environmental data acquisition module, a path deviation calculation module, a path deviation index analysis module, and a path generation and dynamic update module. The environmental sensing module is used to collect environmental data inside the cable tunnel and upload the raw data to the target environmental database; The target environment database is used to store the raw environmental data collected inside the cable tunnel; The path planning module is used to preprocess the raw data and extract environmental parameters of the same area, and calculate the real-time path feasibility index of each node. The map modeling module is used to build a 3D topological map of the tunnel, marking the location of obstacles and passable areas; The real-time environmental data acquisition module is used to collect environmental data of the robot's current location in real time. The path deviation calculation module is used to calculate the deviation threshold range for each path node. The path deviation index analysis module is used to calculate the path deviation index, which measures the degree of deviation between the current environment and the theoretical model. The path generation and dynamic update module is used for dynamic path planning and path adjustment.

[0013] As attached Figure 2 The cable tunnel robot path generation method based on environmental perception and dynamic programming shown herein includes the following steps in its specific implementation: S1: The environmental sensing module collects environmental data inside the cable tunnel and uploads the raw data to the target environmental database.

[0014] In this embodiment, the cable tunnel raw data collected by the environment perception module is obtained by multiple sensors: the AGV obstacle avoidance laser radar measures the target distance by emitting laser pulses, generates three-dimensional point cloud data to accurately identify the obstacle profile, tunnel structure size and spatial coordinates; the RGBD camera uses structured light technology to obtain environmental depth information and synchronously generates three-dimensional point cloud data; the ultrasonic sensor measures the distance of objects by emitting and receiving ultrasonic reflection signals; the IMU unit uses an accelerometer, a gyroscope and a magnetometer to monitor the attitude and motion trajectory of the robot in real time; the Ousile R8 mobile SLAM color three-dimensional laser scanner generates high-density three-dimensional point cloud data through multi-angle laser scanning; the raw data collected by various sensors is not processed and is directly transmitted to the target environment database to provide basic data support for subsequent three-dimensional topological map construction and path planning.

[0015] S2: The path planning module extracts the environment parameters of the same region from the target environment database according to the region range and transmits them to the map modeling module. In this embodiment, it needs to be specifically pointed out that "the same region" means that when the path planning module extracts the environment parameters, it will set the space division standard according to the physical structure or functional attribute of the tunnel, filter out the environment data with completely coinciding geographical range, the same functional attribute or consistent logical partition from the database, which ensures that the environment parameters transmitted to the map modeling module only correspond to the actual environment of the target region, avoids mixing data from adjacent intervals or different functional areas, and thus guarantees the spatial accuracy and data consistency of three-dimensional map modeling.

[0016] It needs to be further pointed out that the environment parameters include multi-dimensional information such as tunnel space structure, obstacle distribution, passable region boundary and robot pose, specifically including geometric dimensions such as the shape, width, height and radian of the tunnel cross section; three-dimensional coordinates, types and sizes of obstacles; horizontal and vertical safety boundaries, path node coordinates of passable regions; filtered point cloud data of laser radar, semantic features extracted by RGBD camera, and real-time robot pose calculated by IMU unit.

[0017] It needs to be explained that the path planning module will first preprocess the original data in the target environment database when extracting the environment parameters of the same area in the target environment database. Specifically, it first receives binary point cloud data packets, image matrices and other multi-source data transmitted by laser radar, RGBD camera devices through TCP / IP, USB interface, and parses three-dimensional coordinates, reflection intensity, timestamps and other information according to the exclusive protocol, and encapsulates them into a standardized data structure containing laser radar point cloud, RGBD data, ultrasonic distance and IMU attitude; In the data cleaning link, the laser radar point cloud is filtered by statistical outlier filtering to remove flying points, and the data amount is reduced by voxel grid downsampling, and the median filtering is applied to the depth map and the sliding window average filtering is applied to the ultrasonic signal, and then the IMU data is fused by complementary filtering to suppress attitude drift; In the coordinate unification process, the database module will determine the external parameter matrix of each sensor relative to the robot coordinate system through hand-eye calibration, convert the original coordinates to the robot coordinate system by using rotation matrix and translation vector, and based on the global clock synchronization mechanism, linear interpolation or nearest neighbor matching is performed on the sensor data of different frequencies based on the IMU time as the reference, to provide accurate data support for subsequent path planning.

[0018] S3: The map modeling module constructs a three-dimensional topological map of the tunnel according to the extracted environment parameters, and marks the obstacle positions, passable regions and path nodes.

[0019] In this embodiment, it needs to be specifically pointed out that, as shown in the accompanying drawings, Figure 3 The steps of constructing the three-dimensional topological map are as follows: A1: Spatial voxelization and obstacle marking; It needs to be explained that the map modeling module divides the target area into equidistant three-dimensional voxels based on the tunnel cross-sectional geometric dimensions and the mileage range, and at the same time, the input obstacle coordinate point set , the passable region boundary point set are standardized in coordinates, and the formula is used, where W, H and L are the width, height and length of the tunnel cross-sectional geometric dimensions, and xmin, ymin, zmin and xmax, ymax and zmax are the original coordinate extreme values, which are used to map the original coordinates to a unified coordinate system; Then the voxel state function S(vi) is used to mark the voxel attributes: , If there exists oj e O, whose standardized coordinates satisfy xj' e [xv, xv + v], yj' e [yv, yv + v] and zj' e [zv, zv + v], then S(vi) = 1; otherwise, S(vi) = 0, and finally, the invalid voxels outside the tunnel wall are removed, and only the internal voxels are retained.

[0020] A2: Path node screening and generation; It needs to be explained that the key nodes are extracted from the passable voxels, and the formula is: , ni is the candidate node, (xv, yv, zv) is the voxel starting coordinate, and the candidate node is down-sampled by octree, and the tunnel branch opening, turning vertex and 1-meter interval of straight line segment are preferentially selected as the key nodes , and the orientation constraint along the tunnel extension direction is added to the node to ensure the continuity of the robot travel direction.

[0021] A3: Node connection relationship construction; It needs to be explained that the graph theory model is adopted in this embodiment to define the node connection rule: the straight line distance ≤1.2 meters is the adjacent node, the connectivity is judged by ray detection, the connection is established when there is no obstacle shielding, and the edge weight is calculated according to the Euclidean distance and the environmental complexity; A4: Topological map optimization and storage; It needs to be explained that the Delaunay triangulation algorithm is used in this embodiment to delete redundant nodes and retain key points that can represent the structural characteristics of the tunnel; the optimized node list, edge connection relationship and obstacle attribute are stored in JSON format, including coordinates, passing constraints and other parameters, which are called by the path planning module to complete the construction of the three-dimensional topological map.

[0022] It needs to be further explained that the Delaunay triangulation is an algorithm for constructing a triangular mesh from a set of discrete points, which ensures that any triangle does not contain other points within its circumcircle in all possible triangulations of the given point set; this feature makes the triangulation result have the advantages of maximizing the minimum angle and avoiding long and narrow triangles, and is widely used in the fields of geometric modeling and mesh generation.

[0023] S4: Real-time environment data acquisition module acquires the environment data of the current position of the robot in real time, and transmits it to the path planning module to calculate the real-time path feasibility index of each path node in the current environment.

[0024] In this embodiment, it needs to be specifically explained that the real-time path feasibility index is calculated by a multi-factor weighted summation model, first, the obstacle detection data of the laser radar, RGBD camera and ultrasonic sensor are uniformly converted into distance values da, and the passing probability Pp is defined: , m is the number of sensor types, β is the sensitivity parameter, and ds is the safety distance threshold; secondly, the collision risk index Rc is calculated based on the speed and distance: , vo is the robot speed, do is the distance of the nearest obstacle, a is the speed weight coefficient, and occ is the obstacle moving probability, which is 0 when it is a static obstacle and 1 when it is a dynamic obstacle; finally, the implementation path feasibility index If is calculated: , w1+w2=1, w1 is the probability weight, w2 is the collision risk weight, and under the conventional value, w1>w2.

[0025] It needs to be explained that the multi-factor weighted summation model realizes the quantitative evaluation of the path node passing feasibility by fusing multi-sensor data and dynamic risk parameters. The passing probability Pp adopts a logistic regression model to convert the deviation of the distance value da detected by each sensor from the safety distance threshold ds into a probability value. When the obstacle distance is less than ds, Pp decreases exponentially with the decrease of the distance, ensuring that the near-distance obstacle significantly inhibits the passing ability. The collision risk index Rc is coupled with the robot speed vo and the dynamic attribute of the obstacle: the Rc of the static obstacle occ=0 is always 0, and the dynamic obstacle linearly increases with the relative speed and distance ratio, reflecting the urgency of dynamic obstacle avoidance.

[0026] It needs to be further explained that the ratio of the weight coefficients w1 and w2 follows the safety first principle to limit the value domain of If to [-1, 1]: when If>0.5, it is determined as a high feasible node, 0.2<If≤0.5 as a medium feasible node, and If≥0.2 needs to be bypassed.

[0027] S5: The path deviation calculation module compares the real-time path feasibility index with the theoretical path benchmark to calculate the deviation threshold range of each path node.

[0028] In this embodiment, it needs to be specifically explained that the specific steps of calculating the deviation threshold range of each path node are as follows: B1: Based on the static planning result of the three-dimensional topological map, the node theoretical feasibility is calculated. , k is the number of paths passing through the node in the historical planning; It needs to be explained that the node theoretical feasibility It is obtained based on the historical planning data statistics of the three-dimensional topological map, and is used to reflect the benchmark passing capacity of the node in the static environment. Specifically, the path deviation calculation module will select all paths passing through the target node from the historical planning records to form a path set, and then calculate and average the real-time feasibility index of each path passing through the node. Here, k refers to the number of effective paths passing through the node in the historical planning, and the generation time of these paths needs to be within 1 hour from the current time to avoid statistical deviation caused by environmental structure changes. The essence of It calculation is the prior probability quantization of node passing capacity: if It is close to 1, it means that the node has maintained high feasibility in the historical planning, such as being in a spacious straight road area; if It is less than 0.5, it means that the node has structural passing risk, such as being near a narrow passage or a fixed obstacle, and such nodes need to increase the safety margin during dynamic planning; B2: Calculate the deviation threshold range [low, high]: , is the preset relative deviation coefficient, is the historical feasibility standard deviation, reflecting environmental uncertainty, is the two-sided quantile of the standard normal distribution.

[0029] It needs to be explained that the calculation of the deviation threshold range [low, high] is based on the node theory feasibility, environmental uncertainty and the preset safety coefficient, which is used to define the normal fluctuation interval of path feasibility. First, take the node theory feasibility It as the benchmark, and adjust It by the preset relative deviation coefficient δ to get the basic fluctuation range. This part of the calculation reflects the fluctuation degree of the node feasibility under normal environment. For example, if the It of a node is 0.8 and δ is 0.2, the basic interval is [0.64, 0.96]. On this basis, is used to compensate for the random fluctuations of the environment, where the historical feasibility standard deviation σ measures the fluctuation intensity of the node's past traffic state. If the node is close to the dynamic operation area or there are temporary obstacles frequently, its σ value will be relatively large; and the two-sided quantile of the standard normal distribution is determined according to the confidence level α, which determines the strictness of the threshold boundary, ensuring that the actual feasibility falls within the set range with a high probability; The lower limit low and the upper limit high calculated have clear decision-making significance: when the real-time feasibility If is lower than low, it means that the current environmental risk is significantly higher than the historical benchmark, and if sudden obstacles or narrow passages appear, the system will trigger global path re-planning; if If is higher than high, it means that the environmental conditions are better than expected, for example, obstacles have been removed, and the system will preferentially explore this high feasibility area to optimize the path; when If is within the interval [low, high], the system only makes local fine-tuning to the path, avoiding frequent re-planning due to minor fluctuations.

[0030] S6: The path deviation index analysis module analyzes the path deviation index based on the deviation threshold range of multiple path nodes, which measures the deviation degree of the current environment from the theoretical model.

[0031] In this embodiment, the deviation index measures the overall deviation of the environment from the theoretical model, which is calculated using the weighted sum of squares method. The specific calculation process is as follows: C1: Calculate the single-node deviation value Δi: , When If(ni) is within the threshold range, Δi=0.

[0032] It should be explained that the calculation of the single-node deviation value Δi aims to quantify the degree of deviation between the real-time feasibility of a single path node and the theoretical threshold range. Specifically, this calculation handles two different cases through a piecewise function: when the real-time feasibility index If(ni) is lower than the corresponding node's threshold lower limit lowi, Δi is calculated using the formula... The calculation shows that the numerator represents the absolute deviation between the current feasibility and the lower limit, while the denominator is the total width of the threshold interval for that node. Dividing the two yields the relative deviation ratio. When If(ni) is higher than the upper threshold highi, Δi is used. The calculation logic is the same: divide the absolute deviation by the interval width to measure the relative degree of deviation. When If(ni) is within the threshold range of [lowi, highi], it indicates that the real-time state of the node conforms to the historical benchmark and expected fluctuations. At this time, Δi is directly assigned a value of 0, representing no significant deviation. This calculation method considers both the direction of deviation and eliminates the influence of the difference in threshold intervals between different nodes through normalization, making the deviation values ​​of each node horizontally comparable and providing a standardized basis for the subsequent calculation of the overall deviation index.

[0033] C2: Calculate the path deviation index Ed: , n is the number of nodes in the current path, wi is the node weight, and γ is the environmental sensitivity coefficient.

[0034] It's important to explain that the path deviation index Ed is a comprehensive assessment of the deviations of all nodes along the entire path, using a weighted sum of squares to highlight the impact of key nodes and smooth out local fluctuations. Specifically, in the formula... First, set the deviation value of each node. Multiply by the corresponding weight wi, where wi reflects the importance of the node in the path; squaring the multiplication amplifies the difference in deviation values, preventing small deviations from masking the result; then summing the weighted squared deviation values ​​of all nodes yields the cumulative deviation of the entire path. Then, through... The average value is processed to eliminate the interference of different numbers of path nodes on the results, making the deviation index of paths of different lengths comparable. Finally, it is multiplied by the environmental sensitivity coefficient γ. The larger the γ is, the more sensitive the system is to environmental changes. It is used to adjust the sensitivity of the deviation index according to the actual application scenario. The larger the final Ed value, the higher the overall deviation between the actual environment of the current path and the theoretical model. The system uses this to determine whether to trigger path replanning or adjust the strategy.

[0035] S7: The path generation and dynamic update module uses the path deviation index. If the deviation exceeds the threshold, it triggers the dynamic programming algorithm to recalculate the optimal path; if the deviation is small, it makes local adjustments based on the theoretical path.

[0036] In this embodiment, when the deviation index exceeds the threshold value, i.e. , Eth is a preset threshold value, the improved Dlite algorithm is used, and the path weight is adjusted in combination with the real-time feasibility index: , g(n) is the actual cost from the starting point to node n, h(n) is the heuristic cost of node n to the end point, is a deviation influence coefficient, the path search priority is dynamically adjusted, the nodes with small deviation influence and high feasibility are preferentially explored, a new optimal path bypassing the deviation area and fitting the real-time environment is quickly generated, when the deviation is small, the edge weight of the adjacent node is fine-tuned on the basis of the theoretical path according to the real-time feasibility index, the path is smoothly adapted to the slight change of the environment through local path reconnection, and the stability and efficiency of the robot are ensured.

[0037] Secondly: the drawings in the disclosed embodiments of the application only involve the structures involved in the disclosed embodiments, other structures can refer to the usual design, and in the case of no conflict, the same embodiment and different embodiments of the application can be combined with each other; Finally: the above only describes the preferred embodiments of the application and is not used to limit the application, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application shall be included in the protection scope of the application.

Claims

1. A method for generating a path for a cable tunnel robot based on environmental perception and dynamic programming, characterized in that, It includes an environmental perception module, a target environment database, a path planning module, a map modeling module, a real-time environmental data acquisition module, a path deviation calculation module, a path deviation index analysis module, and a path generation and dynamic update module, and includes the following steps: S1: The environmental sensing module collects environmental data inside the cable tunnel and uploads the raw data to the target environment database; S2: The path planning module extracts environmental parameters of the same region from the target environment database according to the regional scope and transmits them to the map modeling module; S3: The map modeling module constructs a 3D topological map of the tunnel based on the extracted environmental parameters, marking the locations of obstacles, passable areas, and path nodes; S4: The real-time environmental data acquisition module collects environmental data of the robot's current position in real time and transmits it to the path planning module to calculate the real-time path feasibility index of each path node in the current environment. S5: The path deviation calculation module compares the real-time path feasibility index with the theoretical path benchmark and calculates the deviation threshold range for each path node. S6: The path deviation index analysis module analyzes the deviation threshold range of multiple path nodes to obtain the path deviation index, which measures the degree of deviation between the current environment and the theoretical model. S7: The path generation and dynamic update module uses the path deviation index. If the deviation exceeds the threshold, it triggers the dynamic programming algorithm to recalculate the optimal path; if the deviation is small, it makes local adjustments based on the theoretical path.

2. The cable tunnel robot path generation method based on environmental perception and dynamic programming according to claim 1, characterized in that: The environmental parameters include the tunnel spatial structure, obstacle distribution, passable area boundaries, and robot pose. Specifically, they include the geometric dimensions of the tunnel cross-section, such as shape, width, height, and curvature; the three-dimensional coordinates, type, and size of obstacles; the horizontal and vertical safety boundaries of the passable area and the coordinates of path nodes; the point cloud data filtered by the lidar, the semantic features extracted by the RGBD camera, and the real-time robot pose calculated by the IMU unit.

3. The cable tunnel robot path generation method based on environmental perception and dynamic programming according to claim 1, characterized in that: The steps for constructing a 3D topological map are as follows: A1: Spatial Voxelization and Obstacle Marking; A2: Path node filtering and generation; A3: Establishing node connection relationships; A4: Topology map optimization and storage.

4. The cable tunnel robot path generation method based on environmental perception and dynamic programming according to claim 3, characterized in that: The aforementioned spatial voxelization and obstacle marking refer to the map modeling module dividing the target area into isometric three-dimensional voxels based on the tunnel cross-section geometry and mileage range, while simultaneously marking the input obstacle coordinate point set. Set of boundary points of passable area To standardize coordinates, use the formula W, H, and L are the width, height, and length of the tunnel cross-section, respectively, while xmin, ymin, zmin, xmax, ymax, and zmax are the original coordinate extreme values ​​used to map the original coordinates to a unified coordinate system. Next, the voxel attributes are marked using the voxel state function S(vi): ， If there exists oj∈O, whose standardized coordinates satisfy xj'∈[xv,xv+v], yj'∈[yv,yv+v], zj'∈[zv,zv+v], then S(vi)=1; otherwise, S(vi)=0. Finally, invalid voxels outside the tunnel wall are removed, and only internal voxels are retained.

5. The cable tunnel robot path generation method based on environmental perception and dynamic programming according to claim 3, characterized in that: The model used for path node filtering is: ， (xv,yv,zv) are the initial coordinates of the voxel, and ni is the candidate node.

6. The cable tunnel robot path generation method based on environmental perception and dynamic programming according to claim 1, characterized in that: The real-time path feasibility index is calculated using a multi-factor weighted summation model, and the specific steps are as follows: a1: Calculate the probability of passage Pp: ， Da is the distance value that is uniformly converted from obstacle detection data from LiDAR, RGBD camera, and ultrasonic sensor; m is the number of sensor types; β is the sensitivity parameter; and ds is the safe distance threshold. a2: Calculate the collision risk index Rc: ， vo is the robot speed, do is the distance to the nearest obstacle, a is the speed weighting coefficient, and occ is the probability of obstacle movement. a3: Calculate the feasibility index of the implementation path. If: ， w1+w2=1, where w1 is the passability probability weight and w2 is the collision risk weight.

7. The method for generating a path for a cable tunnel robot based on environmental perception and dynamic programming according to claim 1, characterized in that: The specific steps for calculating the deviation threshold range for each path node are as follows: B1: Calculate the theoretical feasibility of nodes based on the static planning results of the 3D topology map; ， k is the number of paths that pass through this node in the historical planning; B2: Calculate the deviation threshold range [low, high]: ， To preset the relative deviation coefficient, The historical feasibility standard deviation reflects environmental uncertainty. The two-sided quantiles of the standard normal distribution.

8. The method for generating a cable tunnel robot path based on environmental perception and dynamic programming according to claim 1, characterized in that: The deviation index measures the overall degree of deviation between the environment and the theoretical model, and is calculated using the weighted sum of squares method. The specific calculation process is as follows: C1: Calculate the single-node deviation value Δi: ， Δi = 0 when If(ni) is within the threshold range; C2: Calculate the path deviation index Ed: ， n is the number of nodes in the current path, wi is the node weight, and γ is the environmental sensitivity coefficient.

9. The method for generating a path for a cable tunnel robot based on environmental perception and dynamic programming according to claim 1, characterized in that: When the deviation index exceeds the threshold, the path generation and dynamic update module adjusts the path weights in conjunction with the real-time feasibility index. Let g(n) be the actual cost from the starting point to node n, and h(n) be the heuristic cost from node n to the destination. This is the deviation influence coefficient.

Images