Motion planning method of coal mine underground roadway inspection robot
By constructing a grid map and using Harris corner detection and an improved A* algorithm to generate global paths, combined with chord length parameterization and TEB algorithm optimization, the problem of uneven paths in underground coal mine roadways was solved, improving the robot's motion performance and environmental adaptability.
Patent Information
- Application Number
- CN202610016813.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-07
- Publication Date
- 2026-04-24
AI Technical Summary
In the context of underground coal mine roadways, existing technologies often result in unsmooth paths generated by traditional path planning methods, high energy consumption, and limited performance of local optimization algorithms. This leads to discontinuous robot movement and planning failures, making it difficult for robots to move autonomously and safely in complex environments.
By constructing a grid map using LiDAR, identifying obstacle outlines using Harris corner detection, generating a global path using an improved A* algorithm, and performing local optimization using chord-length parameterized cubic B-spline interpolation and the TEB algorithm that integrates potential field gravitational constraints, a smooth and safe trajectory is generated.
It enables the generation of smooth and efficient global paths in underground coal mine roadways, and improves the robot's motion performance and environmental adaptability through local optimization, ensuring the robot's safe movement in complex environments.
Smart Images

Figure CN121918562A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of robot path planning technology, specifically a motion planning method for a coal mine underground roadway inspection robot. Background Technology
[0002] The safety of underground coal mine roadways is directly related to the stable operation of the entire production system. Traditional manual inspection methods are not only inefficient but also expose personnel to dangerous environments. With the development of robotics technology and the coal mining industry's move towards intelligent construction, inspection robots will gradually replace manual labor. However, autonomous movement of robots in complex roadway environments remains a technical challenge that urgently needs to be overcome.
[0003] Currently, path planning methods based on the traditional A* algorithm face severe challenges in tunnel environments. The generated paths often contain numerous unnecessary turns, leading to unsmooth robot motion, increased energy consumption, and even failure to execute, severely restricting the performance of subsequent local optimization algorithms. This causes the robot to frequently get stuck in local optima or fail to plan, affecting the continuity and reliability of inspection tasks. At the local planning level, while the Time Elastic Band (TEB) algorithm can handle the robot's kinematic constraints and generate smooth trajectories, it also has significant shortcomings in the special scenarios of coal mine tunnels. The traditional TEB algorithm mainly relies on the repulsive force field generated by obstacles to maintain a safe distance. In narrow tunnels, the strong repulsive force of obstacles on both sides can easily cause the robot to oscillate in the middle area, or plan an excessively large detour radius to avoid obstacles.
[0004] The corner points of an obstacle's outline are crucial information reflecting its structural characteristics. Existing methods fail to fully utilize this environmental feature, resulting in a lack of foresight and intelligence in the planning process. How to generate a smooth and efficient global path while ensuring safety, and provide a high-quality reference for the local planner, has become a core issue in improving the overall performance of inspection robots. Summary of the Invention
[0005] The purpose of this invention is to provide a motion planning method for a coal mine underground roadway inspection robot to solve the problems in the prior art.
[0006] The objective of this invention can be achieved through the following technical solutions: A motion planning method for a coal mine underground roadway inspection robot includes the following steps: Step 1: Collect information using the lidar mounted on the robot to construct a two-dimensional grid map with the obstacle locations marked; Step 2: For the raster map obtained in Step 1, use the Harris corner detection algorithm to extract the corner points of the obstacle outline, identify the bypass nodes around the corner points, and construct a list C_List; Step 3: Set the planning start point S and planning end point G on the map, and run the improved A* algorithm to generate a global path. The path points are denoted as Qp={q0, q1, ..., qn}. Step 4: Add time series to the path points obtained in Step 3 using the chord length parameterization method, and use cubic B-splines to interpolate the time-x-axis and time-y-axis of the path points to generate an initial reference trajectory with continuity. Step 5: Extract pose points from the trajectory obtained in Step 4 at fixed time intervals, and use them as global references for the TEB algorithm with fused potential field gravitational constraints for local optimization to generate local trajectories. If the local trajectory is invalid, return to Step 3 to replan, and finally generate a safe trajectory for the robot and control the robot's movement.
[0007] Furthermore, in step one, environmental information of underground coal mine roadways is collected by a lidar mounted on the robot, a two-dimensional grid map is constructed using the grid method, and obstacle grids are marked on the map according to the actual location of the obstacles. The obstacle grid is defined as an impassable area, which is prohibited from being expanded in path planning. The remaining grids are defined as free grids, which are passable areas and are allowed to be expanded in path planning.
[0008] Furthermore, the method for determining the bypass node in step two is as follows: taking each identified corner point as the center, traversing its 2*2 grid neighborhood; if there is only one obstacle grid in the neighborhood, then selecting the free grid in the 2*2 neighborhood that is diagonally related to the obstacle grid as the bypass node. There is only one obstacle grid in the four grids centered on the red corner, so there is a detour node indicated by yellow. However, there are multiple obstacle grids in the four grids centered on the blue corner, so there is no detour node.
[0009] Furthermore, the improved A* algorithm in step three uses a target-point-oriented visual bypass node expansion strategy to expand child nodes.
[0010] Furthermore, the bypass node is obtained in step two. Visibility means that if the straight path connecting two free grids does not pass through any obstacle grids, then the two grids are considered to be visible to each other; otherwise, they are not visible. The target-point-oriented visual detour node expansion strategy includes the following steps: S1: Initialize the planned endpoint G to the current target point M; S2: Determine whether the parent node C and the current target point M are visible. If they are visible, directly make M the child node of the parent node C; otherwise, execute S3. S3: Locate the first obstacle grid encountered on the line connecting parent node C and M, retrieve all obstacle grids connected to this obstacle grid, and query the corresponding bypass node in the list C_List to form a candidate set H, and then execute S4; S4: Traverse the candidate set H and select all bypass nodes that are visible to the parent node C as children of C. If there are no bypass nodes visible to the parent node C in the set H, then execute S5. S5: Select the bypass node with the closest Euclidean distance to the parent node C from the candidate set H as the new temporary target point Mnew; if Mnew and M are not the same node, then return to S2 for iteration as the target point; otherwise, determine that the parent node C has no child nodes.
[0011] Furthermore, the specific steps of the improved A* algorithm in step three are as follows: S1: Create lists OpenList and CloseList, and put the planning starting point into OpenList; S2: Determine if OpenList is empty; if OpenList is not empty, select the node with the smallest total cost F(n) as the parent node C, move node C from OpenList to CloseList, and execute S3; if OpenList is empty, determine that the planning has failed. S3: Determine if the parent node C is the planning endpoint. If so, start from the endpoint G and backtrack the parent node to the starting point S to generate a global path and end the planning; otherwise, execute S4. S4: Using the target point-oriented line-of-sight reachable detour node expansion strategy, obtain all child nodes of parent node C, calculate the total cost of each child node, add it to OpenList or update the cost of existing nodes in OpenList, and then return to S2; The formula for calculating the total cost function is: F(n) = G(n) + H(n) (1) In the formula, F(n) represents the total cost of reaching the destination from the starting point through node n, which includes two parts: G(n) and H(n). G(n) represents the actual cost of reaching node n from the starting point; H(n) represents the estimated cost of reaching the destination from node n. The heuristic function H(n) is represented using Euclidean distance, and its calculation formula is as follows: In the formula, (xn, yn) are the coordinates of the extended node n, and (xg, yg) are the coordinates of the endpoint.
[0012] Furthermore, step four includes: The mathematical expression for a cubic B-spline curve is:
[0013] In the formula, k is the degree of the B-spline curve, and the value of k for a cubic B-spline curve is 3; i is the control point number. di is a control point, and there are n+1 control points; t is a node vector, with a number of m+1 nodes satisfying m=n+k+1, and its physical meaning is the time corresponding to passing through a path point; Ni,k(t) is the basis function of the i-th k-th B-spline curve.
[0014] 8. The motion planning method for a coal mine underground roadway inspection robot according to claim 7, characterized in that the node vector is determined according to the following steps: S1: Based on the path point coordinate sequence Qp={q0, q1, ..., qn} generated by the improved A* algorithm, calculate the total chord length of the path as L:
[0015] In the formula, Qp − Qp−1 represents the Euclidean distance between adjacent path points; Calculate the chord length parameters based on the total chord length L:
[0016] S2: Set the first k+1 nodes of the m+1 node vector to 0 and the last k+1 nodes to T. By setting repeated nodes, ensure that the trajectory curve passes through the first and last path points. The relationship between the remaining internal nodes and the chord length parameter is as follows: .
[0017] Furthermore, the specific steps of the TEB algorithm for fusing potential field gravitational constraints in step five are as follows: S1: Define the intercepted trajectory sequence B as the set of variables to be optimized: B = (Q, T) (7) In the formula, B represents the trajectory to be optimized, and Q and T are a series of pose sequences and time interval sequences, respectively; S2: Construct a multi-objective optimization function F(B), and solve for the optimal trajectory that minimizes F(B) using the g2o graph optimization framework. This function consists of multiple weighted sub-objective functions:
[0018] In the formula, ftime is used to minimize the running time; fobs is the obstacle repulsion cost, which is a penalty cost incurred when the distance between the robot and the obstacle is less than the set minimum safe distance; fvel and facc are kinematic constraint costs, which are penalty costs incurred when the robot's speed or acceleration exceeds the limits. fcorner represents the corner gravity cost. The Euclidean distance between the robot's current pose and all bypass nodes in the list C_List is calculated in real time. The nearest bypass node is selected as a local gravity source. The goal is to minimize the distance between the robot's pose and the gravity source, so that the generated optimized trajectory actively approaches the bypass node. γt, γo, γv, γa, and γc are the weighting coefficients for time, obstacle, velocity, acceleration, and corner gravity constraints, respectively, used to adjust the influence of the sub-cost function.
[0019] The beneficial effects of this invention are: 1. This invention first extracts environmental feature points through Harris corner detection to obtain detour nodes. Then, it improves the A* algorithm based on a target-point-guided visual detour node expansion strategy to enhance global path quality. Next, it uses cubic B-spline curves for smooth trajectory generation. Finally, it achieves local optimization and dynamic replanning through an improved TEB algorithm incorporating potential field gravitational constraints. This method not only fully considers the environmental characteristics of roadway obstacles but also establishes effective synergy between global planning and local optimization, enabling the robot to exhibit superior motion performance and environmental adaptability in complex roadway environments, and better applying it to intelligent inspection in underground coal mines. 2. This invention defines path key points by determining the turning points of obstacles in the environment, and uses these key points to plan paths around obstacles and participate in local trajectory optimization. In a grid map, the grayscale difference between obstacles and surrounding free grid cells is significant, and the turning points precisely match the concept of corner points in computer vision. Therefore, the widely used Harris corner detection method is used for identification. Corner points are located at the intersections of grid lines, at the sharp corners of obstacle grid cells, and are not suitable for direct participation in planning. They need to be offset to surrounding safe free grid cells according to predetermined rules. Furthermore, only grid nodes located at the outward protrusions of obstacles are key to bypassing obstacles in path planning. Therefore, rules were formulated to filter out grid nodes that meet the criteria. Attached Figure Description
[0020] The invention will now be further described with reference to the accompanying drawings.
[0021] Figure 1 This is a flowchart of the motion planning method for coal mine underground roadway inspection robots; Figure 2 This is a schematic diagram of a raster map; Figure 3 This is a schematic diagram of Harris corner detection and recognition. Figure 4 This is a schematic diagram of the detour node selection; Figure 5 This is the flowchart of the improved A* algorithm. Detailed Implementation
[0022] 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.
[0023] A motion planning method for a coal mine underground roadway inspection robot, such as Figures 1-5 As shown, it includes the following steps: (1) Collect environmental information of underground roadways in coal mines by using a lidar mounted on a robot, construct a two-dimensional grid map using the grid method, and mark obstacle grids on the map according to the actual location of obstacles; like Figure 2 As shown in the black grid, the obstacle grid is defined as a trespassable area, which cannot be expanded in path planning; The remaining grid cells are defined as free grid cells, which are passable areas and allow for node expansion during path planning; (2) For the raster map obtained in step (1), the Harris corner detection algorithm is used to extract the corner points of the obstacle contours, such as... Figure 3 The green cross markers shown indicate the location of the corner points, and the list C_List is constructed by identifying the surrounding nodes. The method for determining the detour node is as follows: taking each identified corner point as the center, traverse the 2*2 grid neighborhood where it is located. If there is only one obstacle grid in the neighborhood, select the free grid in the 2*2 neighborhood that is diagonally related to the obstacle grid as the detour node. like Figure 4 As shown, there is only one obstacle grid in the four grids centered on the red corner, so there is a detour node indicated by yellow; while there are multiple obstacle grids in the four grids centered on the blue corner, so there is no detour node. (3) Set the planning start point S and planning end point G on the map, run the improved A* algorithm to generate a global path, and record the path points as Qp={q0, q1, ..., qn}; The improved A* algorithm uses a target-point-oriented visual detour node expansion strategy to expand child nodes. The bypass node is obtained in step (2). Visibility means that if the straight path connecting two free grids does not pass through any obstacle grids, then the two grids are considered to be visible to each other; otherwise, they are not visible. The target-point-oriented visual detour node expansion strategy includes the following steps: S1: Initialize the planned endpoint G to the current target point M; S2: Determine whether the parent node C and the current target point M are visible. If they are visible, directly make M the child node of the parent node C; otherwise, execute S3. S3: Locate the first obstacle grid encountered on the line connecting parent node C and M, retrieve all obstacle grids connected to this obstacle grid, and query the corresponding bypass node in the list C_List to form a candidate set H, and then execute S4; S4: Traverse the candidate set H and select all bypass nodes that are visible to the parent node C as children of C. If there are no bypass nodes visible to the parent node C in the set H, then execute S5. S5: Select the bypass node with the closest Euclidean distance to the parent node C from the candidate set H as the new temporary target point Mnew; if Mnew and M are not the same node, then return to S2 for iteration as the target point; otherwise, determine that the parent node C has no child nodes. The flowchart of the improved A* algorithm is as follows: Figure 5 As shown, the specific steps are as follows: S1: Create lists OpenList and CloseList, and put the planning starting point into OpenList; S2: Determine if OpenList is empty; if OpenList is not empty, select the node with the smallest total cost F(n) as the parent node C, move node C from OpenList to CloseList, and execute S3; if OpenList is empty, determine that the planning has failed. S3: Determine if the parent node C is the planning endpoint. If so, start from the endpoint G and backtrack the parent node to the starting point S to generate a global path and end the planning; otherwise, execute S4. S4: Using the target point-oriented line-of-sight reachable detour node expansion strategy, obtain all child nodes of parent node C, calculate the total cost of each child node, add it to OpenList or update the cost of existing nodes in OpenList, and then return to S2; The formula for calculating the total cost function is: F(n) = G(n) + H(n) #(1) In the formula, F(n) represents the total cost of reaching the destination from the starting point through node n, comprising two parts: G(n) and H(n). G(n) represents the actual cost of reaching node n from the starting point; H(n) represents the estimated cost of reaching the destination from node n. The heuristic function H(n) is represented using Euclidean distance, and its calculation formula is as follows: H(n) = (xn−xg)2 + (yn−yg)2#(2) In the formula, (xn, yn) are the coordinates of the extended node n, and (xg, yg) are the coordinates of the endpoint.
[0024] (4) Use the chord length parameterization method to add time series to the path points obtained in step (3), and use cubic B splines to interpolate the time-x-axis and time-y-axis of the path points respectively to generate an initial reference trajectory with continuity. The mathematical expression for the cubic B-spline curve is: F(t) = i = 0ndiNi,k(t)#(3) In the formula, k is the degree of the B-spline curve, and the value of k for a cubic B-spline curve is 3; i is the control point number. di is a control point, and there are n+1 control points; t is a node vector, with a number of m+1 nodes satisfying m=n+k+1, and its physical meaning is the time corresponding to passing through a path point; Ni,k(t) is the basis function of the i-th k-th B-spline curve; The node vector is determined according to the following steps: S1: Based on the path point coordinate sequence Qp={q0, q1, ..., qn} generated by the improved A* algorithm, calculate the total chord length of the path as L: L=p=1nQp − Qp−1 #(4) In the formula, Qp − Qp−1 represents the Euclidean distance between adjacent path points; Calculate the chord length parameters based on the total chord length L: tp=tp−1+Qp − Qp−1L #(5) S2: Set the first k+1 nodes of the m+1 node vector to 0 and the last k+1 nodes to T. By setting repeated nodes, ensure that the trajectory curve passes through the first and last path points. The relationship between the remaining internal nodes and the chord length parameter is as follows: ti+k = 1ki = jj+k−1tp, i=1,2,…,nk#(6) (5) Extract pose points from the trajectory obtained in step (4) at fixed time intervals, use them as global references for the TEB algorithm with fused potential field gravitational constraints for local optimization, generate local trajectories, trigger replanning when the local trajectory is invalid, return to step (3) for replanning, and finally generate a smooth and safe trajectory that satisfies the robot's kinematic constraints and control commands to drive the robot to move.
[0025] The specific implementation steps of the TEB algorithm for integrating potential field gravitational constraints are as follows: S1: Define the intercepted trajectory sequence B as the set of variables to be optimized: B = (Q, T) # (7) In the formula, B represents the trajectory to be optimized, and Q and T are a series of pose sequences and time interval sequences, respectively.
[0026] S2: Construct a multi-objective optimization function F(B), and solve for the optimal trajectory that minimizes F(B) using the g2o graph optimization framework. This function consists of multiple weighted sub-objective functions: F(B) = γtftime+γofobs+γvfvel+γafacc+γcfcorner (8) In the formula, ftime is used to minimize the running time; fobs is the obstacle repulsion cost, which is a penalty cost incurred when the distance between the robot and the obstacle is less than the set minimum safe distance; fvel and facc are kinematic constraint costs, which are penalty costs incurred when the robot's speed or acceleration exceeds the limits. fcorner represents the corner gravity cost. The Euclidean distance between the robot's current pose and all bypass nodes in the list C_List is calculated in real time. The nearest bypass node is selected as a local gravity source. The goal is to minimize the distance between the robot's pose and the gravity source, so that the generated optimized trajectory actively approaches the bypass node. γt, γo, γv, γa, and γc are the weighting coefficients for time, obstacle, velocity, acceleration, and corner gravity constraints, respectively, used to adjust the influence of the sub-cost function.
[0027] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.
Claims
1. A motion planning method for a coal mine underground roadway inspection robot, characterized in that, Includes the following steps: Step 1: Collect information using the lidar mounted on the robot to construct a two-dimensional grid map with the obstacle locations marked; Step 2: For the raster map obtained in Step 1, use the Harris corner detection algorithm to extract the corner points of the obstacle outline, identify the bypass nodes around the corner points, and construct a list C_List; Step 3: Set the planning start point S and planning end point G on the map, run the improved A* algorithm to generate a global path, and denote the path points as Q. p ={q0, q1, ..., qn}; Step 4: Add time series to the path points obtained in Step 3 using the chord length parameterization method, and use cubic B-splines to interpolate the time-x-axis and time-y-axis of the path points to generate an initial reference trajectory with continuity. Step 5: Extract pose points from the trajectory obtained in Step 4 at fixed time intervals, and use them as global references for the TEB algorithm with fused potential field gravitational constraints for local optimization to generate local trajectories. If the local trajectory is invalid, return to Step 3 to replan, and finally generate a safe trajectory for the robot and control the robot's movement.
2. The motion planning method for a coal mine underground roadway inspection robot according to claim 1, characterized in that, Step one involves collecting environmental information of underground coal mine roadways using a lidar mounted on a robot, constructing a two-dimensional grid map using a grid method, and marking obstacle grids on the map according to the actual location of obstacles. The obstacle grid is defined as an impassable area, which is prohibited from being expanded in path planning. The remaining grids are defined as free grids, which are passable areas and are allowed to be expanded in path planning.
3. The motion planning method for a coal mine underground roadway inspection robot according to claim 2, characterized in that, The method for determining the detour node in step two is as follows: take each identified corner point as the center, traverse the 2*2 grid neighborhood where it is located, and if there is only one obstacle grid in the neighborhood, select the free grid that is diagonally related to the obstacle grid in the 2*2 neighborhood as the detour node. There is only one obstacle grid in the four grids centered on the red corner, so there is a detour node indicated by yellow. However, there are multiple obstacle grids in the four grids centered on the blue corner, so there is no detour node.
4. The motion planning method for a coal mine underground roadway inspection robot according to claim 3, characterized in that, The improved A* algorithm in step three uses a target-point-oriented visual detour node expansion strategy to expand child nodes.
5. The motion planning method for a coal mine underground roadway inspection robot according to claim 4, characterized in that, The bypass node is obtained in step two. Visibility means that if the straight path connecting two free grids does not pass through any obstacle grids, then the two grids are considered to be visible to each other; otherwise, they are not visible. The target-point-oriented visual detour node expansion strategy includes the following steps: S1: Initialize the planned endpoint G to the current target point M; S2: Determine whether the parent node C and the current target point M are visible. If they are visible, directly make M the child node of the parent node C; otherwise, execute S3. S3: Locate the first obstacle grid encountered on the line connecting parent node C and M, retrieve all obstacle grids connected to this obstacle grid, and query the corresponding bypass node in the list C_List to form a candidate set H, and then execute S4; S4: Traverse the candidate set H and select all bypass nodes that are visible to the parent node C as children of C. If there are no bypass nodes visible to the parent node C in the set H, then execute S5. S5: Select the bypass node with the closest Euclidean distance to the parent node C from the candidate set H as the new temporary target point Mnew; if Mnew and M are not the same node, then return to S2 for iteration as the target point; otherwise, determine that the parent node C has no child nodes.
6. The motion planning method for a coal mine underground roadway inspection robot according to claim 4, characterized in that, The specific steps of the improved A* algorithm in step three are as follows: S1: Create lists OpenList and CloseList, and put the planning starting point into OpenList; S2: Determine if OpenList is empty; if OpenList is not empty, select the node with the smallest total cost F(n) as the parent node C, move node C from OpenList to CloseList, and execute S3; if OpenList is empty, determine that the planning has failed. S3: Determine if the parent node C is the planning endpoint. If so, start from the endpoint G and backtrack the parent node to the starting point S to generate a global path and end the planning; otherwise, execute S4. S4: Using the target point-oriented line-of-sight reachable detour node expansion strategy, obtain all child nodes of parent node C, calculate the total cost of each child node, add it to OpenList or update the cost of existing nodes in OpenList, and then return to S2; The formula for calculating the total cost function is: F(n) = G(n) + H(n) In the formula, F(n) represents the total cost of reaching the destination from the starting point through node n, which includes two parts: G(n) and H(n). G(n) represents the actual cost of reaching node n from the starting point; H(n) represents the estimated cost of reaching the destination from node n. The heuristic function H(n) is represented using Euclidean distance, and its calculation formula is as follows: In the formula, (xn, yn) are the coordinates of the extended node n, and (xg, yg) are the coordinates of the endpoint.
7. The motion planning method for a coal mine underground roadway inspection robot according to claim 6, characterized in that, Step four includes: The mathematical expression for a cubic B-spline curve is: In the formula, k is the degree of the B-spline curve, and the value of k for a cubic B-spline curve is 3; i is the control point number. di is a control point, and there are n+1 control points; t is a node vector, with a number of m+1 nodes satisfying m=n+k+1, and its physical meaning is the time corresponding to passing through a path point; Ni,k(t) is the basis function of the i-th k-th B-spline curve.
8. The motion planning method for a coal mine underground roadway inspection robot according to claim 7, characterized in that, in, The node vector is determined according to the following steps: S1: Based on the path point coordinate sequence Qp={q0, q1, ..., qn} generated by the improved A* algorithm, calculate the total chord length of the path as L: In the formula, Qp − Qp−1 represents the Euclidean distance between adjacent path points; Calculate the chord length parameters based on the total chord length L: S2: Set the first k+1 nodes of the m+1 node vector to 0 and the last k+1 nodes to T. By setting repeated nodes, ensure that the trajectory curve passes through the first and last path points. The relationship between the remaining internal nodes and the chord length parameter is as follows: 。 9. The motion planning method for a coal mine underground roadway inspection robot according to claim 8, characterized in that, The specific steps of the TEB algorithm for fusing potential field gravitational constraints in step five are as follows: S1: Define the intercepted trajectory sequence B as the set of variables to be optimized: B = (Q, T) (7) In the formula, B represents the trajectory to be optimized, and Q and T are a series of pose sequences and time interval sequences, respectively; S2: Construct a multi-objective optimization function F(B), and solve for the optimal trajectory that minimizes F(B) using the g2o graph optimization framework. This function consists of multiple weighted sub-objective functions: F(B) = γtftime+γofobs+γvfvel+γafacc+γcfcorner (8) In the formula, ftime is used to minimize the running time; fobs is the obstacle repulsion cost, which is a penalty cost incurred when the distance between the robot and the obstacle is less than the set minimum safe distance; fvel and facc are kinematic constraint costs, which incur penalty costs when the robot's speed or acceleration exceeds the limits. fcorner represents the corner gravity cost. The Euclidean distance between the robot's current pose and all bypass nodes in the list C_List is calculated in real time. The nearest bypass node is selected as the local gravity source. The goal is to minimize the distance between the robot's pose and the gravity source, so that the generated optimized trajectory actively approaches the bypass node. γt, γo, γv, γa, and γc are the weighting coefficients for time, obstacle, velocity, acceleration, and corner gravity constraints, respectively, used to adjust the influence of the sub-cost function.
