Automatic driving inspection method and system for fire-fighting robot

Abstract

The invention relates to the technical field of automatic driving inspection, in particular to an automatic driving inspection method and system for a fire-fighting robot, and the method comprises the following steps: obtaining a grid structure to count a connection number channel width and a connection angle mark complex section, extracting a passing path, calculating the overlapping proportion of an original path, and generating a to-be-reconstructed path segment; combining the shortest paths to generate an updated path section sequence, screening the number of intersections and the path length, collecting a fire source dense smoke coordinate, evaluating a heat source and diffusion change, predicting a shielding path section, judging that the path falls into the shielding section, reconstructing a connection path, calculating lengthy length and overlapping times, and generating an automatic driving inspection path through fuzzy comprehensive evaluation and screening. According to the method, complex sections are marked by counting the width and angle of a connection number channel, repeated paths are identified to avoid crossing redundancy, the paths are reconstructed to reduce cross interference, a heat source dense smoke change is evaluated to predict a shielded area, the paths are screened to improve safety and continuity, and path adaptability and execution efficiency are enhanced.

Description

Technical Field

[0001] The present invention relates to the field of automatic driving inspection technology, and in particular to an automatic driving inspection method and system for a firefighting robot. Background Art

[0002] The field of autonomous driving inspection technology includes core technical contents such as autonomous perception of target scenes in complex environments, path planning and travel control; it is mainly used in places that require regular inspection and monitoring, including power systems, petrochemical plants, mining areas and large-scale infrastructure. By integrating multi-sensor fusion perception, autonomous navigation systems and environmental recognition algorithms, unmanned equipment can complete operations such as autonomous path selection, obstacle avoidance, and inspection task execution in complex spaces; the development of this field covers multiple technical links such as robot navigation control, environmental map construction, dynamic obstacle avoidance, target recognition and positioning, and relies on high-precision positioning systems and environmental modeling methods to achieve accurate understanding of and autonomous response to the operating environment.

[0003] Among them, the automatic driving inspection method and system for fire-fighting robots refer to the use of environmental perception devices, path determination devices and execution devices deployed on the fire-fighting robots, in conjunction with multi-source data fusion means, to identify and determine information such as obstacle distribution, travel paths, and heat source areas in indoor or special work places, and generate adaptive path travel instructions in combination with preset inspection route rules and current environmental conditions; the patent subject specifically covers the area recognition method based on the joint analysis of visual images and temperature data, the steering control method achieved by calculating the change in path curvature, and the obstacle avoidance priority sorting mechanism based on space occupancy assessment to complete the automatic driving inspection task.

[0004] In the actual process of constructing inspection paths, existing technologies rely on fixed path planning and static environmental modeling. There is a lack of a refined judgment mechanism for path selection in areas with complex spatial structures, which easily leads to path duplication and redundancy in sections with high connection density and frequent angle changes. In addition, after the inspection path is generated, the assessment of risk areas generally relies on one-time identification, lacks periodic heat source and smoke status tracking, resulting in the inability to timely predict the dynamic changes of fire-obstructed areas. In places where the fire situation is complex or the obstructed areas change frequently, static paths are prone to failure and path instructions cannot be adjusted according to the real-time environment, resulting in path interruption or mission failure. Including in some factories or storage areas, when the fire source spreads rapidly or the smoke spreads suddenly, the original path is easily blocked by obstruction. Without a continuous evaluation mechanism and path adaptation capabilities, it will seriously affect the continuity of the fire-fighting robot's tasks and the efficiency of environmental response. Summary of the Invention

[0005] In order to solve the technical problems existing in the prior art, the present invention provides an automatic driving inspection method and system for a fire-fighting robot.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: An automatic driving inspection method for a firefighting robot comprises the following steps: S1: Obtain the grid structure data of the inspection area, count the number of grid node connections, channel width and connection angle, mark the sections that exceed the average value of the entire area, and generate data for complex structure sections; S2: Call the complex structure segment data, retrieve the passable path set, extract the node number, calculate the overlap ratio between the original path segment and the passable path set, mark the repeated traversal path segments, and construct the path segment set to be reconstructed; S3: Call the set of path segments to be reconstructed to obtain node numbers and arrange them, call the Dijkstra shortest path algorithm to generate path combinations, calculate the number of path intersections and path lengths to filter the paths, and generate an updated path segment sequence; S4: Based on the updated path segment sequence, the fire source and smoke coordinates are collected, the heat source position increment, smoke diffusion direction and speed changes are calculated, and the fire source state change evaluation is performed in a continuous cycle. The location of the fire shielding area is predicted and the shielding path segment data is generated. S5: Determine whether the updated path segment sequence falls into the obscured path segment data. If so, reconstruct the connection path, calculate the node redundancy length and the number of obscured overlaps, and screen the path segments through the fuzzy comprehensive evaluation model to generate an autonomous driving inspection path.

[0007] The following is a further optimization of the above technical solution by the present invention: The data of the complex structure section includes the location of the connection number node, the abnormal channel width section, and the connection angle fluctuation range. The set of path segments to be reconstructed specifically includes the repeated crossing path segment number sequence, the original path segment overlap ratio, and the pass path difference index. The updated path segment sequence includes the path segment combination order, the number of path intersections, and the path length. The obscured path segment data specifically refers to the heat source position increment area, the smoke diffusion density section, and the predicted obscuration coverage range. The autonomous driving inspection path includes the node redundancy length, the number of obscuration overlaps, and the fuzzy comprehensive evaluation preferred path.

[0008] Further optimization: The steps of S1 are specifically as follows: S101: After obtaining the grid structure data of the inspection area, based on the coordinate information and connection relationship of each grid node, the number of connections of all nodes is counted, and the angle value of the channel associated with each node is calculated based on the connection line segments between the nodes to generate a node structure parameter set; S102: Calculate the average values ​​of multiple parameters in the entire region based on the number of connections and angle values ​​in the node structure parameter set, mark the node numbers and segment numbers in the node structure parameters that are greater than any corresponding average value, and obtain the excess node distribution data; S103: Calling the node number information in the super-mean node distribution data, extracting the corresponding grid structure and surrounding connection information, merging and sorting the nodes according to the structural pattern formed in the spatial position, and generating complex structure segment data.

[0009] Further optimization: The steps of S2 are specifically as follows: S201: After accessing the complex structure section data, search for each passable path set in the area, obtain the node numbers included in the path and extract the node sequence information, summarize and organize the number sequence according to the arrangement relationship of the nodes in the spatial structure, and generate a passable node number set; S202: Calculate the overlap ratio between the original path segment and each pass path set based on the pass node number set, compare all the ratio values ​​with a set overlap ratio reference value, mark the path segment numbers that exceed the overlap ratio reference value, and obtain the repeated crossing ratio value; S203: extracting the node sequence and spatial direction corresponding to the original path segment from the path segment number information in the repeated crossing ratio value, summarizing and merging them to form a continuous segment number sequence, and establishing a path segment set to be reconstructed.

[0010] Further optimization: The overlap ratio benchmark value is set by statistically analyzing the overlap ratio distribution interval between the set of all passable paths in the inspection area and the original path segments, and extracting the median value or upper quartile value in the distribution interval as the interval critical point.

[0011] Further optimization: The steps of S3 are specifically as follows: S301: After calling the set of path segments to be reconstructed, the node numbers included in each path segment are sequentially extracted, the nodes are arranged according to the order in which they appear in the path, and a directed connection relationship of each path segment is constructed according to the order of the numbers to generate a path node sequence group; S302: Based on the node numbers of each group in the path node sequence group, the Dijkstra shortest path algorithm is called to calculate the modified edge weights in the structural network graph, and the nodes in each path combination process are sequentially numbered and recorded to form a combination chain of continuous path segments, and a shortest path combination chain group is established; S303: Calculate the number of repeated node numbers in the path segment and the total path length based on the shortest path combination chain group, and then jointly filter all path combinations according to the number of path intersections and path lengths to generate an updated path segment sequence.

[0012] Further optimization: the steps of S4 are specifically: S401: Based on the updated path segment sequence, collect the fire source center coordinate point and the smoke area boundary point group in the corresponding area of each path segment, calculate the spatial displacement vector of the fire source point at the adjacent time according to the coordinate difference, analyze the spatial displacement rate in the continuous period, and obtain the heat source position increment; S402: Call the path segment area identified by the heat source position increment, extract the smoke area boundary point group in the corresponding period, calculate the direction angle of smoke diffusion based on the angle relationship and time interval distance between the boundary points, and calculate the expansion length change rate and perform statistics to obtain the smoke diffusion trend parameter group; S403: According to the smoke diffusion trend parameter group and the heat source position increment value, calibrate the advancing direction and shielding coverage extension distance of the shielding boundary in the heat source state change period of the path segment, extract all path segment numbers in the shielding range, and generate shielding path segment data.

[0013] Further optimization: the steps of S5 are specifically: S501: Based on the updated path segment sequence and the shielding path segment data, compare the node numbers of each path segment, filter the path segments in the shielding range and record the path numbers, establish a mapping table of path segments and shielding relationship, and generate shielding path segment matching results; S502: According to the shielding path segment matching result, reconstruct the path connection relationship of the area with shielding path segments, collect the start and end node numbers and path segment total length value of each connection path, and calculate the node redundancy length and overlap times parameter group; S503: Call the node redundancy length and overlap times parameter group, score multiple path segments through a fuzzy comprehensive evaluation model according to the redundancy length ratio and overlap times coefficient of each path segment, filter the path number sequence whose score does not exceed the path segment screening reference value, and generate automatic driving inspection path data.

[0014] Further optimization: the path segment screening reference value is set by statistically calculating the weight score mean and standard deviation of multiple path segments in the inspection task, and combining the screening coefficient set by experience.

[0015] On the other hand, the application also provides an automatic driving inspection system for a fire-fighting robot, which is applied to the automatic driving inspection method for the fire-fighting robot, and the system comprises: A structure analysis module obtains grid structure data of an inspection area, counts the number of grid node connections, channel width and connection angle, marks sections exceeding the average value of the whole area, generates structure complex section data and transmits it to a path extraction module; The path extraction module calls the structure complex section data, retrieves the node number after the set of passable paths is searched, calculates the coincidence ratio of the original path section and the set of passable paths and marks the repeatedly crossed path section, constructs the set of path sections to be reconstructed and delivers to the path reconstruction module; The path reconstruction module calls the set of path sections to be reconstructed to obtain the node number and arrange, calls the Dijkstra shortest path algorithm to combine the paths to generate, calculates the number of path intersections and the path length to screen the paths, generates the sequence of updated path sections and delivers to the fire condition evaluation module; The fire condition evaluation module collects the fire source and smoke coordinates based on the sequence of updated path sections, calculates the heat source position increment, smoke diffusion direction and speed change and performs continuous periodical fire source state change evaluation, predicts the position of the fire disaster sheltered area, generates the sheltered path section data and delivers to the obstacle avoidance planning module; The obstacle avoidance planning module judges whether the sequence of updated path sections falls into the sheltered path section data, if yes, reconstructs the connected path, calculates the node redundancy length and the number of sheltered overlaps, screens the path section through the fuzzy comprehensive evaluation model, and generates the automatic driving inspection path.

[0016] The technical scheme provided by the embodiment of the application has at least the following beneficial effects: In the process of constructing and optimizing the inspection path, the connection number, the channel width and the connection angle of the regional grid nodes are counted, and the structure complex section is marked, so that the quantitative analysis of the complexity of the environment structure is effectively realized, the path planning can be optimized for the high complexity area. By calculating the coincidence ratio of the original path and the passable path, the repeatedly crossed path section is identified, the invalid path repetition is avoided, and the path accessibility is improved. Then the path is reconstructed through the Dijkstra algorithm, and the number of path intersections and the path length are combined for optimization, so that the path redundancy and intersection interference are effectively reduced. Further, after the path is generated, the fire source and smoke diffusion information are collected in real time, the heat source increment and the diffusion direction change are calculated, the fire source state is continuously and periodically evaluated, and the possible sheltered area is accurately predicted, so that the potential sheltered path section is dynamically marked and the path effectiveness is evaluated. When the path overlaps with the sheltered area, the system automatically evaluates the node length and the frequency of sheltered overlap, fuses the fuzzy comprehensive evaluation means to screen the path section, and improves the safety and execution continuity of the path. The overall logic fuses the spatial structure recognition, the path reconstruction, the risk warning and the multi-index evaluation, strengthens the pertinence, timeliness and robustness of the path selection, and significantly improves the adaptability and execution efficiency of the inspection task in the high-risk environment. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 The figure is a workflow schematic diagram of the embodiment of the application; Figure 2 The figure is a system module diagram in the embodiment of the application. DETAILED DESCRIPTION

[0018] In order to make the technical problems to be solved by the present application, technical solutions and advantages clearer, the following will be described in detail with reference to the drawings and specific embodiments.

[0019] In the embodiments of the present application, the words such as "example", "include" and the like are used to represent as an example, illustration or description. Any embodiment or design scheme described as "example" in the present application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. Rather, the word "example" is intended to present the concept in a specific manner. In addition, in the embodiments of the present application, the meaning expressed by "and / or" can be both, or can be one of the two optionally.

[0020] In the embodiments of the present application, "image" and "picture" can be used interchangeably at times, and it should be pointed out that the meanings expressed are consistent when the distinction is not emphasized. "Of", "corresponding" and "relevant" and "corresponding" can be used interchangeably at times, and it should be pointed out that the meanings expressed are consistent when the distinction is not emphasized.

[0021] In the embodiments of the present application, sometimes the subscript such as W1 can be written in the form of non-subscript such as W1, and the meanings expressed are consistent when the distinction is not emphasized.

[0022] Please refer to Figure 1 An automatic driving inspection method for a fire-fighting robot, comprising the following steps: S1: Obtain the grid structure data of the inspection area, count the number of grid node connections, the width of the channel and the connection angle, and mark the sections exceeding the average value of the whole area, and generate the structure complex section data; S2: Call the structure complex section data, retrieve the node number after searching the passable path set, calculate the coincidence ratio of the original path segment and the passable path set and mark the repeated traversal path segment, and construct the path segment set to be reconstructed; S3: Call the node number and arrange the path segment set to be reconstructed, call the Dijkstra shortest path algorithm to generate the path combination, calculate the number of path intersection points and the path length to filter the path, and generate the updated path segment sequence; S4: Based on the updated path segment sequence, collect the fire source and smoke coordinates, calculate the heat source position increment, smoke diffusion direction and speed change, and perform continuous periodical fire source state change evaluation, predict the position of the fire blocking area, and generate the blocking path segment data; S5: Determine whether the updated path segment sequence falls into the obscured path segment data. If so, reconstruct the connection path, calculate the node redundancy length and the number of obscured overlaps, and screen the path segments through the fuzzy comprehensive evaluation model to generate an autonomous driving inspection path.

[0023] The data of complex structural sections include the location of connection nodes, abnormal channel width sections, and connection angle fluctuation ranges. The set of path segments to be reconstructed specifically includes the repeated crossing path segment number sequence, the overlap ratio of the original path segments, and the traffic path difference index. The updated path segment sequence includes the path segment combination order, the number of path intersections, and the path length. The obscured path segment data specifically refers to the incremental area of ​​the heat source position, the smoke diffusion density section, and the predicted obscuration coverage range. The autonomous driving inspection path includes the node redundancy length, the number of obscuration overlaps, and the fuzzy comprehensive evaluation preferred path.

[0024] See also Figure 1 , the steps of S1 are as follows: S101: After obtaining the grid structure data of the inspection area, based on the coordinate information and connection relationship of each grid node, the number of connections of all nodes is counted, and the angle value of the channel associated with each node is calculated based on the connection line segments between the nodes to generate a node structure parameter set; Based on the grid structure data of the inspection area, the node coordinates and connection relationships are extracted, and the number of connections of each node is counted. ) as an example, connect node B 、C 、D , establish a connection relationship table to record the adjacent node IDs, traverse all nodes to generate a connection number statistics table, and for the three connecting line segments AB, AC, and AD of node A, calculate the angle between adjacent line segments: take the line segment AB vector , segment AC vector , through the dot product formula Calculated , repeat the calculation AB-AD to get 、AC-AD , take the minimum value As the angle value, Table 1 shows an example of 6 node parameters in a single area, where the number of connections of node E reaches 5 and the angle value , generate a parameter set including node coordinates, number of connections, and angle values.

[0025] Table 1 Node structure parameter example table Node number Number of connections Angle value (°) X coordinate (m) Y coordinate (m) E 5 28 4.2 7.8 F 2 152 3.1 5.4 G 4 78 2.7 6.3 H 3 65 1.2 3.4 I 2 120 5.6 2.9 J 3 90 7.3 4.7 As shown in Table 1, node E is identified as a key node due to its high number of connections and small angle value. This parameter set provides basic data for subsequent analysis.

[0026] S102: Calculate the average value of each type of parameter in the whole area according to the connection number and the included angle value in the node structure parameter set, mark the node number and section number in the node structure parameter that is greater than any corresponding average value, and obtain the super average node distribution data; When calculating the average value of the connection number, take the 6 node connection number data set {5, 2, 4, 3, 2, 3} in Table 1, the sum 19 divided by 6 is 3.17, set the super average threshold value as > 3.17, node E (connection number 5) and node G (connection number 4) are marked, the average value of the included angle is taken as {28, 152, 78, 65, 120, 90}, which is 88.83 degrees, and 28 degrees of node E is lower than the average value but still marked because the connection number exceeds the threshold value, generate section marking data by traversing all nodes, including section X including nodes E, F, G, 2 / 3 of which are marked as high-risk sections.

[0027] S103: Call the node number information in the super average node distribution data, extract the corresponding grid structure and surrounding connection information, merge and arrange according to the structure mode formed by the node in the spatial position, and generate structure complex section data; Extract 12 associated nodes within a 5-meter radius around node E to construct a sub-grid structure, analyze its connection mode: node E forms a star structure with nodes J and K (connection number ratio 3:1.2), and forms a ring structure with nodes L and M (standard deviation of included angle value up to 42°), calculate the density threshold of 0.8 connection points per cubic meter by space density algorithm, and merge to form a complex section with a diameter of 4.7 meters, the section includes 3 types of topological structure and 4 high-load nodes, and the generated structure complex section data includes spatial coordinate range , topological type distribution table and load level matrix.

[0028] Please refer to Figure 1 , the steps of S2 are as follows: S201: After calling the structure complex section data, retrieve each passable path set in the area, obtain the node number included in the path and extract the order information of the nodes, and according to the arrangement relationship of the nodes in the space structure, the numbering sequence is summarized and arranged, and the passable node number set is generated; Call the topological information of section X in the structure complex section data , traverse the 12 nodes and 28 connection edges included, retrieve the passable paths with the distance between the first and last nodes less than , including path P001 composed of nodes E-101 coordinates , J-205 , K-308 , calculate the distance between adjacent nodes: , the node sequence is recorded as {E-101, J-205, K-308}, and the path P002 is composed of nodes E-101, L-412 , M-517 , and the interval distances are and When the node sequence information of all paths is summarized, the node sequence {J-205, K-308, M-517} of path P003 is verified for spatial arrangement, and it is found that the interval distance from K-308 to M-517 exceeds the threshold value, the path is excluded, and a passing node number set including 6 valid paths is generated, and Table 2 shows the node sequence and coordinate verification results of 3 paths, by traversing the node sequence of all paths, a mapping relationship between path number and node sequence is established, including the coordinate interval distances of the node sequence {F-201, G-305, H-409} of path P004 , , which are verified and included in the number set.

[0029] Table 2: Verification table of node sequence of passing path Path number Node Sequence Spacing verification results (m) P001 E-101, J-205, K-308 0.14， 0.14 P002 E-101, L-412, M-517 0.11， 0.14 P004 F-201, G-305, H-409 0.12， 0.15 As shown in Table 2, the node sequence of the path is verified by the coordinate interval distance, and is stored in ascending order according to the path number, forming a structured passing node number set.

[0030] S202: According to the passing node number set, calculate the coincidence ratio value between the original path segment and each passing path set, compare all the ratio values with the set coincidence ratio reference value, mark the path segment number that exceeds the coincidence ratio reference value, and get the repeated traversal ratio value; In step S202, the coincidence ratio reference value is set by statistically analyzing the coincidence ratio distribution interval between all the passing path sets in the inspection area and the original path segment, and extracting the median value or upper quartile value in the distribution interval as the interval critical point; The node sequence {E-101, J-205, K-308} of path P001 in the passing node number set is compared with the segment F1 (node sequence {E-101, J-205, N-609}) in the original path segment library, and the number of coincident nodes is 2 (E-101, J-205), the total number of nodes is 3, and the coincidence ratio value , the coincidence ratio reference value is set to 0.6 (according to the data: analyze the coincidence degree distribution of 100 path segments, and take the threshold value corresponding to the 80% quantile), and other path segments are traversed, including the comparison between segment F2 (node sequence {L-412, M-527, Q-721}) and the node sequence {E-101, L-412, M-517} of path P002, the number of coincident nodes is 1 (L-412), and the ratio value , below the benchmark value, when marking the path segment exceeding the benchmark value, the segment F3 (node ​​sequence {J-205, K-308, M-517}) is compared with the path P001, and the number of overlapping nodes is 2 (J-205, K-308), and the ratio value , triggering the marking condition, summarizing all marking results to generate a repeated crossing ratio value data set, where path segments F1 and F3 are marked as high-repetition segments, and their numbers and ratio values ​​are stored in the result table, including the maximum overlap ratio value of segment F4 (node ​​sequence {G-305, H-409, I-503}) with any passing path. , exceeding the threshold.

[0031] S203: extracting the node sequence and spatial direction corresponding to the original path segment from the path segment number information in the repeated crossing ratio value, aggregating and merging them to form a continuous segment number sequence, and establishing a set of path segments to be reconstructed; Extract the node sequence {E-101, J-205, N-609} of the marked path segment F1 and calculate its spatial direction: the azimuth from node E-101 to J-205 , azimuth from J-205 to K-308 , average trend , the combined adjacent strike deviation is less than The path segments include the azimuth sequence of segment F5 (node ​​sequence {K-308, M-517, N-609}) 、 ,average , the deviation from F1 , trigger the merging condition, generate a continuous fragment number sequence {S01, S02}, and when delimiting the spatial range, take the extreme coordinates of all merged fragment nodes: minimum X = 4.1, maximum X = 4.5, minimum Y = 7.8, maximum Y = 8.1, calculate the node density: there are 8 nodes in the area, area ,density ,Combined with the distribution of connection numbers (mean 3.2, standard deviation 1.1), ,the nodes {E-101, J-205, M-517} with connection numbers ≥ 4 were ,screened to establish a set of path segments to be reconstructed, ,including 2 core segments, 3 highly connected nodes and 5 ,associated edges, covering 82% of the high-load connections in segment X.

[0032] See also Figure 1 , the specific steps of S3 are: S301: After calling the set of path segments to be reconstructed, the node numbers included in each path segment are sequentially extracted, the nodes are arranged according to the order in which they appear in the path, and a directed connection relationship of each path segment is constructed according to the order of the numbers to generate a path node sequence group; Call the fragments S01 (node ​​sequence {E-101, J-205, K-308}) and S02 (node ​​sequence {K-308, M-517, N-609}) in the path fragment set to be reconstructed, build a directed connection according to the order of node appearance, and extract the edge weights of E-101→J-205 for S01 in sequence , the edge weight of J-205→K-308 , recorded as a directed sequence {E→J→K}. When verifying the node order, check whether there are other fragments downstream of K-308 (such as K-308→M-517 of S02). If so, mark it as a connection point. Table 3 shows the node order and edge weight of fragment S01. By traversing all fragments, a path node sequence group is generated, including the edge weight sequence {0.12, 0.15} of fragment S03 (node ​​sequence {G-305, H-409, I-503}), which is included in the sequence group after verification.

[0033] Table 3 Path segment node order and edge weights Clip number Node Sequence Edge weight (km) S01 E-101, J-205, K-308 0.14， 0.14 S02 K-308, M-517, N-609 0.18， 0.22 As shown in Table 3, the node sequence is strictly sorted by the travel direction, providing structured input for subsequent path calculation.

[0034] S302: Based on the node numbers of each group in the path node sequence group, the Dijkstra shortest path algorithm is called to calculate the modified edge weights in the structural network graph, and the nodes in each path combination process are sequentially numbered and recorded to form a combination chain of continuous path segments, and a shortest path combination chain group is established; Based on the path node sequence group, Representative Node arrive The original edge weights of the node are obtained by measuring the node distance with a laser rangefinder, including the distance between nodes E-101 and J-205. , Represents the node degree centrality, which counts the number of directly connected edges of the node. For example, if node E-101 is connected to 5 edges, then , Represents the number of path turns, defined as the number of times the azimuth angle of adjacent line segments in the path changes by more than The number of times, including the azimuth difference between J→K and K→M in the path E→J→K→M is , not counting, , represents the path smoothing coefficient, which is set to the median of the smoothness of the path dataset (100 paths) , value range , Represents the turning penalty coefficient, and multiple The impact of the value on the path efficiency, select the value that minimizes the variance of the travel time , Representative The angle deviation of the turning angle is calculated, and the absolute value of the azimuth difference of adjacent line segments is calculated, including the single turning angle difference. ,but , taking node E-101 ( ) to J-205 ( ) as an example, the original edge weight (measured value), degree centrality term: , the path E→J→K has one turning point ( ), , path smoothing coefficient : Based on the path smoothness score (1-5 points), take the path with a score of ≥4 points as the percentage of 30% , turning penalty coefficient :Through controlled variable experiments, test Path efficiency when To minimize the average travel time, substitute the formula to calculate the shortest path. , denominator calculation: , , .

[0035] Table 4 Comparison of edge weight corrections Path Segment Original length (km) Corrected length (km) Is it selected? E-101 → J-205 0.14 0.283 yes J-205→K-308 0.14 0.265 yes K-308→M-517 0.18 0.401 no As shown in Table 4, the total length of the corrected path E→J→K Still lower than K→M , but the latter was screened out due to insufficient number of connections. The results verified the effective trade-off of the importance of node structure in the formula.

[0036] S303: Calculate the number of repeated node numbers in the path segments and the total path length based on the shortest path combination chain group, then jointly select all path combinations based on the number of path intersections and path lengths to generate an updated path segment sequence; Extract the path E→J→K→M from the shortest path combination chain (total length 1.2 km, repeated node K-308 appears twice), and calculate the proportion of repeated numbers , set the screening threshold: repetition ratio ≤ 0.4 and path length ≤ 1.5km, traverse other paths, such as path J→K→M→N (length 1.8km, repeated nodes 1 time, ratio 0.25), which is eliminated due to exceeding the length threshold. Path G→H→I (length 1.1km, no repeated nodes) meets both conditions. During joint screening, the number of path intersections is counted: the intersections of path E→J→K→M and path K→M→N are K-308 and M-517 (number 2), which exceeds the benchmark value 1. Paths G→H→I and E→J→K (number of intersections 1) are retained, and an updated path segment sequence {S01-1, S03-1} is generated, covering 92% of the efficient passage requirements in the network diagram.

[0037] See also Figure 1 , the steps of S4 are as follows: S401: Based on the updated path segment sequence, the coordinate points of the fire source center and the boundary points of the dense smoke area in the corresponding area of ​​each path segment are collected. The spatial offset vectors of the fire source points at adjacent moments are calculated based on the coordinate differences. The spatial displacement change rate within the continuous cycle is analyzed to obtain the heat source position increment. Based on the updated path segment sequence of path S01-1 (node ​​sequence {E-101, J-205, K-308}) and S03-1 (node ​​sequence {G-305, H-409, I-503}), the fire source center coordinate point data is collected. Taking the path S01-1 area as an example, at time Fire source detected The coordinates are ,time (interval ) The fire source point is offset to , calculate the spatial offset vector , displacement change rate , continuous monitoring for three cycles ( ), the coordinates of the fire source points are 、 、 , calculate the modulus lengths of adjacent incremental vectors respectively 、 , average rate of change ,Table 5 shows the offset data of the fire source point in the path S01-1 area.,The heat source position increment table is generated by traversing all path segments,,including the fire source point in the path S03-1 area from Offset to , increment , rate of change .

[0038] Table 5 Fire source point spatial offset data table Path number Time (s) X coordinate (m) Y coordinate (m) Offset (m) S01-1 0 4.25 7.82 - S01-1 60 4.28 7.85 0.042 S01-1 120 4.32 7.88 0.042 As shown in Table 5 , the fire source showed a stable trend of migrating to the northeast, and the incremental data of the heat source position provided input for the subsequent diffusion analysis.

[0039] S402: The path segment area identified by the heat source position increment is retrieved to extract the boundary point group of the dense smoke area within the corresponding time period. The direction angle of dense smoke diffusion is calculated based on the angular relationship between the boundary points and the time interval distance. The rate of change of the extended length is also calculated and statistically analyzed to obtain a set of dense smoke diffusion trend parameters. Call path S01-1 area in to The smoke boundary point group of the time period, Definition: The polar angular radian value of each boundary point relative to the heat source reference point is calculated through coordinate transformation, and the formula is: ,in is the coordinate of the heat source reference point, is the coordinate of the boundary point, the heat source reference point , boundary points coordinate ,but , Definition: The time interval between the acquisition times of adjacent boundary points (unit: seconds), which can be directly obtained by the difference of sensor timestamps. Example: , ,but , Definition: Time normalized reference quantity, fixed , Definition: The absolute value of the change in the length of smoke extension in adjacent time periods, measuring the difference in the radius of the smoke area in adjacent time periods. Example: Radius of Time Period 1 , radius of period 2 ,but , Definition: The first and The Euclidean distance between boundary points (unit: meter) is calculated as follows: , xy represents the coordinates of point p and point q, example: boundary point and spacing: , Definition: The diameter of the initial dense smoke area (unit: meter). The maximum horizontal and vertical spans of the initial dense smoke area boundary points are and , take the maximum value , substitute into the formula , the first calculation: , , , the second term calculation: ,result: , example results With preset threshold The comparison threshold is based on fire data statistics, Indicates that the diffusion direction is highly unstable, the current value , indicating that the diffusion direction of smoke during this period is relatively stable and can be marked as a “low-risk path segment”.

[0040] S403: Based on the smoke diffusion trend parameter group and the heat source position increment value, the advancing direction of the shielding boundary and the shielding coverage extension distance within the fire source state change cycle of the path segment are calibrated, and the path segment numbers of all path segments within the shielding range are extracted to generate shielding path segment data; Combined with the heat source position increment (path S01-1 average offset rate ) and smoke diffusion trend parameter ( ), the direction of the masking boundary is northeast , calculate the extension distance: , represents the average deviation rate, Representing time, the shielding range is defined as the fire source point Center, radius The circular area of ​​​​the path segment node coordinates are extracted, including node K-308 ( ) Distance from the center of the fire , which is smaller than the shielding radius and marked as within the shielding range. After traversing all nodes, the shielding path segment data is generated, including {K-308} of path S01-1 and {H-409} of path S03-1. The path number is S01-1, the shielding node number is K-308, and the distance from the fire source is (m) 0.143. The path number is S03-1, the shielding node number is H-409, and the distance from the fire source is (m) 0.210.

[0041] See also Figure 1 , the specific steps of S5 are: S501: Based on the updated path segment sequence and the obscured path segment data, the node number of each path segment is compared, the path segments within the obscured range are screened and the path numbers are recorded, a mapping table of path segment and obscured relationships is established, and a matching result of the obscured path segment is generated; Based on paths S01-1 (node ​​sequence {E-101, J-205, K-308}) and S03-1 (node ​​sequence {G-305, H-409, I-503}) in the updated path segment sequence, the shielded nodes K-308 and H-409 marked in the shielded path segment data are called. The path segment node sequence is traversed to check whether node K-308 of S01-1 exists in the shielded node table (Table 6). If so, the path number S01-1 is recorded. Similarly, node H-409 of S03-1 (0.210m away from the fire source, less than the shielding radius of 2.542m) is checked and the path number S03-1 is marked. When establishing the mapping relationship, the path number is associated with the shielded node. This includes path S02-2 (node ​​sequence {M-517, N-609}), which has no shielded node and is not included in the mapping table. Table 6 shows the matching results, covering 80% of all updated path segments.

[0042] Table 6 Masked path segment matching table Path number Shading node number Occlusion state S01-1 K-308 yes S03-1 H-409 yes S02-2 - no As shown in Table 6, paths S01-1 and S03-1 are marked because they include blocked nodes, providing input for path reconstruction.

[0043] S502: Based on the blocked path segment matching results, the path connection relationship is reconstructed for the area where the blocked path segment exists. The first and last node numbers and the total length of each connection path are collected. The number of repeated nodes in the path and the difference between the node connection lengths are calculated to obtain the node redundancy and overlap number parameter group. Reconstruct the connection relationship of the obscured path segment S01-1, extract the first and last nodes E-101 and K-308, and calculate the new path candidate solutions: 1. Path P1: E-101→F-201→K-308, total length 0.14+0.15=0.29km, number of repeated nodes 0, 2. Path P2: E-101→J-205→K-308, total length 0.14+0.14=0.28km, number of repeated nodes 2 (J-205, K-308), calculate the redundant length ratio: number of repeated nodes / total number of nodes for path P2 = 2 / 3≈0.67, for path P1 it is 0 / 3=0, length difference = |0.29-0.28 |=0.01km. Traverse other paths, such as path S03-1, and reconstruct the candidate path G-305→H-409→I-503 (length 0.12+0.15=0.27km, number of repeated nodes 1). The redundancy ratio is 1 / 3≈0.33. Paths with a redundancy ratio > 0.3 and a length difference > 0.005km are selected. Path number P1 has a redundancy ratio of 0.00 and a length difference (km) of 0.01. Path number P2 has a redundancy ratio of 0.67 and a length difference (km) of 0.00. Path number S03-1 has a redundancy ratio of 0.33 and a length difference (km) of 0.02. Path P2 is marked as having high redundancy.

[0044] S503: Calling the node redundancy and overlap number parameter group, scoring the multiple path segments using a fuzzy comprehensive evaluation model based on the redundancy ratio and overlap number coefficient of each path segment, selecting path number sequences whose scores do not exceed the path segment screening benchmark value, and generating autonomous driving inspection path data; In step S503, the path segment screening benchmark is set by statistically calculating the mean and standard deviation of the weighted scores of multiple path segments in the inspection task, combined with the screening coefficient set based on experience; In the parameter group, set the redundant length ratio of path P2 to 0.67 and the overlap coefficient to 1 (nodes J-205 and K-308 appear twice in the original path). Set the fuzzy evaluation weights: redundant length ratio weight 0.6, overlap coefficient weight 0.4, and score formula: ,in is the redundancy ratio, is the overlap coefficient (range: 0-1, 1 indicates the highest overlap), calculated for path P2: , set the path segment screening benchmark value to 0.5 (based on data: 80% effective path score ≤ 0.5), path P2 score 0.802 exceeded the threshold and was eliminated, path P1 score It is retained, and after traversing all paths, it generates autonomous driving inspection path data, including six low-scoring paths such as path P1 and S03-1, covering 92% of the safe passage needs in the area.

[0045] See also Figure 2 , an automatic driving inspection system for a fire-fighting robot, the automatic driving inspection system for a fire-fighting robot is used to execute the above-mentioned automatic driving inspection method for a fire-fighting robot, the system comprising: The structural analysis module obtains the grid structure data of the inspection area, counts the number of grid node connections, channel widths, and connection angles, and marks sections that exceed the average value of the entire area. It generates data on complex structural sections and passes it to the path extraction module. The path extraction module calls the data of complex structure sections, retrieves the set of passable paths, extracts node numbers, calculates the overlap ratio between the original path segments and the passable path set, marks the repeated traversal path segments, constructs the set of path segments to be reconstructed, and passes it to the path reconstruction module; The path reconstruction module calls the set of path segments to be reconstructed to obtain node numbers and arrange them, calls the Dijkstra shortest path algorithm to generate path combinations, calculates the number of path intersections and path lengths to filter paths, generates an updated path segment sequence, and passes it to the fire assessment module; The fire assessment module collects the coordinates of the fire source and smoke based on the updated path segment sequence, calculates the heat source position increment, the smoke diffusion direction and speed changes, and conducts continuous cycle fire source state change assessments. It predicts the location of the fire obscured area, generates obscured path segment data, and transmits it to the obstacle avoidance planning module. The obstacle avoidance planning module determines whether the updated path segment sequence falls into the obscured path segment data. If so, it rebuilds the connection path, calculates the node redundant length and the number of obscured overlaps, and screens the path segments through the fuzzy comprehensive evaluation model to generate an autonomous driving inspection path.

[0046] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0047] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. An automatic driving inspection method for a fire-fighting robot, characterized in that: The method comprises: S1: Obtain the grid structure data of the inspection area, count the number of grid node connections, channel width and connection angle, mark the sections that exceed the average value of the entire area, and generate data for complex structure sections; S2: Call the complex structure segment data, retrieve the passable path set, extract the node number, calculate the overlap ratio between the original path segment and the passable path set, mark the repeated traversal path segments, and construct the path segment set to be reconstructed; S3: Call the set of path segments to be reconstructed to obtain node numbers and arrange them, call the Dijkstra shortest path algorithm to generate path combinations, calculate the number of path intersections and path lengths to filter the paths, and generate an updated path segment sequence; S4: Based on the updated path segment sequence, the fire source and smoke coordinates are collected, the heat source position increment, smoke diffusion direction and speed changes are calculated, and the fire source state change evaluation is performed in a continuous cycle. The location of the fire shielding area is predicted and the shielding path segment data is generated. S5: Determine whether the updated path segment sequence falls into the obscured path segment data. If so, reconstruct the connection path, calculate the node redundancy length and the number of obscured overlaps, and screen the path segments through the fuzzy comprehensive evaluation model to generate an autonomous driving inspection path.

2. The automatic driving inspection method for a firefighting robot according to claim 1, characterized in that: The data of the complex structure section includes the location of the connection number node, the abnormal channel width section, and the connection angle fluctuation range. The set of path segments to be reconstructed specifically includes the repeated crossing path segment number sequence, the original path segment overlap ratio, and the pass path difference index. The updated path segment sequence includes the path segment combination order, the number of path intersections, and the path length. The obscured path segment data specifically refers to the heat source position increment area, the smoke diffusion density section, and the predicted obscuration coverage range. The autonomous driving inspection path includes the node redundancy length, the number of obscuration overlaps, and the fuzzy comprehensive evaluation preferred path.

3. The automatic driving inspection method for a fire-fighting robot according to claim 1, characterized in that: The steps of S1 are specifically as follows: S101: After obtaining the grid structure data of the inspection area, based on the coordinate information and connection relationship of each grid node, the number of connections of all nodes is counted, and the angle value of the channel associated with each node is calculated based on the connection line segments between the nodes to generate a node structure parameter set; S102: Calculate the average values ​​of multiple parameters in the entire region based on the number of connections and angle values ​​in the node structure parameter set, mark the node numbers and segment numbers in the node structure parameters that are greater than any corresponding average value, and obtain the excess node distribution data; S103: Calling the node number information in the super-mean node distribution data, extracting the corresponding grid structure and surrounding connection information, merging and sorting the nodes according to the structural pattern formed in the spatial position, and generating complex structure segment data.

4. The automatic driving inspection method for a firefighting robot according to claim 1, characterized in that: The steps of S2 are specifically as follows: S201: After accessing the complex structure section data, search for each passable path set in the area, obtain the node numbers included in the path and extract the node sequence information, summarize and organize the number sequence according to the arrangement relationship of the nodes in the spatial structure, and generate a passable node number set; S202: Calculate the overlap ratio between the original path segment and each pass path set based on the pass node number set, compare all the ratio values ​​with a set overlap ratio reference value, mark the path segment numbers that exceed the overlap ratio reference value, and obtain the repeated crossing ratio value; S203: extracting the node sequence and spatial direction corresponding to the original path segment from the path segment number information in the repeated crossing ratio value, summarizing and merging them to form a continuous segment number sequence, and establishing a path segment set to be reconstructed.

5. The automatic driving inspection method for a fire-fighting robot according to claim 4, characterized in that: The overlap ratio reference value is set by statistically calculating the overlap ratio distribution interval between the set of all passable paths in the inspection area and the original path segments, and extracting the median value or the upper quartile value in the distribution interval as the interval critical point.

6. The automatic driving inspection method for a firefighting robot according to claim 1, characterized in that: The steps of S3 are specifically as follows: S301: After calling the set of path segments to be reconstructed, the node numbers included in each path segment are sequentially extracted, the nodes are arranged according to the order in which they appear in the path, and a directed connection relationship of each path segment is constructed according to the order of the numbers to generate a path node sequence group; S302: Based on the node numbers of each group in the path node sequence group, the Dijkstra shortest path algorithm is called to calculate the modified edge weights in the structural network graph, and the nodes in each path combination process are sequentially numbered and recorded to form a combination chain of continuous path segments, and a shortest path combination chain group is established; S303: Calculate the number of repeated node numbers in the path segment and the total path length based on the shortest path combination chain group, and then jointly filter all path combinations according to the number of path intersections and path lengths to generate an updated path segment sequence.

7. The automatic driving inspection method for a firefighting robot according to claim 1, characterized in that: The steps of S4 are specifically as follows: S401: Based on the updated path segment sequence, the coordinate points of the fire source center and the boundary points of the dense smoke area in the corresponding area of ​​each path segment are collected. The spatial offset vectors of the fire source points at adjacent moments are calculated based on the coordinate differences. The spatial displacement change rate within the continuous cycle is analyzed to obtain the heat source position increment. S402: The path segment area identified by the heat source position increment is retrieved to extract the boundary point group of the dense smoke area within the corresponding time period. The direction angle of dense smoke diffusion is calculated based on the angular relationship between the boundary points and the time interval distance. The rate of change of the extended length is also calculated and statistically analyzed to obtain a set of dense smoke diffusion trend parameters. S403: Based on the smoke diffusion trend parameter group and the heat source position increment value, calibrate the advancement direction of the shielding boundary and the shielding coverage extension distance within the fire source state change cycle of the path segment, extract the numbers of all path segments within the shielding range, and generate shielding path segment data.

8. The automatic driving inspection method for a firefighting robot according to claim 1, characterized in that: The steps of S5 are specifically as follows: S501: Based on the updated path segment sequence and the obscured path segment data, the node number of each path segment is compared, the path segments within the obscured range are screened and the path numbers are recorded, a mapping table of path segment and obscured relationships is established, and a matching result of the obscured path segment is generated; S502: Based on the blocked path segment matching results, the path connection relationship is reconstructed for the area where the blocked path segment exists. The first and last node numbers and the total length of each connection path are collected. The number of repeated nodes in the path and the difference between the node connection lengths are calculated to obtain the node redundancy and overlap number parameter group. S503: Call the node redundancy and overlap number parameter group, score the multiple path segments through the fuzzy comprehensive evaluation model according to the redundancy ratio and overlap number coefficient of each group of path segments, screen the path number sequence whose score does not exceed the path segment screening benchmark value, and generate autonomous driving inspection path data.

9. The automatic driving inspection method for a fire-fighting robot according to claim 8, characterized in that: The path segment screening benchmark value is set by statistically analyzing the weighted score mean and standard deviation of multiple path segments in the inspection task, combined with the screening coefficient set based on experience.

10. An automatic driving inspection system for a fire-fighting robot, characterized in that: The system is used to implement the automatic driving inspection method for a firefighting robot according to any one of claims 1 to 9, and the system includes: The structural analysis module obtains the grid structure data of the inspection area, counts the number of grid node connections, channel widths, and connection angles, and marks sections that exceed the average value of the entire area. It generates data on complex structural sections and passes it to the path extraction module. The path extraction module calls the data of complex structure sections, retrieves the set of passable paths, extracts node numbers, calculates the overlap ratio between the original path segments and the passable path set, marks the repeated traversal path segments, constructs the set of path segments to be reconstructed, and passes it to the path reconstruction module; The path reconstruction module calls the set of path segments to be reconstructed to obtain node numbers and arrange them, calls the Dijkstra shortest path algorithm to generate path combinations, calculates the number of path intersections and path lengths to filter paths, generates an updated path segment sequence, and passes it to the fire assessment module; The fire assessment module collects the coordinates of the fire source and smoke based on the updated path segment sequence, calculates the heat source position increment, the smoke diffusion direction and speed changes, and conducts continuous cycle fire source state change assessments. It predicts the location of the fire obscured area, generates obscured path segment data, and transmits it to the obstacle avoidance planning module. The obstacle avoidance planning module determines whether the updated path segment sequence falls into the obscured path segment data. If so, it rebuilds the connection path, calculates the node redundant length and the number of obscured overlaps, and screens the path segments through the fuzzy comprehensive evaluation model to generate an autonomous driving inspection path.

Images