Robot path planning method based on topology awareness

Abstract

The invention relates to the technical field of path planning, in particular to a robot path planning method based on topology awareness, which comprises the following steps of: constructing grids and counting four-way shielding frequency, marking sparse transition density levels according to frequency threshold values, identifying sparse grid clusters and calculating boundary ratios, and marking passage priority or constraint segments. Analyzing path section direction stability, screening continuous priority edges, calculating edge connection fluctuation identification structure sensitive sections, and counting execution failure proportions to mark potential unaccessible path sections. According to the method, the grid shielding frequency in the obstacle distribution diagram is quantitatively calculated, the consistency of the planned path in the global navigation direction is improved, the critical structure area is accurately recognized in combination with path section boundary topological connection density analysis, the adaptive capacity of the path in the structure change area is effectively improved, and the navigation accuracy is improved. The potential inaccessible state of the path segment is dynamically labeled, and the response efficiency and execution stability of path planning to complex environment changes are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of path planning, in particular to a robot path planning method based on topological perception. BACKGROUND

[0002] The technical field of path planning mainly studies how to generate an optimal or feasible path for a mobile body from a starting position to a target position in a known or unknown environment. This field covers techniques such as geometric path planning, sampling space planning, graph search algorithms, numerical optimization methods, and reinforcement learning path generation. The core goals include obstacle avoidance, path smoothness, computational efficiency, and path length optimization, and it is widely used in intelligent manufacturing, warehouse logistics, service robots, surgical navigation systems, and mobile platform autonomous navigation scenarios.

[0003] Among them, the robot path planning method is a technical solution for realizing autonomous motion control of mobile robots. Its purpose is to automatically calculate the path that the robot needs to take from the current position to the target position according to environmental information and task goals, ensuring that the path is spatially feasible and meets the requirements of obstacle avoidance, efficiency, and path constraints. The method can be applied to navigation robots, AGV cars, inspection robots, and other systems to realize indoor and outdoor autonomous navigation, task path planning and execution, and improve system intelligence and operational efficiency.

[0004] Traditional planning methods using graph search or sampling space methods have difficulty accurately reflecting the spatial passability state in obstacle-dense or locally complex regions, lack a hierarchical expression mechanism for local topological structure characteristics, and cause path passability misjudgment when traversing structure boundary ambiguous regions during path generation. In global path direction control, only the shortest distance or cost function is considered, and the direction continuity of path segments is ignored, resulting in direction jump phenomenon in the planned path, affecting the smoothness and execution efficiency of the actual motion of the robot. In addition, the processing granularity is rough in the utilization of path execution history data, and the path validity label is not dynamically adjusted according to the path call record, resulting in weak adaptability of path planning to environmental changes in long-term operation. SUMMARY

[0005] The purpose of the present application is to solve the shortcomings in the prior art, and a robot path planning method based on topological perception is proposed.

[0006] In order to achieve the purpose, the present application adopts the following technical scheme: a robot path planning method based on topological perception, comprising the following steps: S1: Obtain a two-dimensional obstacle distribution map in a robot working area, establish a fixed-size grid cell set in the area, calculate the four-direction average blocking frequency of each grid, and perform interval judgment on the frequency and a preset blocking frequency threshold, to generate grid-level obstacle blocking level label information; S2: Call the grid-level obstacle blocking level label information, calculate the boundary grid number of each cluster area, mark the cluster area as a passing priority section or a constraint section, and generate a topological map passable area initial mapping structure; S3: Based on the topological map passable area initial mapping structure, obtain the line direction vector in each pair of path sections, extract the maximum deviation value and the median deviation value in the angle sequence and perform difference operation, select the path section that meets the stable direction feature, and generate a direction stable path section mapping result; S4: Call the direction stable path section mapping result, calculate the ratio difference value of the average number of path section two ends and the total number of neighborhood edges according to the path section start and end nodes, if the difference value is greater than the boundary fluctuation coefficient threshold, mark the corresponding section path as a structure critical section, and generate a structure sensitive path section identification result.

[0007] The grid-level obstacle blocking level label information includes sparse level area annotation, transition level area annotation, dense level area annotation, direction blocking statistical layer and grid number index, the topological map passable area initial mapping structure is specifically a passing priority area set, a constraint area set, a space connected cluster label, a boundary identification bitmap and a passable section space coordinate set, the direction stable path section mapping result includes a continuity priority path section set, a path section direction deviation rate label, a path section target alignment angle information, a path section direction stability score and a path section sequence number table, and the structure sensitive path section identification result specifically refers to a structure critical path section set, a path node connection edge fluctuation label, a three-order adjacent node relationship graph, a boundary jump section identification set and a topological overload flag.

[0008] The grid-level obstacle blocking level label information includes sparse level area annotation, transition level area annotation, dense level area annotation, direction blocking statistical layer and grid number index, the topological map passable area initial mapping structure is specifically a passing priority area set, a constraint area set, a space connected cluster label, a boundary identification bitmap and a passable section space coordinate set, the direction stable path section mapping result includes a continuity priority path section set, a path section direction deviation rate label, a path section target alignment angle information, a path section direction stability score and a path section sequence number table, and the structure sensitive path section identification result specifically refers to a structure critical path section set, a path node connection edge fluctuation label, a three-order adjacent node relationship graph, a boundary jump section identification set and a topological overload flag. S111: Obtain a two-dimensional obstacle distribution map in a robot working area, establish a fixed-size grid cell set in the area, calculate the four-direction average blocking frequency of each grid, and perform interval judgment on the frequency and a preset blocking frequency threshold, to generate grid-level obstacle blocking level label information; S112: Based on the grid average blocking frequency information, call the average frequency value of each grid and perform interval judgment with the set blocking frequency threshold, perform section identification operation in the interval where the frequency value is located, set the label attribute after judging that the corresponding section is a low frequency section, a medium frequency section or a high frequency section, and generate a frequency interval label list; S113: According to the frequency interval label list, each grid is classified and labeled according to the label attribute, the grid number index and the label corresponding relationship are called, the space labels of the three types of sparse, transition and dense labels are implemented, and the grid level obstacle blocking level label information is obtained.

[0009] The application improves that the obtaining step of the topological graph passable region initial mapping structure is specifically: S211: The grid level obstacle blocking level label information is called, the four-neighbor grid connectivity identification is performed based on the spatial adjacency relationship of the sparse labeled grid according to the sparse level labeling unit, the aggregation processing is performed on the grid groups that are continuously distributed in space, the closed region set composed of each group of adjacent grids is extracted and identified, and the sparse region connectivity cluster number index is generated; S212: The sparse region connectivity cluster number index is called, the boundary grid number and the core region grid point number of each cluster region are obtained according to the grid number information of the cluster region, the ratio of the boundary grid number to the core grid number is calculated, the structure compactness difference value is obtained by operation, and the structure boundary ratio difference sequence is generated; S213: According to the structure boundary ratio difference sequence, judgment processing is performed with the set structure compactness ratio threshold value, whether the structure compactness difference value is less than the structure compactness ratio threshold value is judged, if yes, the corresponding cluster region is marked as a passable segment, and if not, the cluster region is marked as a constraint segment, the cluster region is classified and marked, and the topological graph passable region initial mapping structure is obtained.

[0010] The application improves that the obtaining step of the direction stable path segment mapping result is specifically: S311: Based on the topological graph passable region initial mapping structure, the passable node set included in the passable segment is obtained, the double node pairs are combined in the node set according to the spatial coordinate sequence, the start and end coordinate points of each path segment are extracted and the path segment center line vector is constructed, the direction included angle is calculated in combination with the current preset global motion direction vector of the robot, and the path segment direction offset angle sequence is generated; S312: The path segment direction offset angle sequence is called, the maximum value and the median value of the path segment direction offset angle are calculated, the difference value is extracted as the direction fluctuation intensity factor, the path length value and the included angle change rate of the path segment are collected, the joint offset factor is calculated, the direction offset joint intensity value is obtained by operation, the direction offset joint intensity value of each path segment is compared with the direction offset stability threshold value, the path segments meeting the direction consistency are screened, and the direction consistency path segment sequence is obtained; S313: According to the direction consistency path segment sequence, the marked path segments in the original topological graph structure are marked, the path segments are assigned with the edge weight identification attribute of the continuity priority edge, and the direction stable path segment mapping result is established.

[0011] The application improves that the obtaining step of the structural sensitive path segment identification result is specifically: S411: calling the direction stable path segment mapping result, according to the start and end nodes of the path segment, counting the number of topological connection edges of each node, recording the arithmetic mean of the number of edges of the two end nodes of the path segment, and establishing the average edge number information of the two end nodes of the path segment; S412: according to the average edge number information of the two end nodes of the path segment, obtaining the third-order adjacent node set of each path segment, and counting the total number of connection edges of each node in the set, combining the network density background and the node distribution center of gravity index of the path segment in the topological graph, and calculating the structural boundary deviation intensity value to obtain a structural boundary deviation intensity value sequence; S413: calling the structural boundary deviation intensity value sequence, judging whether the structural boundary deviation intensity value is greater than the set boundary fluctuation coefficient threshold, if the condition is met, the corresponding path segment is marked as a structural critical segment, and the structural sensitive path segment identification result is established.

[0012] The application improves that the method further includes the following steps: S5: obtaining the path segment in the structural sensitive path segment identification result, calculating the path call failure ratio, comparing the failure ratio with the passable state reference ratio, if the ratio is higher than the set ratio threshold, performing a potential non-passable label calibration operation on the path segment, and generating dynamic passable topological path labeling information; The dynamic passable topological path labeling information includes a potential non-passable path segment index, a path segment historical failure comparison table, a dynamic path label graph, a passable state level label, and a path segment state change record.

[0013] The application improves that the obtaining step of the dynamic passable topological path labeling information is specifically: S511: obtaining the structural sensitive path segment identification result, reading the call number and the actual number of successful record times of each path segment in the robot continuous path call record, performing difference calculation on the two data, and corresponding processing with the path segment index to generate a path segment call failure number sequence; S512: based on the path segment call failure number sequence, according to the failure number and the total call number of the path segment, performing failure frequency ratio calculation, and comparing each item with the set passable state reference ratio, extracting the path segment index higher than the threshold value and summarizing to generate a path segment passable failure ratio value set; S513: calling the path segment passable failure ratio value set, according to the path segment index satisfying the threshold condition, labeling the path segment with a potential non-passable attribute label, and completing state writing in the topological graph structure to establish dynamic passable topological path labeling information.

[0014] Compared with the prior art, the application has the advantages and positive effects that: In the application, by quantitatively calculating the grid occlusion frequency in the obstacle distribution map and dividing the sparse, transition, dense levels according to the interval attribution, the hierarchical expression of the passability of the operation area is realized, the fine description of the path accessibility structure is enhanced by combining the boundary compactness evaluation of the connected cluster, the direction stability of the path segment is identified by using the direction vector angle difference operation, the consistency of the planned path in the global navigation direction is improved, the critical structure area is accurately identified by combining the boundary topology connection density analysis of the path segment, the adaptability of the path in the structure change area is effectively improved, and the response efficiency and execution stability of the path planning to the complex environment change are improved by introducing the path calling frequency and execution failure proportion linkage mechanism, the potential impassable state of the path segment is dynamically marked, and the environmental adaptability and dynamic robustness of the overall path are strengthened. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 The method flowchart of the application is shown in the following table: Figure 2 The flowchart of obtaining the grid-level obstacle occlusion level label information of the application is shown in the following table: Figure 3 The flowchart of obtaining the initial mapping structure of the topological map passable area of the application is shown in the following table: Figure 4 The flowchart of obtaining the direction stable path segment mapping result of the application is shown in the following table: Figure 5 The flowchart of obtaining the structure sensitive path segment identification result of the application is shown in the following table: Figure 6 The flowchart of obtaining the dynamic passable topological path marking information of the application is shown in the following table. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical scheme and advantages of the application more clear and explicit, the application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the application and not to limit the application.

[0017] In the description of the application, it should be understood that the orientations or positional relationships indicated by the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like are based on the orientations or positional relationships shown in the drawings, and are only for the convenience of describing the application and simplifying the description, and therefore cannot be understood as indicating or implying that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the application. In addition, in the description of the application, the meaning of "multiple" is two or more, unless otherwise specifically limited.

[0018] Please refer to Figure 1 The present application provides a technical solution: a robot path planning method based on topology perception, comprising the following steps: S1: Obtain the two-dimensional obstacle distribution map in the robot working area, establish a set of fixed-size grid cells in the area, collect the number of obstacle projections in four orthogonal directions of the grid, calculate the average frequency of four-directional obstacle blocking of each grid, and perform interval judgment on the frequency and the preset blocking frequency threshold, call the interval attribution label to generate three levels of grid label types of sparse, transition and dense, and generate grid-level obstacle blocking level label information; The blocking frequency threshold can be defined as the average value of the number of obstacle projections in each direction in the unit area grid, which is usually obtained according to the laser scanning blocking count of each direction, and is used to distinguish the complexity level of the environment structure; S2: Call the grid-level obstacle blocking level label information, identify the connected cluster area of the corresponding type grid in space according to the sparse level labeled unit, calculate the number of boundary grids for each cluster area, extract the ratio of the number of boundary grids to the number of core area grid points, and judge the ratio with the compactness ratio threshold, mark the cluster area as a priority section or a constraint section according to the judgment result, and generate an initial mapping structure of the topological map passable area; The compactness ratio threshold can be defined as the ratio of the boundary length to the area, which is commonly used to identify whether the area is edge or concentrated, and the reference source is the compactness measurement method in urban space modeling; S3: Based on the initial mapping structure of the topological map passable area, extract the double-node path segment according to the passable node set included in the priority section, obtain the line direction vector in each pair of path segments, calculate the included angle between the line direction vector and the current robot global expected motion direction vector, extract the maximum deviation value and the median deviation value in the included angle sequence and perform difference operation, compare the difference result with the direction deviation stability threshold, filter the path segments that meet the stable direction feature, and mark the path segments in the original topological graph structure as continuous priority edges, and generate the direction stable path segment mapping result; The line direction vector is the unit vector between the start point and the end point of the path segment, and the included angle can be calculated by inverse cosine; the direction deviation stability threshold is an important parameter for controlling the smoothness of the navigation path, which can be determined through simulation debugging or set according to the minimum turning angle path rule; S4: Call the direction stable path segment mapping result, according to the path segment start and end nodes, count the number of topological connection edges of each node and the total number of edges in the adjacent three-order nodes, calculate the ratio difference value of the average value of the number of edges at both ends of the path segment and the total number of edges in the neighborhood, and if the difference value is greater than the boundary fluctuation coefficient threshold, mark the corresponding path segment as a structure critical section, and generate a structure sensitive path segment identification result; The boundary fluctuation coefficient threshold can be set as a standard deviation ratio of the number of connected edges, which is used to identify edge sections and bottleneck sections in the topology graph. The relevant concepts can refer to the edge density change judgment criteria in the graph neural network. S5: Obtain the path section in the structure-sensitive path section identification result, read the difference between the frequency of the path section appearing in the robot continuous path call record and the number of path execution success records, calculate the path call failure rate, compare the failure rate with the passable state reference ratio, and if it is higher than the set ratio threshold, perform a potential non-passable label labeling operation on the path section, and generate dynamic passable topology path labeling information; The passable state reference ratio refers to the experience threshold of path historical call failure ratio, which can be set in combination with the task failure rate record in the robot path planning system; The grid-level obstacle blocking level label information includes sparse level area annotation, transition level area annotation, dense level area annotation, direction blocking statistical layer and grid number index. The topology passable area initial mapping structure specifically includes passable priority area set, constraint area set, space connected cluster label, boundary identification bitmap and passable section space coordinate set. The direction stable path section mapping result includes continuity priority path section set, path section direction offset rate label, path section target alignment angle information, path section direction stability score and path section sequence number table. The structure-sensitive path section identification result specifically refers to the structure-critical path section set, path node connection edge fluctuation label, three-order adjacent node relationship graph, boundary jump section identification set and topology overload flag. The dynamic passable topology path labeling information includes potential non-passable path section index, path section historical failure comparison table, dynamic path label graph, passable state level label and path section state change record.

[0019] Please refer to Figure 2 The acquisition steps of the grid-level obstacle blocking level label information are as follows: S111: Obtain a two-dimensional obstacle distribution map in the robot working area, establish a fixed grid element set in the area, collect the obstacle blocking count value of each grid in four orthogonal directions, and calculate the average value of the count value in each direction. An average blocking frequency data set based on the four-direction blocking of the grid is established, and grid average blocking frequency information is generated; Obtain the two-dimensional obstacle distribution map in the robot working area. When a set of grid cells with fixed side length is established in the area, first, according to the known spatial boundary coordinate point set read in the environment map file, a grid system with a size of 1 m x 1 m is constructed, the unique number information of each cell is obtained, and the number matrix is generated in the two-dimensional plane. Then, the obstacle blocking count of each grid cell in the north, east, south and west directions is collected. The number of times blocked in the four directions is counted and recorded as an integer array. Taking grid G1 as an example, if the north is 2 times, the east is 1 time, the south is 3 times, and the west is 0 times, then the direction blocking array is {2, 1, 3, 0}. Then, the average operation is performed on the four direction values of each grid to obtain the average blocking frequency of the grid. For example, the average frequency of G1 is (2+1+3+0) / 4=1.5. The average frequency data sequence of all grids is formed, and a two-dimensional array is established as the average blocking frequency data set. The result of this step is the grid average blocking frequency information. Table 1 Grid direction blocking frequency calculation table As shown in Table 1, the table shows the number of times of blocking in four directions and the average blocking frequency of some grids. The frequency value is used for subsequent interval identification and label determination.

[0020] S112: Based on the grid average blocking frequency information, the average frequency value of each grid is called and interval judgment is performed with the set blocking frequency threshold. The section recognition operation is performed in the interval where the frequency value is located. After determining that the corresponding section is a low frequency section, a medium frequency section or a high frequency section, the label attribute is set, and the frequency interval label list is generated. Based on the grid average blocking frequency information, the average frequency value of each grid is called and interval judgment is performed with the set blocking frequency threshold. The blocking frequency threshold is 1 and 2, which are two interval breakpoints. The setting basis is the statistical boundary of the average blocking frequency μ and the standard deviation σ of the overall grid of the region. The first threshold is set to μ−0.5σ, and the second threshold is set to μ+0.5σ. When μ=1.5 and σ=0.5, the first threshold is calculated to be 1.25, and the second threshold is calculated to be 1.75. Therefore, 1 and 2 are respectively rounded to 1 and 2 as the interval limit values. The low frequency interval is set to [0, 1), the medium frequency interval is set to [1, 2), and the high frequency interval is set to [2, ∞). The average frequency value of each grid is brought into the set interval judgment formula to perform section attribution recognition operation. For example, if the average frequency value of G1 is 1.5, it belongs to the medium frequency section. If the average frequency value of a certain grid is 2.25, it belongs to the high frequency section. If it is 0.8, it belongs to the low frequency section. Interval judgment uses inequality joint expression for judgment. Finally, each grid is attributed to its corresponding frequency label category and its label information is recorded to form a structured table composed of grid number and section label. The result is the frequency interval label list. S113: According to the frequency interval label list, each grid is classified and labeled according to the label attribute, the grid number index and the label corresponding relationship are called, the space labels of the three types of sparse, transition and dense labels are implemented, and the grid level obstacle blocking level label information is obtained; According to the frequency interval label list, all grids are classified and processed according to their assigned label attributes. First, the grid with low frequency label is classified as a sparse area grid, the medium frequency is classified as a transition area grid, and the high frequency is classified as a dense area grid. The numbers of different areas are marked in the grid matrix, and the mapping relationship table of label and space position is constructed based on the grid index information. The grid ID and its space coordinates and corresponding label are filled into the space structure array one by one. Finally, a complete sparse, transition, dense three type obstacle labeling layer information is established in the two-dimensional area graph, and the grid level obstacle blocking level label information is obtained.

[0021] Please refer to Figure 3 , the acquisition steps of the initial mapping structure of the topological graph are as follows: S211: Calling the grid level obstacle blocking level label information, based on the spatial adjacency relationship of the sparse labeled grid, the four-adjacent grid connectivity recognition is performed, the spatial continuous distribution grid group is aggregated, the closed area set composed of each group of adjacent grids is extracted and identified, and the sparse area connectivity cluster number index is generated; In the grid level obstacle blocking level label information, based on the spatial adjacency relationship of the sparse labeled grid, first, all the cell data marked as sparse label in the grid obstacle level label information is extracted, the spatial mapping table is established according to the two-dimensional coordinates, and then the four-adjacent judgment operation is performed on each sparse grid, that is, whether there is other sparse cell in the upper, lower, left and right four directions is detected. If there is, a connection relationship is established, and then all the grid cells connected with each other are aggregated according to the spatial connectivity, the grid set that can be connected with each other is extracted, and an independent closed area is formed. For example, if G1, G2 and G4 are spatially adjacent and marked as sparse area, they are classified into a connected cluster C1. After the aggregation of all sparse areas is completed, the aggregated clusters are numbered and identified one by one, and the information such as the grid number set corresponding to each cluster and the spatial boundary range is recorded. Finally, the sparse area connectivity cluster number index is obtained. S212: Calling the sparse area connectivity cluster number index, according to the grid number information of the cluster area, the boundary grid number and the core area grid point number of each cluster area are obtained, the ratio of the boundary grid number to the core grid number is calculated, and the formula is as follows: ; The structure compactness difference value is obtained by operation, and the structure boundary ratio difference sequence is generated; Among them, represents the structure compactness difference value, the normalized value of the boundary grid of the region, the normalized value of the core grid of the region, the normalized value of the frequency of the change of the boundary direction in the region, the growth adjustment term of the core density, the denominator term for constructing the structural expansion reference value, the entire fraction is multiplied by to constitute the structural fluctuation suppression factor; call the grid number information of each cluster region in the sparse area connectivity cluster number index, obtain the corresponding boundary grid number, core grid number, and boundary direction change frequency of each cluster region ; ; ; ; The formula consists of three parts: the numerator part is used to comprehensively express the combined expansion index of the relative redundancy of the boundary and the aggregation degree of the core region. When the boundary is large and the core is small, the value increases. At the same time, the logarithmic term controls the increase of the core grid influence when the number increases. The denominator part is used as a structural expansion scale standardization factor to make the results of different scale clusters comparable, and multiplied by plays a role in suppressing the structure disturbance, which is used to reduce the influence of the area with drastic direction change. The calculation result is the dispersion intensity of the compactness of the cluster structure. The larger the value, the less compact the structure. Taking cluster C2 as an example, assuming that the number of boundary grids is 10, the number of core grids is 30, and the frequency of direction change is 2, the formula is calculated as follows: First, normalize the processing: ; ; ; Then substitute the formula to calculate: ; Table 2 Calculation table of structural compactness difference value As shown in Table 2, the structural compactness difference value of cluster C2 is the largest, indicating that the boundary and core structure are unbalanced and the spatial structure is divergent, and the compactness is the weakest.

[0022] S213: According to the structure boundary ratio difference sequence, a judgment process is performed with the set structure compactness ratio threshold value to determine whether the structure compactness difference value is less than the structure compactness ratio threshold value. If it is true, the corresponding cluster region is marked as a passing priority section. If it is not true, it is marked as a constraint section. The cluster region is classified and marked to obtain the initial mapping structure of the topological map passable region.

[0023] According to the results in the structure boundary ratio difference sequence, the structure difference value of each cluster needs to be compared with the structure compactness ratio threshold value. The threshold value is the average value of all difference values in the current cluster set plus 0.25 times the standard deviation as the boundary reference. If the structure compactness difference value is 0.034 and the standard deviation is 0.016, the threshold value is set to 0.034 + 0.25 x 0.016 = 0.038. Therefore, the determination boundary is set to 0.038. If it is marked as a passing priority section, otherwise it is marked as a constraint section. According to the data in Table 2, C1, C3, C4, and C5 are less than the threshold value, which are passing priority sections, and C2 is a constraint section. The results show that the internal core structure of the C2 region is too loose or the boundary distribution is abnormal, which needs to be avoided in subsequent path construction. After the marking of all cluster regions is completed, the initial mapping structure of the topological map passable region is established, which includes the number of each cluster, the boundary compactness classification attribute, and the spatial position index. The advantage of the formula is that by introducing the boundary change frequency and the core expansion term, the compactness quantization result can cover multi-dimensional structure state, and effectively construct the basis for determining the path connectivity graph. The results show that the structure difference value not only reflects whether the boundary is closed, but also quantifies whether the core region has a coherent direction, further supporting the hierarchical construction of the region accessibility structure in path passing planning.

[0024] Please refer to Figure 4 The steps for obtaining the direction stable path segment mapping result are as follows: S311: Based on the initial mapping structure of the topological map passable region, the passable node set included in the passing priority section is obtained. The double node pairs are combined in the node set in the order of spatial coordinates. The start and end coordinate points of each path segment are extracted and the path segment centerline vector is constructed. The direction angle calculation is performed in combination with the current preset global motion direction vector of the robot to generate the path segment direction offset angle sequence. Based on the initial mapping structure of the topological graph passable area, all node sets contained in the path segments marked as passable priority segments are obtained, and the two-dimensional spatial position coordinates of each node in the map coordinate system are extracted. The node number is arranged in sequence from left to right and from top to bottom according to the spatial relative position, and a combination pair between two adjacent nodes is constructed to form a plurality of double-node combination pairs, each combination pair representing the start and end points of a path segment. Then, for each node pair, the midline vector of the path segment is generated by two-point coordinate difference calculation, which represents the direction from the start point to the end point. For example, if the nodes are , , the midline vector of the path segment is , and the current preset global motion direction of the robot is , then the included angle calculation is performed between each path segment midline vector and the direction vector, and the included angle calculation adopts the vector dot product formula: ; After obtaining the direction included angle, it is converted into an angle unit and a sequence of direction offset angles of all path segments is established, which is used for subsequent analysis of the direction fluctuation intensity, and finally a sequence of path segment direction offset angles is obtained; S312: The maximum value and the median value of the path segment direction offset angle are calculated by calling the sequence of path segment direction offset angles, and the difference value is extracted as the direction fluctuation intensity factor. The path length value and the included angle change rate of the path segment are collected, and the joint offset factor calculation is performed, using the formula: ; The direction offset joint intensity value is obtained by operation, and the direction offset joint intensity value of each path segment is compared with the direction offset stability threshold value to screen the path segments that meet the direction consistency, and a sequence of direction consistent path segments is obtained; wherein represents the direction offset joint intensity value, represents the maximum value of the path segment direction offset angle, represents the median value of the direction offset angle, represents the normalized length value of the path segment, represents the normalized value of the direction included angle change rate, represents the normalized value of the path density between the start and end points of the path segment, and the denominator controls the upper limit of the influence of the high-density path segment; The direction offset angle of all path segments in the sequence of path segment direction offset angles is called, and the maximum value and the median value are calculated, and the difference value between the two reflects the overall direction fluctuation degree, which is defined as the direction fluctuation intensity factor. Then, the normalized path length , the normalized value of the direction included angle change rate and the normalized path density value between the path start and end points , the joint strength value of the direction offset is calculated by introducing the joint offset factor formula: ; The structure logic of the formula is as follows: The direction change fluctuation range term is constructed, for amplifying the influence of the path segment with fast changing included angle, the distance influence scale is introduced, constituting the denominator to adjust the high-density path segment and suppress the offset contribution of the excessively dense path. The formula calculation result represents the joint strength of the direction offset of each path segment, and the smaller the value, the more stable the direction. Taking the path segment as an example, its parameters are as follows: , , , , ; then the formula is: ; Table 3: Calculation table of the joint strength value of the direction offset of the path segment As shown in Table 3, the joint strength value of the direction offset of the path segment D4 is the largest, indicating that the direction change fluctuation is the most severe, and the direction stability of the path segments D3 and D5 is the best. The benefit of the formula is that by considering the path length, direction change amplitude and density factors at the same time, the direction offset stability evaluation is more spatially continuous and consistent in motion trend.

[0025] S313: According to the sequence of the direction consistency path segment, mark the screened path segment in the original topological graph structure, assign the path segment with the edge weight identification attribute of the continuity priority edge, and establish the mapping result of the direction stable path segment.

[0026] According to the sequence of the direction consistency path segment, all the offset strength values The path segments below the preset stability threshold are marked, and the threshold is set to 4.000. Referring to the path segment D1 to D5 data, the path segments D3 and D5 that meet the condition are screened out. Then, the path segments are searched and located in the original topology graph structure, and the edge identification of the path segments in the topology graph is updated. The edge weight attribute of the path segments is marked as a continuity priority edge attribute, which is used to provide path screening basis in subsequent path planning. The direction stable path segment mapping result is established. The structure serves as an index basis for the direction consistent path segment in the robot navigation path screening, and ensures the consistency of the path continuity identification and the traceability of the graph structure. Finally, the structure contains the path segment number, direction stability identification, edge position index and other fields, which are used for downstream task calling.

[0027] Referring to Figure 5 , the acquisition step of the structure sensitive path segment identification result is specifically: S411: The direction stable path segment mapping result is called, the number of topology connection edges of each node is counted according to the start and end nodes of the path segment, the arithmetic mean of the number of edges connected to the two end nodes of the path segment is recorded, and the average number of edges connected to the two end nodes of the path segment is established. The information of all path segments in the direction stable path segment mapping result is called, and the start node number and end node number of each path segment are extracted in turn. The node position is located in the topology graph structure, and the number of directly connected edges of each node is counted. The number of edges connected to the two nodes of each path segment is respectively recorded as and , and the arithmetic mean is calculated, which represents the average topology connection strength of the path segment boundary structure, and is used to measure the dense degree of the connection relationship of the two ends of the path segment in the overall graph structure. For example, if the number of edges connected to the start node of a path segment is 6 and the number of edges connected to the end node is 8, the average number of edges is . In this way, all path segments are traversed, the path segment number and the corresponding average number of edges are recorded, and the complete average number of edges connected to the two ends of the path segment is established. The data provides a connection baseline reference for subsequent structure boundary discrimination.

[0028] S412: According to the average number of edges connected to the two ends of the path segment, the third-order adjacent node set of each path segment is obtained, and the total number of connected edges of each node in the set is counted. Combined with the network density background and the node distribution center of gravity index of the path segment in the topology graph, the formula is used: ; to obtain the structure boundary deviation strength value, and the structure boundary deviation strength value sequence is obtained. wherein represents the structure boundary deviation strength value, represents the total number of connected edges of the third-order adjacent nodes of the path segment normalized value, the average connection edge number normalization value of the two end nodes of the path segment, the path segment associated with the number of path segments in the topology graph normalization value, the node barycenter index normalization value of the area where the path segment is located, the adjustment factor of the structure barycenter index, usually in the range of 0.1 to 0.5, used to control the index weight, is a stable function of the number of coupled path segments, is used to enhance the boundary difference response under high-density structure; According to the obtained average edge number information of the two ends of the path segment, for each path segment, respectively determine the starting node and the ending node in the topology graph structure, further find the three-order adjacent node set of the node outward, and traverse all nodes in the set, count the number of connected edges of each node, and sum to obtain the total number of connected edges of the path segment in the three-order neighborhood, denoted as , which reflects the local topology density. The and the obtained in the previous step are combined, and are brought into the structure boundary deviation enhancement calculation formula: ; Wherein, represents the number of connections of the current path segment and other path segments in the topology graph normalization value, which controls the number of intersections of the path segment in the graph, represents the barycenter concentration degree of the node in the path segment area in the spatial distribution, represents the adjustment factor of the barycenter index, the value range is set to 0.3, and the deviation response gain is adjusted according to the concentration degree of the topology graph area structure. The calculation logic of the formula is: the numerator is the difference square value of the neighborhood and the path end, which represents the connection structure deviation degree; the square root in the denominator is used to control the expansion of the path intensity background, is a coupling term used to balance the influence of the number of connections on the structure diffusion; the whole is multiplied by , which further enhances the response of the path segment with barycenter to the boundary deviation. Taking path segment P1 as an example, the parameters are as follows: , , , , ; Substituting the calculation gives: ; Table 4 Structure boundary deviation intensity value example table ​​​​​​​​​​As shown in Table 4, the path segment P3 with a high deviation intensity value represents a large deviation of its adjacent structure from the average structure at both ends, which is a structural mutation point. The structural boundary deviation intensity value sequence is used for subsequent boundary fluctuation judgment in structural discrimination.

[0029] S413: Call the structural boundary deviation intensity value sequence, judge whether the structural boundary deviation intensity value is greater than the set boundary fluctuation coefficient threshold, if the condition is met, mark the corresponding path segment as a structural critical segment, and establish a structural sensitive path segment recognition result.

[0030] Call the deviation intensity value of each path segment corresponding to the obtained structural boundary deviation intensity value sequence , and compare it with the system set boundary fluctuation coefficient threshold to determine whether it belongs to the structural transition area. According to the previous structural dispersion analysis and path segment density fluctuation range, the threshold is set to 20.0. When the path segment deviation intensity value is greater than this threshold, it is considered that the boundary is at the structural mutation position from dense to sparse or from sparse to dense, and the path segment is marked as a structural critical segment and included in the structural sensitive area path set; otherwise, it is not marked. Taking the path segment as an example, P1, P3, P5 all exceed the threshold 20.0, and are marked as structural critical segments, and the rest do not meet the conditions. The marking result is associated with the path segment number to establish a structural sensitive path segment recognition result, which is used for the next path selection or graph structure updating process. The final structural identification will be stored in three dimensions of path segment ID, deviation intensity, and sensitive segment.

[0031] Please refer to Figure 6 , the steps of obtaining dynamic accessible topology path marking information are as follows: S511: Obtain the structural sensitive path segment recognition result, read the number of calls and the number of successful records of each path segment in the robot continuous path call record, and calculate the difference between the two data, and process the path segment index to generate a path segment call failure number sequence; Obtain all path segment indexes in the structural sensitive path segment recognition result, and call the call log of each path segment in the robot path planning system continuous job record in turn, extract the call number and corresponding successful record number of each path segment. The call number is denoted as , and the successful record number is denoted as , which comes from the path planning history table and task completion feedback record table in the control system scheduling module. Taking path segment number P1 as an example, if its total call number is 80 times, of which the number of successful completion records is 68 times, the difference is The difference represents the number of failures of the path segment, indicating the frequency of the segment failing to complete the planning due to navigation failure, obstacle avoidance timeout, or dynamic path conflict during use. The number of failures of each path segment is collected according to the path segment number to establish a path segment call failure number sequence, which is a key basis for subsequent passage state stability judgment.

[0032] According to the generated path segment call failure number sequence, the failure number corresponding to each path segment number is and the total number of calls are proportionally converted, and the failure proportion calculation formula is executed: ; The calculation logic is to first obtain the failure rate of each path segment and uniformly convert it to a percentage format, representing the passage failure frequency of each path segment. Further, the result is compared with the system set passable state reference proportion threshold value, which is 10%. This value is set based on the average path planning error tolerance rate of the robot platform in the indoor working environment and the multiple path segment scheduling data backtest results. If the number of failures of path segment P1 is 12 and the total number of calls is 80, then the failure proportion is: ; Therefore, the failure proportion of path segment P1 is higher than the set threshold value of 10%, which is a potential path segment with low passage stability. In this way, the failure proportion values of all path segments are calculated, the path segment numbers of all failure proportions exceeding the threshold value are extracted, and the number, failure number, call number, and calculated failure proportion are recorded to form a path segment passage failure proportion value set.

[0033] Table 5: Path segment passage failure proportion example table As shown in Table 5, the failure proportions of path segments P1, P3, and P6 all exceed the threshold value, and need to enter the state calibration process.

[0034] S512: Based on the path segment call failure number sequence, according to the failure number and the total number of calls of the path segment, the failure frequency proportion calculation is executed, and the set passable state reference proportion is compared item by item, the path segment index higher than the threshold value is extracted and summarized, and the path segment passage failure proportion value set is generated; The path segment numbers whose call path segment passing failure ratio value set is higher than the system set threshold value 10% are updated in the topology graph structure, the edges between the path segment nodes are set in the potential inaccessible state, and a specific identifier is given to identify and avoid in the subsequent path scheduling process. The state writing operation is realized by updating the path edge weight additional field in the topology graph data structure, for example, the edge weight field "access_status" of P1, P3 and P6 is adjusted from "normal" to "uncertain". The identifier will be used as a path segment risk weight reference in the dynamic path scheduling module. After completing the state writing of all the marked path segments, the dynamic accessible topology path marking information is formed, which is used for path fusion optimization processing in the next stage.

[0035] S513: Call the path segment passing failure ratio value set, mark the potential inaccessible attribute label of the path segment according to the path segment index satisfying the threshold condition, complete the state writing in the topology graph structure, and establish the dynamic accessible topology path marking information.

[0036] The above is only a preferred embodiment of the present application, and does not limit the present application in other forms. Any skilled person in the art can use the disclosed technology to make changes or modifications as equivalent embodiments applied to other fields, but any simple modification, equivalent change and modification made on the basis of the technical essence of the present application to the above embodiments without departing from the technical solution content of the present application still belongs to the protection scope of the present application technical solution.

Claims

1. A robot path planning method based on topology perception, characterized in that: The following steps are involved: S1: Obtain a two-dimensional obstacle distribution map within the robot's operating area, establish a fixed-size grid unit set in the area, calculate the average occlusion frequency in the four directions of each grid, and perform interval judgment between the frequency and the preset occlusion frequency threshold to generate grid-level obstacle occlusion level label information; S2: calling the grid-level obstacle occlusion level label information, calculating the number of boundary grids for each cluster area, marking the cluster area as a passable priority segment or a constraint segment, and generating an initial mapping structure of the passable area of ​​the topological map; S3: Based on the initial mapping structure of the passable area of ​​the topological map, obtain the midline direction vector of each pair of path segments, extract the maximum offset value and the median offset value in the angle sequence and perform a difference operation, screen the path segments that meet the stable direction feature, and generate a direction-stable path segment mapping result; S4: Call the mapping result of the directionally stable path segment, and calculate the ratio difference between the average number of edges at both ends of the path segment and the total number of edges in the neighborhood according to the start and end nodes of the path segment. If the difference is greater than the boundary fluctuation coefficient threshold, the corresponding segment path is marked as a structural critical segment, and a structurally sensitive path segment identification result is generated.

2. The robot path planning method based on topology perception according to claim 1 is characterized in that: The grid-level obstacle occlusion level label information includes sparse level area labeling, transition level area labeling, dense level area labeling, directional occlusion statistical layer and grid number index; the initial mapping structure of the passable area of ​​the topological map is specifically a set of passable priority areas, a set of constraint areas, a spatial connectivity cluster label, a boundary identification bitmap and a passable segment spatial coordinate set; the directionally stable path segment mapping result includes a continuity priority path segment set, a path segment direction offset rate label, a path segment target alignment angle information, a path segment direction stability score and a path segment sequence number table; the structurally sensitive path segment identification result specifically refers to a set of structural critical path segments, a path node connection edge fluctuation label, a third-order adjacent node relationship graph, a boundary jump segment identification set and a topological overload flag.

3. The robot path planning method based on topology perception according to claim 2 is characterized in that: The steps for obtaining the grid-level obstacle occlusion level label information are specifically as follows: S111: Obtain a two-dimensional obstacle distribution map within the robot's operating area, establish a set of grid cells with fixed side lengths in the area, collect obstacle occlusion counts for each grid in four orthogonal directions, and average the counts in each direction to establish an average occlusion frequency dataset for the grid based on the four-directional occlusions, thereby generating grid average occlusion frequency information; S112: Based on the average occlusion frequency information of the grids, the average frequency value of each grid is called and compared with the set occlusion frequency threshold to perform interval judgment, perform a segment identification operation for the interval where the frequency value is located, determine whether the corresponding segment is a low frequency segment, a medium frequency segment, or a high frequency segment, and then set a label attribute to generate a frequency interval label list; S113: Based on the frequency interval label list, each grid is classified and labeled according to the label attributes, the correspondence between the grid number index and the label is called, and spatial labeling of three types of labels, namely sparse, transitional, and dense, is implemented to obtain grid-level obstacle occlusion level label information.

4. The robot path planning method based on topology perception according to claim 3 is characterized in that: The steps for obtaining the initial mapping structure of the accessible area of ​​the topological map are specifically as follows: S211: Calling the grid-level obstacle occlusion level label information, performing four-adjacent grid connectivity identification based on the spatial adjacency relationship of the sparsely labeled grids according to the sparse level labeling unit, aggregating the spatially continuously distributed grid groups, extracting the aggregated areas, and identifying the closed area set formed by each group of adjacent grids, and generating a sparse area connected cluster number index; S212: calling the sparse area connected cluster number index, obtaining the number of boundary grids and the number of grid points in the core area of ​​each cluster area according to the grid number information of the cluster area, calculating the ratio of the number of boundary grids to the number of core grids, calculating the structural compactness difference value, and generating a structural boundary ratio difference sequence; S213: Based on the structural boundary ratio difference sequence and the set structural compactness ratio threshold, a judgment process is performed to determine whether the structural compactness difference value is less than the structural compactness ratio threshold. If so, the corresponding cluster area is marked as a traffic priority segment; if not, it is marked as a constraint segment. The cluster area is classified and marked to obtain the initial mapping structure of the accessible area of ​​the topological map.

5. The robot path planning method based on topology perception according to claim 4 is characterized in that: The steps for obtaining the mapping result of the directional stable path segment are specifically as follows: S311: Based on the initial mapping structure of the passable area of ​​the topological map, a passable node set included in the passable priority segment is obtained, two node pairs are combined in the node set according to the spatial coordinate order, the start and end coordinate points of each path segment are extracted and the path segment centerline vector is constructed, and the direction angle is calculated in combination with the current preset global motion direction vector of the robot to generate a path segment direction offset angle sequence; S312: Calling the path segment direction offset angle sequence, calculating the maximum and median values ​​of the path segment direction offset angles, extracting the difference as a direction fluctuation intensity factor, collecting the path length values ​​and angle change rates of the path segments, calculating a joint offset factor, and calculating a direction offset joint intensity value. The direction offset joint intensity value of each path segment is compared with a direction offset stability threshold, and path segments that meet direction consistency are screened to obtain a direction consistent path segment sequence. S313: According to the direction-consistent path segment sequence, a marking operation is performed on the screened path segments in the original topology structure, and the path segments are assigned edge weight identification attributes of continuity priority edges to establish a direction-stable path segment mapping result.

6. The robot path planning method based on topology perception according to claim 5 is characterized in that: The steps for obtaining the structure-sensitive path segment identification result are specifically as follows: S411: calling the direction-stable path segment mapping result, counting the number of topological connection edges of each node according to the start and end nodes of the path segment, recording the arithmetic average of the number of edges of the nodes at both ends of the path segment, and establishing the average number of edges at both ends of the path segment; S412: Based on the average number of edges at both ends of the path segment, a set of third-order adjacent nodes at both ends of each path segment is obtained, and the total number of connected edges of each node in the set is counted. Combined with the network density background of the path segment in the topological graph and the node distribution center of gravity index, a structural boundary deviation strength value is calculated to obtain a structural boundary deviation strength value sequence; S413: calling the structure boundary deviation strength value sequence to determine whether the structure boundary deviation strength value is greater than a set boundary fluctuation coefficient threshold; if the condition is met, marking the corresponding path segment as a structure critical segment, and establishing a structure sensitive path segment identification result.

7. The robot path planning method based on topology perception according to claim 6 is characterized in that: The method further comprises the following steps: S5: Obtaining path segments from the structure-sensitive path segment identification results, calculating a path call failure ratio, and comparing the failure ratio with a passable state benchmark ratio. If the ratio is higher than a set ratio threshold, performing a potential unpassable labeling operation on the path segment to generate dynamic passable topology path labeling information; The dynamic accessible topology path annotation information includes a potential inaccessible path segment index, a path segment history failure comparison table, a dynamic path label map, a accessible status level label, and a path segment status change record.

8. The robot path planning method based on topology perception according to claim 7 is characterized in that: The steps for obtaining the dynamic accessible topology path annotation information are specifically as follows: S511: Obtain the structure-sensitive path segment recognition result, read the number of calls that appear in the robot's continuous path call record and the number of successful records of actual completion for each path segment, call the two data to perform difference calculation, and perform corresponding processing with the path segment index to generate a sequence of the number of path segment call failures; S512: Based on the sequence of the number of call failures for the path segments, a failure frequency ratio is calculated according to the number of failures for the path segments and the total number of calls. The ratio is compared item by item with a set passable state benchmark ratio, and the path segment indexes exceeding the threshold are extracted and summarized to generate a path segment passable failure ratio value set. S513: Call the path segment failure ratio value set, mark the path segment with a potential unpassable attribute label according to the path segment index that meets the threshold condition, complete the status writing in the topology structure, and establish dynamic passable topology path marking information.

Images