Multi-target path optimization method, equipment and medium

Abstract

The invention discloses a multi-target path optimization method, equipment and a medium, and relates to the technical field of path planning. The method comprises the following steps: constructing a weighted graph model of a to-be-planned area and verifying an adjacent matrix to obtain verified weighted graph data; shortest path distances among all node pairs are calculated to form an all-source shortest path distance matrix and a precursor record set, and connectivity detection and feasible region pruning processing are carried out on unreachable node pairs; generating an optimized access sequence covering all target nodes based on the processed full-source shortest path metric data and a nearest neighbor greedy strategy, accumulating composite path cost, and performing feasibility check of time windows, capacity and risk constraints to generate an initial path sequence and accumulated cost data; and expanding adjacent node pairs in the initial path sequence into a specific executable path on the original graph according to the precursor record set, performing neighborhood optimization on the executable path by adopting a local search operator, and outputting an optimized final path scheme and an accumulated cost report.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of path planning, in particular to a multi-objective path optimization method, device and medium. BACKGROUND

[0002] With the popularization of smart park inspection, logistics distribution scheduling and multi-agent collaborative navigation applications, path planning technology has become the core support for improving task execution efficiency and resource utilization. There are mainly two kinds of path planning methods in current engineering practice. One is to rely on global optimization models such as the traveling salesman problem or the vehicle routing problem, to seek the theoretical optimal solution through complex mathematical modeling and accurate search, but the calculation and storage overhead increases exponentially with the node size, which is difficult to apply to large-scale real-time scenarios. The other is to use simple heuristic rules or empirical strategies to quickly generate feasible paths, which has improved the calculation efficiency, but due to the lack of systematic optimization mechanism and constraint processing capability, it often leads to unstable solution quality, unreasonable resource allocation and other prominent problems in complex graph structures or dynamic environments.

[0003] The existing methods are generally single in the path evaluation dimension. Most schemes only use geometric distance as the core cost index, which is difficult to effectively integrate time window constraints, energy consumption limits, risk area avoidance and other multi-dimensional factors. This single evaluation mechanism often leads to time conflicts, energy shortages or safety risks in the actual execution of the generated path, which seriously restricts the global quality and practicality of the path scheme. At the same time, the traditional heuristic path construction process lacks dynamic feedback and adaptive adjustment capability. Once the initial path sequence is generated, it cannot be optimized online according to real-time environmental changes, which makes the robustness and adaptability of the path planning result in dynamic scenarios significantly insufficient.

[0004] In addition, most existing technologies do not fully utilize graph structure information, and fail to effectively integrate prior metric information such as all-source shortest path distance matrix into the path generation process. The path construction stage still needs to repeatedly perform local exploration and cost calculation on the original graph, which not only increases the computational overhead, but also may lead to missed path connectivity detection or improper handling of weakly connected areas. On the other hand, existing systems often rely on manual experience to set cost weights, heuristic thresholds and other parameters, which need to be frequently adjusted manually in different scenarios. This not only reduces the portability of the scheme, but also introduces additional uncertainty, making it difficult to maintain stable path planning results when the scale expands or constraints change. SUMMARY

[0005] The embodiments of the present application provide a multi-objective path optimization method, device and medium to solve the above technical problems.

[0006] In one aspect, the embodiments of the present application provide a multi-objective path optimization method, comprising: construct a weighted graph model of the region to be planned, represent the path cost relationship between nodes in the form of an adjacency matrix, and perform symmetry, non-negativity, dimensional consistency and connectivity checks on the adjacency matrix to generate the checked weighted graph data; Based on the checked weighted graph data, use the shortest path algorithm to calculate the shortest path distance between all node pairs, form the all-source shortest path distance matrix and the predecessor record set, and perform connectivity detection and feasible region pruning processing on the unreachable node pairs; Based on the processed all-source shortest path metric data and the nearest neighbor greedy strategy, generate an optimized access sequence covering all target nodes, and at each step, accumulate the composite path cost in real time, and perform feasibility verification of time window, capacity and risk constraints to generate the initial path sequence and cumulative cost data. When all target nodes have been visited, expand the adjacent node pairs in the initial path sequence into specific executable paths on the original graph according to the predecessor record set, and use local search operators to optimize the executable paths in the neighborhood to reduce the total path cost, and output the optimized final path scheme and cumulative cost report when the termination condition is met.

[0007] In an implementation manner of the present application, a weighted graph model of the region to be planned is constructed, the path cost relationship between nodes is represented in the form of an adjacency matrix, and the adjacency matrix is checked for symmetry, non-negativity, dimensional consistency and connectivity to generate the checked weighted graph data, specifically including: Obtain node data and edge data of the region to be planned, and based on the node data and edge data, construct an adjacency matrix; wherein the node data includes a location identifier, the edge data includes a connection relationship identifier, the element value of the adjacency matrix represents the edge weight value between the corresponding nodes, and the edge weight value in the adjacency matrix is calculated based on at least one of path distance, time, energy consumption or risk; Symmetry check is performed on the adjacency matrix to correct the asymmetric edge weight values found into symmetric values, and the symmetric adjacency matrix is checked for non-negativity to replace the negative weight edges found with preset positive weight values; The non-negative adjacency matrix is checked for dimensional consistency, the edge weight values with inconsistent dimensions are converted to a uniform unit, and the adjacency matrix with consistent dimensions is checked for connectivity, isolated nodes or edges are identified and marked through a graph traversal algorithm, and a connected component report is generated; Based on the connected component report, the abnormal edges are corrected, and the final checked weighted graph data is generated.

[0008] In an implementation form of the present application, based on the verified weighted graph data, the shortest path algorithm is used to calculate the shortest path distance between all node pairs, to form the all-source shortest path distance matrix and the predecessor record set, specifically comprising: Each node is selected as a source point in turn, and the improved Dijkstra algorithm is executed to calculate the shortest distance vector and the predecessor vector from the source point to all other nodes; The shortest distance vectors of all source points are superimposed to form the all-source shortest path distance matrix, and the predecessor vectors of all source points are integrated to form the predecessor record set.

[0009] In an implementation form of the present application, connectivity detection and feasible region pruning processing are performed on the unreachable node pairs, specifically comprising: The all-source shortest path distance matrix is subjected to unreachable node pair detection to identify node pairs with infinite distance values, and to generate unreachable marking data; Based on the unreachable marking data, connectivity analysis is performed to determine the maximum connected subgraph by using a graph traversal algorithm, and feasible region pruning processing is performed to eliminate non-connected nodes, thereby generating a pruned node set; Based on the pruned node set, the all-source shortest path distance matrix and the predecessor record set are updated to generate the corrected all-source shortest path metric data.

[0010] In an implementation form of the present application, based on the processed all-source shortest path metric data and the nearest neighbor greedy strategy, an optimized access sequence covering all target nodes is generated, specifically comprising: The path starting node is selected, and the current node state, the unvisited node set, the path sequence and the cumulative cost variable are initialized; In each iteration, the node with the shortest distance to the current node is selected from the unvisited node set as the next node based on the all-source shortest path distance matrix, and the nearest neighbor selection data is generated; Based on the nearest neighbor selection data, the path sequence is updated, the next node is appended to the path sequence, and the updated path sequence data is generated.

[0011] In an implementation form of the present application, the composite path cost is accumulated in real time after each selection, and the feasibility of time window, capacity and risk constraints is verified to generate the initial path sequence and the cumulative cost data, specifically comprising: Based on the composite cost function, the composite cost from the current node to the next node is calculated and accumulated into the total cost variable in real time to generate the cumulative cost data; the composite cost function integrates the distance cost, the time cost, the energy consumption cost and the risk penalty cost; performing feasibility checking on the next node based on time window, capacity and risk constraints, and generating a constraint checking report; the feasibility checking includes time window conflict detection, capacity overrun checking and risk area avoidance; adjusting the path sequence or composite cost parameters according to the constraint checking report, generating an initial path sequence after feasibility processing and cumulative cost data, and repeating the iteration process until the set of unvisited nodes is empty, and completing the path sequence generation.

[0012] In an implementation manner of the application, when all target nodes are visited, adjacent node pairs in the initial path sequence are unfolded into specific executable paths on the original graph according to the predecessor record set, specifically including: obtaining adjacent node pair data in the initial path sequence, and based on the predecessor record set, finding a sequence of intermediate nodes through which the shortest path of each adjacent node pair passes on the original graph; splicing the sequences of intermediate nodes between all node pairs, removing duplicate nodes after splicing, generating a preliminary executable path, and performing smoothness checking on the preliminary executable path to generate an optimized executable path; verifying the connectivity and validity of the executable path in the original graph, ensuring that all node pairs are correctly unfolded based on the predecessor record, generating a corresponding verification report, and based on the verification report, modifying the abnormal unfolding sequence, re-tracing and splicing the modified node sequence to generate a final executable path; storing the detailed node sequence and edge sequence of the final executable path, and outputting a complete path description containing all intermediate nodes.

[0013] In an implementation manner of the application, a local search operator is used to optimize the executable path in a neighborhood to reduce the total path cost, and when a termination condition is met, an optimized final path scheme and a cumulative cost report are output, specifically including: initializing local optimization iteration parameters and improvement thresholds, and using a path exchange operator to perform neighborhood search on the executable path to generate candidate path variants; calculating the total composite cost value of each candidate path variant, and comparing the cost difference between the candidate path variant and the current path, if an improved solution is found and the constraint condition is met, the candidate path variant is taken as a new current solution, and neighborhood search and acceptance judgment are repeatedly performed until the termination condition is met; generating a final optimized executable path sequence, calculating the cost components and the total composite cost of the final path, and outputting a comprehensive report; the comprehensive report includes the path sequence, the cost details of each dimension, and the constraint satisfaction situation.

[0014] On the other hand, the embodiments of the application also provide a multi-target path optimization device, the device comprising: at least one processor; and a memory connected with the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform a multi-objective path optimization method as described above.

[0015] In another aspect, the embodiments of the present application also provide a non-volatile computer storage medium, which stores computer executable instructions, and the computer executable instructions, when executed, implement a multi-objective path optimization method as described above.

[0016] The embodiments of the present application provide a multi-objective path optimization method, device and medium, which at least have the following beneficial effects: By constructing a weighted graph model and performing multi-dimensional verification, the accuracy and consistency of the path planning basic data can be ensured, and the problem of infeasible path or low efficiency caused by data abnormalities is eliminated from the source; based on the verified data, the all-source shortest path distance matrix and the predecessor record set are pre-calculated, and the connectivity detection and feasible region pruning are implemented, which significantly improves the efficiency and stability of path search, and avoids the calculation overhead and connectivity risk caused by repeated exploration of local path in traditional methods; by combining the nearest neighbor greedy strategy and the real-time composite cost accumulation mechanism, the system integrates the feasibility verification of multi-dimensional constraints such as time window, capacity and risk while dynamically generating an optimized access sequence, so that the path planning process is not only efficient, but also adaptive to complex environmental changes, overcoming the defects of single evaluation dimension and lack of dynamic adjustment capability in the prior art; by expanding the abstract sequence into an executable path of the original graph and applying a local search operator for neighborhood optimization, the total path cost is reduced, the practicality and quality of the path scheme are enhanced, and the dependence on manual parameter tuning is reduced through an automated process, improving the migratability and robustness of the method in different scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0017] The accompanying drawings, which are included to provide a further understanding of the present application, constitute a part of this application and illustrate embodiments of the present application and together with the description serve to explain the present application, but do not limit the present application in any way. In the drawings: Figure 1 A flowchart of a multi-objective path optimization method provided by the embodiments of the present application; Figure 2 An internal structure diagram of a multi-objective path optimization device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0020] Figure 1 This is a flowchart illustrating a multi-objective path optimization method provided in an embodiment of this application.

[0021] The analysis method involved in the embodiments of this application can be implemented by a terminal device or a server, and this application does not impose any special limitations on it. For ease of understanding and description, the following embodiments are all described in detail using a server as an example.

[0022] It should be noted that the server can be a single device or a system composed of multiple devices, i.e., a distributed server. This application does not make any specific limitations on this.

[0023] like Figure 1 As shown in the embodiment of this application, a multi-objective path optimization method includes: Step 101: Construct a weighted graph model of the area to be planned, represent the path cost relationship between nodes in the form of an adjacency matrix, and verify the symmetry, nonnegativity, dimensional consistency and connectivity of the adjacency matrix to generate the verified weighted graph data.

[0024] In this embodiment, the system acquires node data and edge data of the area to be planned. Node data includes location identifiers, used to uniquely identify the spatial or logical location of each node. For example, in a smart park scenario, the location identifier could be a device number, GPS coordinates, or building code. Edge data includes connection relationship identifiers, used to describe the topological connections between nodes, such as road numbers or passage names. This data can come from a geographic information system, building information model, or task planning database, and can be automatically imported through a data interface. Based on this data, the system constructs an initial adjacency matrix. The adjacency matrix is ​​a two-dimensional array structure, where the row and column indices correspond to node identifiers, and the element values ​​represent the edge weights between the corresponding nodes. It is understood that the edge weight is a composite metric, which can be calculated based on at least one of path distance, travel time, energy consumption, or risk assessment. For example, in a logistics and distribution scenario, the edge weight may simultaneously include transportation distance and estimated travel time, and a comprehensive cost value is obtained through weighted calculation.

[0025] The system performs a symmetry check on the initial adjacency matrix. The symmetry check is an important validation step for undirected graph structures, aiming to ensure consistency in bidirectional path costs between nodes. Specifically, the system traverses the non-diagonal elements of the matrix, comparing the weights of each pair of symmetric positions (e.g., W[i, j] and W[j, i]). If an asymmetry is detected, such as data inconsistency due to one-way traffic restrictions, the system automatically performs a correction process. Correction methods include taking the average, taking the maximum, or setting a uniform value based on business rules. For example, in a campus inspection path, if the round-trip time of a certain passage differs due to slope differences, the system can achieve symmetry by taking the average. After completing the symmetry check, the system performs a non-negativity check on the corrected adjacency matrix. The non-negativity check aims to eliminate negative weight edges, as negative weights can cause ambiguity in shortest path calculations. The system scans all edge weights, identifies values less than zero, which may be caused by data collection errors or transmission abnormalities, and replaces them with a pre-set positive weight (such as zero or the smallest reasonable positive value) or directly removes the edge. It should be noted that this processing ensures the mathematical rigor of subsequent algorithm execution.

[0026] The system performs a dimension consistency check on the non-negative adjacency matrix. Dimension consistency means that all edge weights should have a unified unit of measurement to avoid calculation bias caused by direct comparison of different dimensions. Specifically, the system identifies the dimension type of the edge weight, such as distance, time, or energy consumption, and performs unit conversion for inconsistent dimensions, such as converting meters to kilometers or seconds to minutes. After conversion, the system normalizes the weights to fall within a standard numerical range, eliminating the impact of magnitude differences on path evaluation. For example, in a multi-UAV coordination scenario, the system needs to unify battery consumption (milliampere-hours) and flight distance (meters) into dimensionless standardized values. Subsequently, the system performs a connectivity check on the adjacency matrix with consistent dimensions. Connectivity check is achieved through graph traversal algorithms such as depth-first search or breadth-first search, which traverse all reachable nodes from any node, identify and mark isolated nodes or edges, and generate a connected component report. The report describes the connected structure of the graph in detail, including the number of connected components, the nodes included in each component, and boundary information.

[0027] The system corrects abnormal edges based on the connected component report. Abnormal edges include missing edges, weight abnormal edges or logical conflict edges that cause the graph to be disconnected. The correction methods include automatically supplementing key edges (such as adding a virtual path to ensure connectivity), adjusting abnormal weights, or removing invalid edges. For example, in a digital twin park, if two areas are not connected due to modeling omissions, the system can automatically add a connected path based on the actual spatial location. After all corrections are completed, the system generates the final verified weighted graph data, which includes a normalized adjacency matrix, a node attribute table, and a verification log, providing high-quality input for subsequent all-source shortest path calculation. It can be understood that the multi-level verification mechanism of the embodiment fundamentally improves the quality of path planning data and lays a solid foundation for generating feasible and efficient path solutions.

[0028] Step 102, based on the verified weighted graph data, the shortest path algorithm is used to calculate the shortest path distance between all node pairs, forming an all-source shortest path distance matrix and a predecessor record set, and performing connectivity detection and feasible region pruning processing on unreachable node pairs.

[0029] In this embodiment, the system performs all-source shortest path calculation based on the verified weighted graph data. It can be understood that the all-source shortest path refers to the shortest path between any two nodes in the graph, and its distance information is stored in the distance matrix, and the path composition information is stored in the predecessor record set. Specifically, the system selects each node as a source point in turn and performs an improved Dijkstra algorithm. It should be noted that the improved Dijkstra algorithm introduces a priority queue optimization and an early termination mechanism based on the traditional algorithm. The priority queue uses a min-heap data structure to quickly locate the current shortest distance node, significantly reducing the time complexity. The early termination mechanism terminates the search in advance under certain conditions, such as terminating the calculation of the current source point when all target nodes have been processed. For example, in the intelligent park inspection scenario, taking a certain device monitoring point as the source point, the algorithm calculates the shortest distance from this point to all other monitoring points in the park and records the predecessor node information on the path. After the calculation of each source point is completed, the system obtains the shortest distance vector and the predecessor vector, where the distance vector stores the shortest distance from the source point to all nodes, and the predecessor vector stores the direct predecessor node identifier of each node on the shortest path.

[0030] Next, the system superimposes the calculation results of all source points to form a complete data structure. Specifically, the shortest distance vectors of each source point are combined by rows to construct a full-source shortest path distance matrix, which is a square matrix structure, with row and column indices corresponding to node numbers, and element values representing the shortest path distance between corresponding node pairs. At the same time, the predecessor vectors of all source points are integrated to form a predecessor record set, which is a three-dimensional data structure, with the first dimension index representing the source point, the second dimension index representing the target point, and the stored value being the predecessor node on the shortest path from the source point to the target point. It should be noted that this structured storage method allows the shortest path between any node pair to be quickly reconstructed by backtracking the predecessor records. For example, in a logistics distribution path planning, the distance matrix can quickly query the shortest driving distance between any two distribution points, and the predecessor record set can restore the specific driving route.

[0031] Then, the system performs connectivity detection and feasible region pruning. Specifically, the full-source shortest path distance matrix is detected for unreachable node pairs, and the system scans all elements in the distance matrix to identify node pairs with infinite distance values, which represent pairs of points that do not have a connected path in the original graph. After detection, an unreachable marking data is generated, and all unreachable node pairs and their associated information are recorded in a list form. For example, in a multi-UAV cooperative inspection scenario, due to building obstructions or flight restricted areas, some monitoring points may not be directly accessible, and these point pairs will be marked as unreachable.

[0032] Based on the unreachable marking data, the system performs connectivity analysis to determine the maximum connected subgraph by graph traversal algorithm. Specifically, the depth-first search or breadth-first search algorithm is used to traverse all reachable nodes from any node, identify connected components in the graph, and select the connected component with the most nodes as the maximum connected subgraph. Then, feasible region pruning is performed to automatically exclude non-connected nodes, which are usually located in isolated areas or cannot access the main network due to path network defects. For example, in a warehouse logistics environment, if a shelf area becomes isolated due to temporary obstacles, the system will exclude the nodes in that area from the planning range. After pruning, a pruned node set is generated, which only contains all nodes in the maximum connected subgraph.

[0033] Finally, the system updates the all-source shortest path metric data based on the pruned node set. Specifically, the distance matrix is updated, the rows and columns related to the pruned nodes are removed, a distance sub-matrix containing only the connected nodes is reconstructed, and the predecessor record set is updated synchronously, with the records related to the pruned nodes being deleted, to ensure the consistency of the data structure. It can be understood that this dynamic updating mechanism ensures that subsequent path planning is only performed in the feasible connected area, avoiding invalid calculations in unreachable areas. After generating the revised all-source shortest path metric data, the system also pre-generates multi-dimensional cost tensors such as time cost and energy cost, and aligns them with the shortest path metric, providing complete data support for the subsequent composite cost evaluation in the path construction stage. It should be noted that through the comprehensive processing of the present embodiment, the system establishes a stable and reliable path metric space, laying a solid foundation for generating high-quality path solutions.

[0034] Step 103: Based on the processed all-source shortest path metric data and the nearest neighbor greedy strategy, an optimized access sequence covering all target nodes is generated, and the composite path cost is accumulated in real time at each step, and the feasibility test of time window, capacity and risk constraints is performed, generating an initial path sequence and cumulative cost data.

[0035] In the present embodiment, the system first selects a path starting node. The starting point selection can use fixed rules (such as specifying the warehouse location), heuristic rules (such as selecting the central node), or multi-start point parallel exploration mechanisms. After selecting the starting point, the system initializes the core state variables required for path construction, sets the current node as the starting point, establishes a set of unvisited nodes containing all nodes to be visited, initializes an empty path sequence, and sets the cumulative cost variable to zero. Illustratively, in the park inspection scenario, the starting point can be set as the main control center, the unvisited node set contains all equipment points that need to be inspected, and the path sequence initially contains only the main control center node.

[0036] In each iteration, based on the all-source shortest path distance matrix, the system selects the node with the shortest distance from the current node as the next node from the unvisited node set. Specifically, the system queries the row vector corresponding to the current node in the distance matrix, filters out the nodes belonging to the unvisited set, compares the distance values of these nodes and selects the node corresponding to the minimum value, thereby generating nearest neighbor selection data, which records the current node, the selected next node and their distance information. It should be noted that this selection mechanism based on pre-computed distance matrix avoids the overhead of online path calculation, significantly improving the construction efficiency. Based on the nearest neighbor selection data, the system updates the path sequence, appends the next node to the end of the existing sequence to generate updated path sequence data. At the same time, the system updates the current node state and the unvisited node set to ensure the correct advancement of the iteration state.

[0037] After each step of node selection, the system calculates the composite cost from the current node to the next node based on a composite cost function. The composite cost function is a multi-dimensional evaluation model that integrates multiple factors such as distance cost, time cost, energy consumption cost, and risk penalty cost. It can be understood that each cost component is combined through a weighting coefficient, and the weight parameter can be dynamically configured according to the specific scene. Exemplarily, in the logistics distribution scene, the distance cost reflects the transportation mileage, the time cost considers the traffic conditions, the energy consumption cost reflects the vehicle power consumption, and the risk cost evaluates the road safety level. The system accumulates the calculated composite cost into the total cost variable in real time to generate cumulative cost data containing detailed dimensions.

[0038] At the same time, the system performs feasibility verification of multi-dimensional constraints. The time window conflict detection verifies whether the node access time is within the preset time range, and generates a detailed constraint verification report recording the satisfaction of each constraint condition and its deviation degree. For example, in express delivery, the receiver's acceptable time period is checked; the capacity overrun check evaluates whether the resource consumption exceeds the limit, such as whether the AGV power consumption exceeds the battery capacity; the risk area avoidance identifies whether the node is located in a forbidden area or a high-risk area. Specifically, the system dynamically adjusts the path construction strategy according to the verification report. For minor constraint conflicts, the composite cost can be adjusted by adding a penalty term; for serious conflicts, the next node needs to be reselected or the access order needs to be adjusted. Exemplarily, when a delivery point is detected to exceed the time window, the system can automatically delay its access order or adjust the driving speed to meet the time requirement.

[0039] The system repeatedly performs the above iteration process, completing node selection, cost accumulation, and constraint verification in each round until the set of unvisited nodes is empty. Finally, the initial path sequence after feasibility processing and the complete cumulative cost data are generated. It should be noted that this method of combining greedy construction and real-time verification not only guarantees the efficiency of path generation, but also ensures the feasibility of the scheme in a complex constraint environment. It can be understood that through the implementation of this embodiment, the system can quickly generate an optimized access sequence covering all target nodes while considering multi-dimensional actual constraints.

[0040] Step 104, when all target nodes are visited, the adjacent node pairs in the initial path sequence are expanded into specific executable paths on the original graph according to the predecessor record set, and a local search operator is used to optimize the executable paths in the neighborhood to reduce the total path cost. When the termination condition is met, the optimized final path scheme and cumulative cost report are output.

[0041] In this embodiment, when all target nodes are visited, the system obtains adjacent node pair data in the initial path sequence, which records the adjacent node combinations in the visiting sequence in the form of ordered pairs. Based on the predecessor record set, the system performs path backtracking for each adjacent node pair to find the intermediate node sequence through which the shortest path on the original graph passes. Illustratively, in the AGV navigation scenario, if the initial sequence contains nodes A→B→C, the system needs to expand the specific paths of A→B and B→C respectively. Specifically, for a node pair (u, v), the system first finds the predecessor node of v corresponding to the source point u in the predecessor set, and then continues to query the predecessor node to the predecessor node of u until it backtracks to the starting point u, and finally obtains the complete intermediate node sequence.

[0042] In the splicing process, the system automatically identifies and removes duplicate nodes between adjacent sequences, such as the endpoint of the previous path and the starting point of the next path being the same node, to ensure the generation of continuous preliminary executable paths. It should be noted that the de-duplication process can avoid redundant nodes in the path, ensuring the simplicity and execution efficiency of the path. Subsequently, the system performs smoothness checking on the preliminary executable path to detect and eliminate unnecessary turning points or stopping points, generating an optimized executable path. It can be understood that in the mobile robot navigation scenario, path smoothness directly affects the stability and energy efficiency of motion control.

[0043] Then, the system verifies the connectivity and effectiveness of the executable path in the original graph. Connectivity verification can ensure that all adjacent nodes are indeed reachable in the actual road network, and effectiveness verification checks whether the path meets basic geometric and kinematic constraints. During the verification process, a detailed verification report is generated, recording the verification results of each path segment. Based on the verification report, the system automatically corrects abnormal expansion sequences. For connectivity-abnormal segments, the system reselects alternative paths for backtracking; for effectiveness-abnormal segments, the system adjusts the intermediate node sequence or inserts auxiliary nodes. Illustratively, in the UAV inspection path, if a segment of the path is blocked by an obstacle and cannot be traveled, the system will automatically plan a detour path. After the correction is completed, the system re-performs backtracking and splicing to generate the final executable path and store its detailed node sequence and edge sequence information.

[0044] After obtaining the executable path, the system first initializes the local optimization parameters and uses the path exchange operator to perform neighborhood search on the executable path to generate candidate path variants. It should be noted that the local optimization parameters include control parameters such as the maximum number of iterations, the improvement threshold, and the time upper limit. Specifically, the 2-opt operator reconstructs the path sequence by exchanging two edges in the path, and the Or-opt operator adjusts the path structure by moving consecutive node blocks. Illustratively, in the logistics distribution path optimization, the 2-opt operation may exchange the connection order between two distribution points, thereby shortening the total travel distance.

[0045] For each candidate path variant, the system calculates its total composite cost value, taking into account multiple dimensions such as distance, time, energy consumption, and risk. By comparing the cost difference between the candidate path and the current path, the system adopts an improved acceptance strategy, and if an improved solution is found and all constraint conditions are met, the candidate path is taken as the new current solution. It can be understood that this acceptance strategy ensures that the optimization process is always directed towards reducing the cost. The system repeatedly performs neighborhood search and acceptance judgment until the termination condition is met, such as reaching the maximum number of iterations or not finding an improved solution for a certain number of generations.

[0046] Finally, the system generates the final optimized executable path sequence and calculates the cost components and total composite cost in each dimension. The output comprehensive report includes a complete path sequence description, detailed cost breakdown in each dimension (such as distance, time, energy consumption, etc.), and a summary of constraint satisfaction. It should be noted that this multi-dimensional report output provides a comprehensive basis for the evaluation and execution of the path scheme. Exemplarily, in the smart park inspection system, the output report not only includes the inspection route, but also details the estimated time consumption, energy consumption statistics, and risk assessment, providing data support for operation and maintenance decision-making, and realizing the complete transformation from abstract sequence to executable path.

[0047] The above is the method embodiment of the present application. Based on the same inventive concept, the present application also provides a multi-objective path optimization device, the structure of which is as shown in Figure 2 .

[0048] Figure 2 The internal structure of a multi-objective path optimization device provided by the present application is shown in Figure 2 . The device includes: at least one processor; and a memory in communication connection with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to: build a weighted graph model of the region to be planned, represent the path cost relationship between nodes in the form of an adjacency matrix, and perform symmetry, non-negativity, dimension consistency, and connectivity verification on the adjacency matrix to generate verified weighted graph data; based on the verified weighted graph data, use the shortest path algorithm to calculate the shortest path distance between all node pairs, form a full-source shortest path distance matrix and a predecessor record set, and perform connectivity detection and feasible region pruning processing on unreachable node pairs; Based on the processed all-source shortest path metric data and the nearest neighbor greedy strategy, an optimized access sequence covering all target nodes is generated, and the composite path cost is accumulated in real time after each step selection, and the feasibility test of time window, capacity and risk constraints is performed, to generate an initial path sequence and accumulated cost data; When all target nodes are accessed, the adjacent node pairs in the initial path sequence are unfolded into specific executable paths on the original graph according to the predecessor record set, and a local search operator is used to optimize the executable paths in the neighborhood to reduce the total path cost, and when the termination condition is met, the optimized final path scheme and the accumulated cost report are output.

[0049] The embodiments of the present application also provide a non-volatile computer storage medium, which stores computer executable instructions, and the computer executable instructions can perform the following when executed: A weighted graph model of the region to be planned is constructed to represent the path cost relationship between nodes in the form of an adjacency matrix, and the adjacency matrix is checked for symmetry, non-negativity, dimensional consistency and connectivity to generate the checked weighted graph data; Based on the checked weighted graph data, a shortest path algorithm is used to calculate the shortest path distance between all node pairs to form an all-source shortest path distance matrix and a predecessor record set, and the connectivity detection and feasible region pruning processing are performed on the unreachable node pairs; Based on the processed all-source shortest path metric data and the nearest neighbor greedy strategy, an optimized access sequence covering all target nodes is generated, and the composite path cost is accumulated in real time after each step selection, and the feasibility test of time window, capacity and risk constraints is performed, to generate an initial path sequence and accumulated cost data; When all target nodes are accessed, the adjacent node pairs in the initial path sequence are unfolded into specific executable paths on the original graph according to the predecessor record set, and a local search operator is used to optimize the executable paths in the neighborhood to reduce the total path cost, and when the termination condition is met, the optimized final path scheme and the accumulated cost report are output.

[0050] Each of the embodiments in the present application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment mainly explains the difference from other embodiments. Especially, the device and medium embodiments are described simply because they are basically similar to the method embodiments, and the relevant parts can be referred to the part of the method embodiments.

[0051] The device and medium provided by the embodiments of the present application are one-to-one corresponding to the method, so the device and medium also have the similar beneficial technical effects as the method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the device and medium will not be described here.

[0052] Those skilled in the art will appreciate that embodiments of the present application can be readily used as software, hardware, or a combination of software and hardware. In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0053] The present application is described in reference to the flowchart illustrations and / or block diagrams according to the embodiments of the application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processing system, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams. Figure 1 means for carrying out one or more functions specified in the flowchart illustrations and / or block diagrams.

[0054] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams. Figure 1 means for carrying out one or more functions specified in the flowchart illustrations and / or block diagrams.

[0055] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams. Figure 1 means for carrying out one or more functions specified in the flowchart illustrations and / or block diagrams.

[0056] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0057] The memory can include non-persistent memory, random access memory (RAM), and / or non-volatile memory, such as read only memory (ROM) or flash memory, among others. The memory is an example of computer-readable media.

[0058] Computer-readable media includes permanent and non-permanent, movable and non-movable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible by a computing device. According to the definition herein, computer-readable media does not include transitory media such as modulated data signals and carriers.

[0059] It should also be noted that the terms "comprising", "containing", or any other variant thereof are intended to cover non-exclusive inclusions, so that a process, method, article or apparatus that includes a list of elements does not only include those elements, but also includes other elements not explicitly listed, or further includes elements inherent in such a process, method, article or apparatus. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus that includes the element.

[0060] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application shall be included in the scope of claims of the present application.

Claims

1. A multi-objective path optimization method, characterized in that, The method comprises: Construct a weighted graph model of the region to be planned to represent the path cost relationship between nodes in the form of an adjacency matrix, and perform symmetry, non-negativity, dimensional consistency and connectivity checks on the adjacency matrix to generate the checked weighted graph data; Based on the checked weighted graph data, use the shortest path algorithm to calculate the shortest path distance between all node pairs to form a full-source shortest path distance matrix and a predecessor record set, and perform connectivity detection and feasible region pruning processing on unreachable node pairs; Based on the processed full-source shortest path metric data and the nearest neighbor greedy strategy, generate an optimized access sequence covering all target nodes, and at each step, accumulate the composite path cost in real time after selection, and perform feasibility verification of time window, capacity and risk constraints to generate an initial path sequence and accumulated cost data; When all target nodes have been accessed, expand the adjacent node pairs in the initial path sequence into specific executable paths on the original graph according to the predecessor record set, and use local search operators to optimize the executable paths in the neighborhood to reduce the total path cost, and output the optimized final path scheme and accumulated cost report when the termination condition is met.

2. The multi-objective path optimization method of claim 1, wherein, Construct a weighted graph model of the region to be planned to represent the path cost relationship between nodes in the form of an adjacency matrix, and perform symmetry, non-negativity, dimensional consistency and connectivity checks on the adjacency matrix to generate the checked weighted graph data, specifically comprising: Obtain node data and edge data of the region to be planned, and based on the node data and edge data, construct an adjacency matrix; wherein the node data includes a location identifier, the edge data includes a connection relationship identifier, the element value of the adjacency matrix represents the edge weight value between the corresponding nodes, and the edge weight value in the adjacency matrix is calculated based on at least one of path distance, time, energy consumption or risk; Perform symmetry checking on the adjacency matrix to correct the asymmetric edge weight values found into symmetric values, and perform non-negativity checking on the symmetric adjacency matrix to replace the negative weight edges found with preset positive weight values; Perform dimensional consistency detection on the non-negative adjacency matrix, perform unit conversion on the edge weight values with inconsistent dimensions, and perform connectivity checking on the adjacency matrix with consistent dimensions to identify and mark isolated nodes or edges through a graph traversal algorithm to generate a connected component report; Based on the connected component report, modify the abnormal edges, and generate the final checked weighted graph data.

3. The multi-objective path optimization method of claim 1, wherein, Based on the checked weighted graph data, use the shortest path algorithm to calculate the shortest path distance between all node pairs to form a full-source shortest path distance matrix and a predecessor record set, specifically comprising: Select each node as a source point in turn, execute the improved Dijkstra algorithm to calculate the shortest distance vector and predecessor vector from the source point to all other nodes; Superimpose the shortest distance vectors of all source points to form a full-source shortest path distance matrix, and integrate the predecessor vectors of all source points to form a predecessor record set.

4. The multi-objective path optimization method of claim 1, wherein, Perform connectivity detection and feasible region pruning processing on unreachable node pairs, specifically comprising: Performing unreachable node pair detection on the all-pairs shortest path distance matrix, identifying node pairs with infinite distance values, and generating unreachable label data; Based on the unreachable label data, performing connectivity analysis, determining the maximum connected subgraph through graph traversal algorithm, and performing feasible region pruning processing to eliminate non-connected nodes, generating pruned node set; Based on the pruned node set, updating the all-pairs shortest path distance matrix and the predecessor record set, and generating corrected all-pairs shortest path metric data.

5. The multi-objective path optimization method of claim 1, wherein, Based on the processed all-pairs shortest path metric data and the nearest neighbor greedy strategy, an optimized access sequence covering all target nodes is generated, specifically including: Selecting a path starting node and initializing current node state, unvisited node set, path sequence, and cumulative cost variable; In each iteration, based on the all-pairs shortest path distance matrix, the node with the shortest distance from the current node is selected as the next node from the unvisited node set, generating nearest neighbor selection data; Based on the nearest neighbor selection data, update the path sequence by appending the next node to the path sequence, and generate updated path sequence data.

6. The multi-objective path optimization method of claim 1, wherein, Real-time cumulative composite path cost after each selection, and feasibility check of time window, capacity and risk constraints, generating initial path sequence and cumulative cost data, specifically including: Based on the composite cost function, calculate the composite cost from the current node to the next node and accumulate it into the total cost variable, generating cumulative cost data; the composite cost function integrates distance cost, time cost, energy cost and risk penalty cost; Based on the time window, capacity and risk constraints, the feasibility of the next node is checked, and a constraint check report is generated; the feasibility check includes time window conflict detection, capacity overrun check and risk area avoidance; According to the constraint check report, adjust the path sequence or composite cost parameters, generate the initial path sequence and cumulative cost data after feasibility processing, and repeat the iteration process until the unvisited node set is empty, complete the path sequence generation.

7. The multi-objective path optimization method of claim 1, wherein, When all target nodes are visited, the adjacent node pairs in the initial path sequence are unfolded into specific executable paths on the original graph according to the predecessor record set, specifically including: Obtain adjacent node pair data in the initial path sequence, and based on the predecessor record set, find the intermediate node sequence through which each adjacent node pair passes on the original graph; Splice the intermediate node sequence between all node pairs, remove duplicate nodes after splicing, generate preliminary executable paths, and perform smoothness check on the preliminary executable paths, generate optimized executable paths; Verify the connectivity and effectiveness of the executable paths in the original graph, ensure that all node pairs are correctly unfolded based on the predecessor record, generate a corresponding verification report, and based on the verification report, modify the abnormal unfolding sequence, retrace and splice the modified node sequence, generate the final executable path; Store the detailed node sequence and edge sequence of the final executable path, and output the complete path description containing all intermediate nodes.

8. The multi-objective path optimization method of claim 1, wherein, The local search operator is used to optimize the executable path in a neighborhood to reduce the total path cost, and an optimized final path scheme and cumulative cost report are output when a termination condition is met, specifically including: Initializing local optimization iteration parameters and improvement thresholds, and using a path exchange operator to perform neighborhood search on the executable path to generate candidate path variants; Calculating the total composite cost of each candidate path variant, and comparing the cost difference between the candidate path variant and the current path. If an improved solution is found and the constraint condition is met, the candidate path variant is taken as a new current solution, and neighborhood search and acceptance judgment are repeatedly performed until the termination condition is met; Generating a final optimized executable path sequence, calculating the dimensional cost components and total composite cost of the final path, and outputting a comprehensive report. The comprehensive report includes the path sequence, dimensional cost details, and constraint satisfaction.

9. A multi-objective path optimization device, comprising: The device comprises: at least one processor; and a memory connected in communication with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform a multi-objective path optimization method according to any one of claims 1-8.

10. A non-transitory computer storage medium storing computer-executable instructions that, when executed, cause a computer to perform: The computer executable instructions, when executed, implement a multi-objective path optimization method according to any one of claims 1-8.

Images