Vehicle path planning method based on adaptive optimization algorithm

Abstract

The present invention provides a vehicle path planning method based on an adaptive optimization algorithm, relating to the field of adaptive algorithm application technology. The method comprises: step 1, based on current order demand, vehicle resources, and preset constraints, generating an initial path plan set that meets time window, number of splits, vehicle model matching, and temperature zone requirements through a step-by-step progressive decision-making process; step 2, based on the initial path plan set, initiating multiple independent optimization links, each of which dynamically adjusts local path nodes using a greedy strategy to generate a set of optimized candidate plans; step 3, evaluating the set of optimized candidate plans in real time, determining a final candidate plan based on cost savings rate, demand satisfaction rate, and soft service quality indicators, and updating the final candidate plan to the current baseline plan. The present invention uses an adaptive optimization algorithm to perceive dynamic events in real time, assess their impact, and intelligently adjust the path, thereby improving the timeliness and accuracy of path planning.

Description

Technical Field

[0001] The present invention relates to the technical field of adaptive algorithm application, and in particular to a vehicle path planning method based on an adaptive optimization algorithm. Background Art

[0002] While traditional vehicle routing technology can achieve basic route design for school shuttles during fixed hours, it still leaves room for improvement when dealing with dynamic variables. Traditional technology relies on fixed reservation information for route planning, making it difficult to quickly adapt to changing ridership demand. For example, during the morning rush hour on a Friday, the pre-set list included 120 junior high students. However, 15 minutes before departure, 23 students canceled their rides due to parents choosing to drive themselves. At the same time, 18 elementary school students living farther away had to join the shuttle bus queue due to their parents' unexpected work arrangements. Due to the lack of real-time data access, traditional planning methods often rely solely on the original route. This results in a large number of empty seats at the original junior high stop, while the waiting areas for newly requested elementary students are not included in the route. This ultimately leads to the dual problems of wasted transportation resources and extended waiting times for some students.

[0003] Furthermore, faced with the regular morning and evening rush hour congestion around schools, as well as occasional temporary traffic controls, traditional route optimization strategies are relatively rigid. For example, during the Wednesday evening rush hour, a shuttle bus was traveling along its scheduled route to an intersection 3 kilometers from the school when a sudden traffic accident caused a one-way road closure. This intersection, a major thoroughfare connecting to student housing, was closed. Traditional planning techniques rely on preset road lengths and average travel times, lacking real-time access to dynamic road closure information from traffic authorities and unable to dynamically adjust parameters based on real-time congestion data. In this situation, the shuttle bus might continue toward the closed intersection until it was near, only to be forced to detour. During this detour, the final stop order for surrounding communities was not recalculated, causing the original 40-minute commute to extend to 75 minutes. Some students arrived home nearly half an hour later than usual, and the detour route included repeated detours, increasing vehicle energy consumption and operating costs. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a vehicle path planning method based on an adaptive optimization algorithm, which can respond to the dynamic riding demand and road condition changes in school buses in real time and realize dynamic path adjustment.

[0005] In order to solve the above technical problems, the technical solutions of the present invention are as follows:

[0006] In a first aspect, a vehicle path planning method based on an adaptive optimization algorithm comprises:

[0007] Step 1: Based on current order demand, vehicle resources, and pre-set constraints, a step-by-step progressive decision-making process is used to generate an initial set of routing solutions that meet time window, number of splits, vehicle type matching, and temperature zone requirements.

[0008] Step 2: Based on the initial path solution set, multiple independent optimization links are started. Each link dynamically adjusts the local path nodes through a greedy strategy to generate a set of optimization candidate solutions.

[0009] Step 3: Conduct real-time evaluation of the optimized candidate solution set, determine the final candidate solution based on cost savings rate, demand satisfaction rate, and soft service quality indicators, and update the final candidate solution to the current baseline solution;

[0010] Step 4: According to the current baseline solution, when a dynamic event is detected within the closed triangle area formed by the order distribution center detection point, the vehicle dispatch center detection point, and the transportation hub detection point, local link optimization is triggered and dynamic correction parameters are generated based on the real-time position relationship between the triangle vertices;

[0011] Step 5: Calculate the geometric center of gravity offset of the affected path segment based on the dynamic correction parameters, perform directional adjustments on the offset path segment, and then update the global solution library to generate a path adjustment result covering the dynamic area.

[0012] Step 6: Set a solution survival threshold based on the updated global solution library. When the solution optimization period is greater than or equal to the threshold, eliminate the old solution based on the dynamic regional topology association characteristics, and generate an alternative optimization link based on the dynamic correction parameters.

[0013] Furthermore, based on current order demand, vehicle resources, and pre-set constraints, a step-by-step progressive decision-making process is used to generate a set of initial routing solutions that meet the time window, number of splits, vehicle model matching, and temperature zone requirements, including:

[0014] Based on the temperature zone attributes and time window restrictions of the orders, multi-dimensional clustering is performed on the current order demand to generate multiple order subsets with the same temperature zone and time window overlap ≥ the preset threshold;

[0015] Based on the order subset, the vehicle resource library is traversed for models that meet the temperature zone adaptation conditions. The candidate vehicle set whose vehicle capacity covers the total order demand in the subset is determined. The order subset with a capacity greater than the single load capacity is split, ensuring that the number of splits is ≤ the preset threshold. The task set to be assigned is generated, including the split order units.

[0016] Match the set of tasks to be assigned with the candidate vehicle set, and generate initial path segments that meet the vehicle model loading capacity constraints based on the distance priority between the vehicle's real-time location and the order delivery point;

[0017] Perform time window conflict detection on the initial path segments. If a time window conflict exists, dynamically adjust the node order of the path segments based on the vehicle speed to generate a set of conflict-free feasible path segments.

[0018] The paths in the same temperature zone and geographically adjacent in the set of feasible path segments are combined and optimized to generate a set of initial path solutions including multi-vehicle collaborative scheduling and multi-order consolidated delivery.

[0019] Furthermore, based on the initial path solution set, multiple independent optimization links are started. Each link dynamically adjusts the local path nodes through a greedy strategy to generate a set of optimization candidate solutions, including:

[0020] Nodes whose path cost ratio is greater than a preset ratio are extracted from the initial path solution set as highly sensitive adjustment objects to generate a candidate node set;

[0021] For different optimization links, based on the distance deviation between the candidate node set and the vehicle's current position, differentiated node adjustment priorities are assigned to each link, generating a link-specific local adjustment range.

[0022] Within the local adjustment range, greedy operations are performed on highly sensitive nodes. If the time window margin of the node's path is sufficient, an attempt is made to insert adjacent unassigned orders. If the time window is tight, an exchange is performed with adjacent path nodes to generate candidate sub-plans within the link.

[0023] The candidate sub-schemes generated by each link are verified for conflicts, the schemes that conflict with the vehicle temperature zone adaptability are eliminated, and the remaining sub-schemes are merged into a set of optimized candidate schemes.

[0024] Furthermore, the different optimization links include a first link that dynamically calculates the deviation amplitude difference based on the real-time distance deviation between the candidate node and the current vehicle position to generate a distance-sensitive adjustment priority sequence; and a second link that quantifies the path interaction intensity based on the geographical overlap rate between the path where the candidate node is located and the adjacent path to generate an overlap-sensitive adjustment priority sequence.

[0025] Furthermore, the set of optimized candidate solutions is evaluated in real time, and the final candidate solution is determined based on cost savings rate, demand satisfaction rate, and soft service quality indicators. The final candidate solution is then updated to the current baseline solution, including:

[0026] For each plan in the set of optimized candidate plans, the difference between the total transportation cost and the historical benchmark plan is calculated, and the plan with a cost saving rate ≥ the preset threshold is determined to form a set of primary cost reduction plans;

[0027] Verify the demand coverage of the primary fee reduction plan set to check whether each plan covers all key order nodes. If there are any urgent orders that are not covered, mark them as demand-missing plans and eliminate them, generating a secondary demand-compliant plan set.

[0028] Evaluate the service quality of the secondary demand compliance solution set, calculate the comprehensive service quality score based on the order delivery on-time rate and the frequency of route adjustment, and generate a priority queue for the solutions based on the score;

[0029] Determine the candidate solutions with the final ranking in comprehensive scores from the solution priority queue, and simultaneously activate the compensation mechanism for unmet demand. Re-inject the unallocated orders from the eliminated demand-missing solutions into the initial path solution set to generate a pool of orders to be reallocated.

[0030] The final candidate solution is updated to the current baseline solution, and the path replanning link of the order pool to be reallocated is triggered.

[0031] Furthermore, the service quality of the secondary demand compliance solution set is evaluated. Based on the order delivery on-time rate and the frequency of route adjustment, a comprehensive service quality score is calculated. The solution priority queue is generated by sorting by score, including:

[0032] Extract the matching deviation between the actual delivery time of orders in each plan and the preset time window, use the ratio of the total deviation time to the total number of orders as the benchmark indicator, and generate the on-time rate ranking sequence of each plan;

[0033] Based on the punctuality ranking sequence, the number of node order adjustments triggered by each solution during the path optimization process is counted to generate an adjustment frequency sequence associated with the punctuality.

[0034] Based on the punctuality ranking sequence and the adjustment frequency sequence, determine the solutions where the total punctuality deviation is less than the preset upper limit and the adjustment frequency is less than the preset limit, forming a preliminary set of preferred solutions;

[0035] For the solutions in the preliminary preferred solution set, they are first sorted in ascending order according to the total deviation of punctuality. If the total deviation is the same, they are sorted again in ascending order according to the adjustment frequency, and the number of unserviced orders in the path coverage area is extracted. The final solution priority queue is generated by sorting in ascending order according to the number of unserviced orders.

[0036] Furthermore, according to the current benchmark solution, when a dynamic event is detected within the closed triangle area formed by the order distribution center detection point, the vehicle dispatch center detection point, and the transportation hub detection point, local link optimization is triggered, and dynamic correction parameters are generated based on the real-time position relationship between the triangle vertices, including:

[0037] The position data of three detection points within the triangular area are collected in real time. If the offset between the real-time coordinates of any detection point and the reference position is greater than the preset tolerance threshold, it is determined to be a dynamic disturbance event and the coordinate sequence is extracted;

[0038] Based on the coordinate sequence, the real-time change rate of each side length of the triangle and the dynamic offset angle of the vertex angle are calculated, and the dynamic event impact intensity index is generated in combination with the current path node distribution density;

[0039] According to the dynamic event impact intensity index and the path node distribution density of the area passed through in the current benchmark scheme, the path redirection priority and detour compensation distance are dynamically calculated to generate dynamic correction parameters.

[0040] Furthermore, based on the dynamic correction parameters, the geometric center of gravity offset of the affected path segment is calculated, and the global solution library is updated after directional adjustment of the offset path segment to generate a path adjustment result covering the dynamic area, including:

[0041] Based on the path redirection priority in the dynamic correction parameters, the affected path segments within the triangle area are extracted, and the offset between the geometric center of gravity and the preset reference position is calculated;

[0042] Performing a directional adjustment operation on the affected path segment according to the offset direction and the detour distance compensation value to obtain an adjusted path segment;

[0043] Perform constraint penetration verification on the adjusted path segments to check whether the detour nodes are greater than the vehicle temperature zone adaptation range, and whether the path expansion violates the limit on the number of splits;

[0044] The verified adjusted path segments are spliced with the unaffected paths, the global solution library is updated, and the path adjustment results with dynamic area coverage are generated.

[0045] Furthermore, a solution survival threshold is set based on the updated global solution library. When the solution optimization period is greater than or equal to the threshold, the old solution is eliminated based on the dynamic regional topology association characteristics, and an alternative optimization link is generated based on the dynamic correction parameters, including:

[0046] Adjust the result stability based on the frequency and path of dynamic events, dynamically calculate the solution survival threshold, and compare it with the current solution optimization cycle;

[0047] If the optimization period is greater than or equal to the threshold, the matching degree between the path node and the current area center offset is extracted, and the solutions with matching degree less than the preset critical value are eliminated;

[0048] Based on the detour compensation distance and priority classification rules set in the dynamic correction parameters, alternative optimized paths are dynamically generated, including prioritizing the reuse of verified detour path segments and achieving smooth connection with the current vehicle's real-time position; if there are no available detour segments, a set of alternative nodes is generated according to the detour compensation distance threshold for path reconstruction.

[0049] In a second aspect, a computer-readable storage medium stores a program, which implements the method when executed by a processor.

[0050] The above solution of the present invention includes at least the following beneficial effects:

[0051] By dynamically adjusting parameters in real time through an adaptive optimization algorithm, it can quickly adapt to complex and changing traffic conditions and order demands. Compared to traditional algorithms, this method can improve path planning efficiency by over 30%, shorten calculation time, and improve the real-time and accuracy of path planning, ensuring that delivery tasks are completed on time. This method comprehensively considers multiple cost factors such as vehicle load, driving distance, and fuel consumption. By optimizing the path, it reduces invalid driving mileage, fuel consumption, and vehicle wear and tear. The adaptive mechanism enables the algorithm to automatically adjust the search strategy based on actual conditions. Whether facing sudden traffic control, temporary additional orders, or unexpected vehicle failures, it can quickly re-plan the path to ensure the stable operation of the delivery network, greatly enhancing the path planning system's adaptability to complex scenarios.

[0052] Fully utilize the carrying capacity of each vehicle, avoid empty or overloaded vehicles, improve vehicle utilization, and achieve rational resource allocation, thereby reducing vehicle investment while ensuring delivery service quality. Efficient and accurate route planning ensures on-time and accurate delivery of goods, reduces delivery delays and errors, and improves customer satisfaction with logistics services. The extensive data accumulated during route planning provides rich operational information, which can be used through data analysis to identify potential problems and optimization opportunities, and formulate long-term development strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a flowchart of a vehicle path planning method based on an adaptive optimization algorithm provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0054] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0055] like Figure 1 As shown, an embodiment of the present invention proposes a vehicle path planning method based on an adaptive optimization algorithm, the method comprising the following steps:

[0056] Step 1: Based on current order demand, vehicle resources, and pre-set constraints, a step-by-step progressive decision-making process is used to generate an initial set of routing solutions that meet time window, number of splits, vehicle type matching, and temperature zone requirements.

[0057] Step 2: Based on the initial path solution set, multiple independent optimization links are started. Each link dynamically adjusts the local path nodes through a greedy strategy to generate a set of optimization candidate solutions.

[0058] Step 3: Conduct real-time evaluation of the optimized candidate solution set, determine the final candidate solution based on cost savings rate, demand satisfaction rate, and soft service quality indicators, and update the final candidate solution to the current baseline solution;

[0059] Step 4: According to the current baseline solution, when a dynamic event is detected within the closed triangle area formed by the order distribution center detection point, the vehicle dispatch center detection point, and the transportation hub detection point, local link optimization is triggered and dynamic correction parameters are generated based on the real-time position relationship between the triangle vertices;

[0060] Step 5: Calculate the geometric center of gravity offset of the affected path segment based on the dynamic correction parameters, perform directional adjustments on the offset path segment, and then update the global solution library to generate a path adjustment result covering the dynamic area.

[0061] Step 6: Set a solution survival threshold based on the updated global solution library. When the solution optimization period is greater than or equal to the threshold, eliminate the old solution based on the dynamic regional topology association characteristics, and generate an alternative optimization link based on the dynamic correction parameters.

[0062] In this embodiment of the present invention, a step-by-step, progressive decision-making process is used to generate an initial set of routing solutions based on order requirements, vehicle resources, and pre-set constraints. This ensures that the solutions precisely meet time window, number of splits, vehicle type matching, and temperature zone requirements. This process ensures the compliance and feasibility of routing plans, avoids resource waste and delivery delays caused by solutions being out of sync with actual operating conditions, and improves the initial quality and reliability of the solutions. Multiple independent optimization links are initiated, and a greedy strategy is used to dynamically adjust local path nodes to generate a set of optimized candidate solutions. Parallel processing improves optimization efficiency, and the greedy strategy can quickly capture local final solutions and explore potential optimization opportunities within the initial solutions. Optimizing the path from multiple perspectives across different links increases the diversity of candidate solutions, providing a richer selection for obtaining optimal routing solutions. The set of optimized candidate solutions is evaluated in real time based on cost savings, demand fulfillment, and soft service quality indicators. The final candidate solution is determined and updated as the current baseline solution. This evaluation system considers cost, demand, and service quality to ensure that the solution is the final solution with comprehensive benefits. This solution is then updated as the baseline solution, effectively controlling operating costs and improving economic efficiency and market competitiveness.

[0063] When a dynamic event is detected within the closed triangle formed by the order distribution center, vehicle dispatch center, and transportation hub inspection point, local link optimization is triggered, and dynamic correction parameters are generated based on the real-time positional relationships between the triangle's vertices. This dynamic event perception mechanism, based on spatial topology, accurately locates the affected area and promptly initiates targeted optimization. The generation of dynamic correction parameters ensures that the optimization strategy is closely aligned with actual dynamic changes, improving the path planning system's response speed and accuracy to emergencies and ensuring the stability of delivery tasks. The geometric center of gravity offset of the affected path segments is calculated based on the dynamic correction parameters. After targeted adjustments are made to these offset path segments, the global solution library is updated to generate a path adjustment result covering the dynamic area. By quantifying the geometric center of gravity offset, precise control of path adjustments is achieved, avoiding the additional costs associated with over-adjustments. Targeted adjustments ensure path validity while minimizing the impact on other path segments. New solutions that adapt to dynamic changes are quickly generated and updated in the global solution library, ensuring the consistency and timeliness of global path planning. A threshold for the survival of solutions is set based on the updated global solution library. When the solution optimization period exceeds the threshold, the old solution is eliminated based on the topological correlation characteristics of the dynamic area, and an alternative optimized link is generated based on the dynamic correction parameters. This mechanism establishes a survival of the fittest mechanism for solutions, preventing outdated solutions from affecting planning results due to their inability to adapt to environmental changes. It combines the dynamic regional topology association characteristics to generate alternative links, ensuring that new solutions can seamlessly connect with dynamic changes and continuously maintain the advancedness and efficiency of path planning solutions.

[0064] In a preferred embodiment of the present invention, step 1, based on current order demand, vehicle resources, and pre-set constraints, generates a set of initial routing solutions that meet time window, number of splits, vehicle type matching, and temperature zone requirements through a step-by-step progressive decision-making process, which may include:

[0065] Step 110 , performing multi-dimensional clustering on the current order demand based on the temperature zone attributes and time window restrictions of the orders, and generating multiple order subsets with the same temperature zone and a time window overlap greater than or equal to a preset threshold;

[0066] Step 111: Based on the order subset, traverse the vehicle resource library for models that meet the temperature zone adaptation conditions, determine a set of candidate vehicles whose vehicle capacity covers the total order demand in the subset, and split the order subsets with a capacity greater than the single load capacity, ensuring that the number of splits is ≤ a preset threshold, and generate a to-be-assigned task set including the split order units;

[0067] Step 112 , matching the set of tasks to be assigned with the set of candidate vehicles, and generating initial path segments that meet the vehicle model loading capacity constraints based on the distance priority between the real-time location of the vehicle and the order delivery point;

[0068] Step 113: Perform time window conflict detection on the initial path segments. If a time window conflict exists, dynamically adjust the node order of the path segments based on the vehicle's speed to generate a set of conflict-free feasible path segments.

[0069] In step 114 , the paths in the same temperature zone and geographically adjacent in the set of feasible path segments are combined and optimized to generate an initial set of path solutions including multi-vehicle coordinated scheduling and multi-order consolidated delivery.

[0070] In this embodiment of the present invention, all current order data is extracted from the order management system, including order number, temperature zone attributes (such as room temperature, refrigerated, frozen, etc.), delivery address, and time window limit (earliest delivery time and latest delivery time). With the temperature zone attribute as the primary clustering dimension, an empty set of different temperature zone categories (such as room temperature, refrigerated, and frozen) is created. All orders are traversed and classified into corresponding temperature zone category sets based on the order temperature zone attributes. In each temperature zone category set, for any two orders, and , calculate the overlapping duration of their time windows , assuming the order The time window is ,Order The time window is , then the overlapping duration , and calculate the total duration , overlap ,in Indicates the actual overlap duration of the time windows of two orders. Indicates the total coverage time after the time windows of two orders are combined. When the overlap ≥ preset threshold (such as 70%), the order and Mark as clusterable. Using a clustering algorithm (such as the DBSCAN algorithm), generate multiple order subsets within each temperature zone category set based on the orders marked as clusterable. The orders in each order subset are in the same temperature zone and the time window overlap meets the requirements.

[0071] For each order subset, the vehicle resource library is accessed to identify vehicle models that match the temperature zone attributes of the order subset. For example, for the refrigerated order subset, refrigerated truck models are identified to form a candidate vehicle list. The total cargo volume of all orders in the order subset is calculated and compared with the loading capacity of each model in the candidate vehicle list. Models from the candidate vehicle list whose loading capacity fully covers the total cargo volume of the order subset are selected to form the candidate vehicle set. For cases where the total cargo volume of the order subset exceeds the vehicle's single load capacity, a heuristic splitting strategy is employed. Orders are preferentially split based on similar cargo type and weight, ensuring that the cargo volume of each split order unit is less than the vehicle's single load capacity and that the number of splits is less than a preset threshold (e.g., three). After the split is complete, the split order units are integrated with the original order to generate a set of pending tasks. Using a geographic information system (GIS), the real-time location coordinates (latitude and longitude) of each vehicle in the candidate vehicle set and the location coordinates of each order delivery point in the pending tasks set are obtained. The straight-line distance between the vehicle's real-time location and the order delivery point is calculated using the Haversine formula. ,in is the radius of the Earth, 、 are the latitude and longitude of the vehicle’s location, 、 The system prioritizes vehicle-order combinations based on the latitude and longitude of the order delivery point, calculating the distance from the smallest to the largest. Combinations with closer distances have higher priority. Order units in the pending task set are then assigned to candidate vehicles in this order. A route is planned for each vehicle, starting from its current location and proceeding to its assigned order delivery point. Initial route segments are generated that meet the vehicle's load capacity constraints. During the assignment process, vehicle load capacities are monitored in real time to prevent overloading.

[0072] For each initial path segment, the estimated time for the vehicle to arrive at each order delivery point is calculated based on the vehicle's average driving speed, the distance between each order delivery point, and the order's time window limit. If the estimated time for the vehicle to arrive at a certain order delivery point is earlier than the earliest delivery time or later than the latest delivery time, the path segment is determined to have a time window conflict. When a time window conflict is detected, the simulated annealing algorithm is used to adjust the order of the order delivery point nodes in the path segment. The objective function is to reduce the time window conflict. By continuously trying new node order combinations, accepting or rejecting new combinations, the node order that minimizes the time window conflict is gradually found, and a set of conflict-free feasible path segments is generated. For the path segments in the feasible path segment set, the geographical distance between path segments in the same temperature zone is calculated based on the location coordinates of the order delivery points. By setting a distance threshold (such as 5 kilometers), geographically adjacent path segments are determined. For the determined route segments that are geographically adjacent and in the same temperature zone, multi-model collaborative scheduling and multi-order consolidated delivery strategies are considered, and a genetic algorithm is used for combinatorial optimization. With the shortest total driving distance and lowest delivery cost as the optimization goals, the route segments are merged and replanned to generate an initial route solution set that includes multi-model collaborative scheduling and multi-order consolidated delivery.

[0073] Multi-dimensional clustering categorizes orders that match temperature zones and time windows, reducing computational effort and improving order processing efficiency. This also ensures that orders within the same order subset meet temperature and time requirements during delivery, enhancing delivery rationality. It identifies compatible vehicle models and rationally splits orders to efficiently match vehicle resources with order demand, preventing overloaded or underloaded vehicles, improving vehicle utilization, and reducing transportation costs. Vehicles are prioritized based on their real-time location and distance to order delivery points, shortening total mileage, reducing delivery time, and improving delivery efficiency while also reducing fuel consumption and operating costs. It effectively detects and resolves time window conflicts, ensuring on-time order delivery, improving customer satisfaction, and enhancing market competitiveness. By combining and optimizing routes within the same temperature zone and geographically adjacent locations, it enables coordinated scheduling of multiple vehicle models and consolidated delivery of multiple orders, further optimizing resource allocation, increasing vehicle load rates, and reducing delivery costs.

[0074] In a preferred embodiment of the present invention, the above step 2, based on the initial path solution set, starts multiple independent optimization links, each link dynamically adjusts the local path nodes through a greedy strategy to generate a set of optimization candidate solutions, which may include:

[0075] Step 220 , extracting nodes whose path cost ratio is greater than a preset ratio from the initial path solution set as highly sensitive adjustment objects, and generating a candidate node set;

[0076] Step 221, for different optimized links, assigning differentiated node adjustment priorities to each link based on the distance deviation between the candidate node set and the current vehicle position, and generating a link-specific local adjustment range, specifically including: the different optimized links including the first link dynamically calculate the deviation amplitude difference based on the real-time distance deviation between the candidate node and the current vehicle position, and generate a distance-sensitive adjustment priority sequence;

[0077] The second link quantifies the path interaction strength based on the geographical overlap ratio between the candidate node’s path and the adjacent paths, and generates an overlap-sensitive adjustment priority sequence;

[0078] Step 222: Within the local adjustment range, perform a greedy operation on the highly sensitive node. If the time window margin of the path where the node is located is sufficient, try to insert the adjacent unassigned order; if the time window is tight, exchange with the adjacent path node to generate a candidate sub-plan within the link;

[0079] Step 223 : Conflict verification is performed on the candidate sub-schemes generated by each link, and schemes that conflict with the vehicle temperature zone adaptability are eliminated, and the remaining sub-schemes are merged into an optimized candidate scheme set.

[0080] In this embodiment of the present invention, detailed information for each route in the initial set of route plans is extracted from a database. This information includes the vehicle's specific route, such as the streets and intersections it passes through from the warehouse to the order delivery point. The actual distance of each route segment is obtained using a map measurement tool. The vehicle's fuel consumption coefficient for each route segment can be found in the vehicle's manual. The toll rates for each route segment can be found in the toll table published by the transportation department. Vehicle wear parameters specific to the vehicle model, such as the relationship between tire wear and component aging and mileage, are also available. Order information associated with each node (including order delivery points and transit points), including cargo weight, volume, and the customer's requested earliest and latest delivery times, is also available. Costs are calculated for each route. Fuel consumption costs are calculated by multiplying the distance of each route segment by the vehicle's fuel consumption coefficient for that route segment. Tolls are directly determined based on the toll rate for the route segment. Vehicle wear and tear costs are estimated based on mileage and the wear and tear parameters of the vehicle model. These three costs are summed to obtain the total route cost. Next, the cost of each node's corresponding route segment is calculated and divided by the total route cost to determine the node's contribution to the total route cost. For example, if the cost of a node's corresponding road segment is 100 yuan, and the total cost of the route is 500 yuan, then the node's cost contribution is 20%. Set a preset cost contribution ratio, such as 35%. Each node on the route is examined individually, and those with a cost contribution greater than this preset ratio are selected. These selected nodes, whose adjustments will have a greater impact on reducing the cost of the entire route, are included in the candidate node set.

[0081] With the help of Geographic Information System (GIS), we can obtain the accurate geographic coordinates of each node in the candidate node set and the real-time coordinates of each vehicle involved in the path planning. Based on these coordinates, we can calculate the actual straight-line distance from the vehicle's current position to each candidate node, and at the same time, we can obtain the theoretical driving distance of the vehicle to each candidate node according to the planned route from the initial path planning solution. For example, the vehicle is currently at The candidate node is Land, calculated by GIS Land The actual straight-line distance to the destination is 15 kilometers, while the theoretical distance according to the planned route is 20 kilometers. For each candidate node, the theoretical driving distance is subtracted from the actual distance to obtain the distance deviation value. If the actual distance is greater than the theoretical distance, it means that the currently planned path may have unreasonable situations such as detours when reaching the node, which requires special attention; even if the actual distance is less than the theoretical distance, this deviation value is recorded. For different optimization links, the candidate nodes are sorted according to the calculated distance deviation values, and different screening criteria are set for each link. For example, link 1 specifically selects nodes with distance deviation values greater than 10 kilometers, and link 2 selects nodes with distance deviation values between 5-10 kilometers. In order of priority, each link is given an exclusive local adjustment range, and the set of nodes that are focused on adjustment for each link is clearly defined. In this way, each link can perform its duties during optimization and work simultaneously to improve the overall optimization efficiency.

[0082] Within the local adjustment range defined for each link, for each highly sensitive node, the time window margin for the vehicle's arrival is calculated based on the average vehicle speed on each road segment, the distance between the node and adjacent nodes, and the earliest and latest delivery times for orders corresponding to the node. Specifically, the vehicle's estimated arrival time is subtracted from the latest delivery time for the order at the node. A criterion is set: if the time window margin is greater than 1.5 times the estimated travel time to the next node, the time window margin is considered sufficient; if this criterion is not met, the time window margin is considered tight. For example, if a vehicle is expected to arrive at a node at 10:00, the latest delivery time for orders at the node is 11:00, and the travel time to the next node is estimated to be 30 minutes, then if 1 hour is greater than 1.5 times 30 minutes (i.e., 45 minutes), the time window margin is sufficient. When a highly sensitive node has sufficient time window margin, a search is conducted within a certain geographic area (e.g., a 5-kilometer radius) centered on that node for unassigned orders. For each unassigned order found, a simulation is performed to insert it before or after the node in the current route. The resulting total route cost (including increases or decreases in fuel consumption due to changes in travel distance, changes in tolls, and so on) and time are calculated. Unassigned orders that reduce total route cost after insertion and still meet the order window requirements for each node are prioritized and inserted into the current route, generating a new route plan, known as a candidate sub-plan within the link.

[0083] When the time window of a highly sensitive node is tight, the system finds adjacent nodes on the current path. It then attempts to swap the positions of the highly sensitive node with the adjacent nodes. The resulting total cost (in terms of driving distance, fuel consumption, tolls, and other factors) is calculated, as well as whether the vehicle's arrival time at each node still meets the order's time window. If the swap reduces the total cost and meets the time window requirements of all nodes, the swapped path is retained as a candidate sub-plan within the link. All candidate sub-plans generated for each link are then individually checked for vehicle temperature zone compatibility. The order's temperature zone requirements (e.g., ambient, refrigerated, or frozen) are extracted from the candidate sub-plans. The temperature zone type of the vehicle executing the plan is then obtained and compared with the order's temperature zone requirements. If a mismatch occurs, such as a refrigerated order in the plan but the assigned vehicle is a normal temperature vehicle, the candidate sub-plan is determined to have a temperature zone compatibility conflict and is removed from the candidate list. The candidate sub-plans that have passed the temperature zone adaptability conflict verification are gathered together and put into a unified set of optimized candidate plans. The plans in this set have all been locally optimized and meet the vehicle temperature zone adaptation requirements.

[0084] Accurately calculating node cost contributions and identifying highly sensitive nodes provides precise direction for optimization efforts, focusing optimization resources on the key links that have the greatest impact on path costs. This avoids meaningless adjustments to all nodes and clarifies optimization priorities, providing a clear and actionable goal for reducing path costs. Distance deviation-based allocation of priorities and adjustment ranges to different optimization links enables efficient, multi-link parallel optimization. Each link focuses on optimizing nodes with different characteristics, fully utilizing computing resources and eliminating duplication of effort and resource waste during the optimization process. This approach fully exploits the optimization potential of all aspects of path planning, resulting in superior overall optimization results and ultimately more reasonable, efficient, and operationally relevant path planning results. Flexible greedy operations are implemented based on time window margins, inserting unassigned orders when the time window allows. This fully utilizes vehicle capacity, improves vehicle loading rates and order delivery efficiency, reduces empty vehicle travel, and reduces operating costs. Node swapping during tight time windows ensures on-time order delivery and avoids delivery delays due to time conflicts. While ensuring timely delivery, this approach reduces costs through rational route adjustments, enhancing the overall effectiveness and practicality of route planning solutions in practical applications. A rigorous temperature zone compatibility conflict verification mechanism fundamentally eliminates the risk of cargo damage caused by vehicle temperature zone mismatches, ensuring cargo quality and safety while safeguarding customer interests. Verified solutions are consolidated into a collection to ensure the final solution is highly feasible and effective in practical applications.

[0085] In a preferred embodiment of the present invention, step 3, performing real-time evaluation on the set of optimized candidate solutions, determining a final candidate solution based on cost savings rate, demand satisfaction rate, and soft service quality indicators, and updating the final candidate solution as the current baseline solution, may include:

[0086] Step 330 , for each of the optimization candidate solutions, calculate the difference between the total transportation cost and the historical benchmark solution, determine the solution with a cost saving rate ≥ a preset threshold, and form a primary cost reduction solution set;

[0087] Step 331: Verify the demand coverage of the primary price reduction solution set to check whether each solution covers all key order nodes. If there are any urgent orders that are not covered, mark them as demand-missing solutions and remove them, generating a secondary demand-compliant solution set.

[0088] Step 332: Perform a service quality assessment on the set of secondary demand compliance solutions. Based on the on-time delivery rate of orders and the frequency of route adjustments, a comprehensive service quality score is calculated. The solutions are sorted by the scores to generate a priority queue. Specifically, the following steps are performed: extracting the matching deviation between the actual delivery time of orders in each solution and the preset time window, using the ratio of the total deviation duration to the total number of orders as a benchmark indicator, and generating a ranking sequence of on-time rates for each solution.

[0089] Based on the punctuality ranking sequence, the number of node order adjustments triggered by each solution during the path optimization process is counted to generate an adjustment frequency sequence associated with the punctuality.

[0090] Based on the punctuality ranking sequence and the adjustment frequency sequence, determine the solutions where the total punctuality deviation is less than the preset upper limit and the adjustment frequency is less than the preset limit, forming a preliminary set of preferred solutions;

[0091] For the solutions in the initial set of preferred solutions, they are first sorted in ascending order by the total deviation of on-time rate. If the total deviation is the same, they are sorted again in ascending order by adjustment frequency. The number of unserved orders in the path coverage area is extracted and sorted in ascending order by the number of unserved orders to generate the final solution priority queue;

[0092] Step 333: Determine the candidate solution with the final ranking in the comprehensive score from the solution priority queue, and simultaneously activate the compensation mechanism for unmet demand. Re-inject the unallocated orders from the eliminated demand-missing solutions into the initial path solution set to generate a pool of orders to be reallocated.

[0093] Step 334: Update the final candidate solution to the current benchmark solution, and trigger the path replanning link of the order pool to be reallocated.

[0094] In an embodiment of the present invention, the transportation cost database is automatically connected to extract detailed cost data for each plan in the set of optimized candidate plans. For fuel costs, the actual fuel consumption records of vehicles corresponding to each plan in different road sections are retrieved from the vehicle fuel consumption monitoring system, and the total fuel cost is calculated in combination with the current oil price; tolls are obtained by querying the toll details of the driving sections of each plan in the electronic toll collection system; vehicle wear and tear costs are calculated based on the vehicle mileage, service life and wear and tear standard table. At the same time, the total transportation cost data of the past benchmark plan is retrieved. This data is stored in the past plan database and contains the total cost information of the plans that have been identified as better in the past period of time. For each candidate plan, according to the formula "cost saving rate = ×100%”. For example, the total cost of the previous benchmark plan is 8,000 yuan, and the fuel cost of a candidate plan is 2,500 yuan, the toll is 1,200 yuan, and the vehicle wear and tear fee is 800 yuan, so the total cost is 4,500 yuan. The cost saving rate of this plan is ×100% = 43.75%. The cost savings rate of all candidate solutions is automatically calculated. The company sets a threshold for cost savings, such as 25%, and compares the cost savings rate of each candidate solution against this threshold. Solutions with cost savings rates ≥ 25% are identified and form a preliminary set of cost reduction solutions. Solutions that do not meet the threshold are automatically marked and removed from the evaluation pool. An analysis report for the solutions that do not meet the threshold is also generated, explaining the reasons for failure. The order management department conducts a comprehensive evaluation of all orders based on multiple dimensions, including order urgency, product value, and customer importance. Urgent orders, such as expedited document orders requiring delivery within two hours; high-value product orders, such as electronic products valued at over 100,000 yuan; and orders from important customers, such as long-term VIP clients, are identified as key order nodes. Information on these key order nodes is compiled into a detailed list, including order number, customer name, product information, and urgency. Each solution in the preliminary set of cost reduction solutions is analyzed individually, and using a geographic information system (GIS) and order matching algorithm, the order delivery points in the solutions are compared with the list of key order nodes. The specific operation is that for each plan, the coordinates of all order delivery points in the path are extracted and matched with the coordinates of the key order nodes. If the coordinates of a key order node do not appear in the plan path, and the order is an urgent order, the plan is determined to be a demand-missing plan. For example, if a plan path does not contain a delivery point for an urgent drug delivery order, the plan is marked as a demand-missing plan. The detected demand-missing plans are automatically removed from the primary fee reduction plan set, and a list of removed plans is generated, recording in detail the name, number, and missing key order information of the removed plans. The remaining plans form a secondary demand compliance plan set to ensure that the plans in this set cover at least all key order nodes.

[0095] The order delivery on-time rate data corresponding to each plan in the secondary demand compliance plan set is obtained from the order delivery record database. The specific statistical method is to calculate the ratio of the number of orders delivered on time to the total number of orders in the past period of time (such as one month) for each plan. For example, a plan delivered a total of 150 orders in the past month, of which 135 were delivered on time. The on-time rate is × 100% = 90%. Furthermore, the route adjustment frequency data for each plan during actual execution is obtained from the route adjustment monitoring system. This data records the time, reason, and adjusted route information of each route adjustment in real time. A service quality score calculation rule is set, such as "Comprehensive service quality score = Order delivery on-time rate × 0.7 - Route adjustment frequency × 0.3." Based on this rule, a comprehensive score is calculated for each plan. For example, if a plan has an on-time rate of 85% and three route adjustments, the comprehensive score is 85% × 0.7 - 3 × 0.3 = 0.595 - 0.9 = -0.305. This automatically calculates the comprehensive scores for all plans. Based on the calculated comprehensive service quality scores, plans in the secondary demand compliance plan set are ranked from high to low. Plans with higher scores are ranked higher in the queue. A plan priority queue is generated, clearly displaying the ranking of each plan. A detailed score comparison chart is also provided, allowing managers to easily identify differences in service quality between plans. The candidate with the highest overall score is directly selected from the solution priority queue and determined as the final candidate. This solution achieves a relatively final balance in terms of cost savings, demand coverage, and service quality. A final candidate report is automatically generated, detailing the solution's various indicators, advantages, and comparisons with other solutions. For the demand-missing solutions eliminated in step 331, the unallocated order information is summarized and organized. Through the order management system and the initial path solution set database, these unallocated orders are reinserted into the initial path solution set, forming a pool of orders to be reallocated. At the same time, the path planners and dispatchers are automatically notified of new orders that need to be reallocated and provided with detailed information on the orders to be reallocated.

[0096] All information about the final candidate solution, including routing, vehicle allocation, estimated costs, and order delivery sequence, is updated to the current baseline solution database, replacing the existing baseline solution information. An update log is generated, recording the update time, operator, and content. At the same time, baseline solution update notifications are automatically sent to relevant departments and personnel to ensure that all parties are kept up to date with the latest routing benchmarks. For orders in the pool of pending reallocation orders, a dedicated rerouting process is automatically initiated. First, currently available vehicle information, including vehicle location, load, and remaining mileage, is obtained from the vehicle resource management system. Then, combined with road congestion and traffic restriction data provided by the real-time traffic information system, a routing algorithm is used to reroute these orders. During the planning process, the relevant parameters of the current baseline solution are referenced to ensure that the newly planned route is consistent with the overall delivery strategy. Ultimately, an optimized reallocation solution, the current baseline solution, is generated.

[0097] By precisely calculating cost savings and selecting solutions based on strict thresholds, the system can quickly and accurately identify solutions with competitive cost control advantages. This allows companies to focus resources on solutions that truly reduce transportation costs during route planning decisions, avoiding wasted time and effort on less cost-effective solutions, directly saving the company significant operating costs. Furthermore, the detailed cost data calculation and analysis process provides a reliable basis for cost management and budgeting. A rigorous demand coverage verification mechanism fundamentally ensures that key order nodes are included in delivery plans, avoiding business losses caused by missing urgent or important customer orders, such as customer complaints and order cancellations. This effectively maintains the company's service reputation and customer relationships. By eliminating solutions with missing demand, the system ensures the completeness and reliability of the order delivery solution set within the sub-standard demand compliance solution set, improving the stability of the order processing process and customer satisfaction. A comprehensive service quality assessment based on on-time delivery rate and route adjustment frequency fully considers the two key factors of delivery service timeliness and stability. Through scientific comprehensive scoring and ranking, the service quality of each solution can be clearly and accurately distinguished. Enterprises can give priority to solutions with high service quality, thereby effectively improving customer satisfaction and enhancing the competitiveness of the enterprise in the market.

[0098] In a preferred embodiment of the present invention, in step 4 above, according to the current benchmark scheme, when a dynamic event is detected within the closed triangle area formed by the order distribution center detection point, the vehicle dispatch center detection point, and the transportation hub detection point, local link optimization is triggered, and dynamic correction parameters are generated based on the real-time positional relationship between the triangle vertices. This may include:

[0099] Step 440 : collecting position data of three detection points within the triangular area in real time. If the offset between the real-time coordinates of any detection point and the reference position is greater than a preset tolerance threshold, it is determined to be a dynamic disturbance event and a coordinate sequence is extracted.

[0100] Step 441 , based on the coordinate sequence, calculate the real-time change rate of each side length of the triangle and the dynamic offset angle of the vertex angle, and generate a dynamic event impact intensity index in combination with the current path node distribution density;

[0101] Step 442 : Dynamically calculate the path redirection priority and detour compensation distance based on the dynamic event impact intensity index and the path node distribution density of the area passed through in the current benchmark solution, and generate dynamic correction parameters.

[0102] In an embodiment of the present invention, high-precision GPS positioning devices and Internet of Things sensors are deployed in order distribution centers, vehicle dispatch centers, and transportation hubs to collect location data of detection points once a second, including longitude and latitude information accurate to the meter level, as well as altitude data. These data are transmitted in real time via the network to a dedicated database for path planning. The database is equipped with a data verification mechanism to automatically eliminate abnormal data (such as obvious jumps and data that exceeds a reasonable range) to ensure the accuracy and validity of the collected data. The benchmark position coordinates of the three detection points are retrieved from the current benchmark solution database, and the offset is calculated for each detection point. For example, assuming its base longitude is , the real-time longitude is , the base latitude is , the real-time latitude is , use the Pythagorean theorem to calculate the offset distance on the plane , the tolerance threshold is set to 50 meters in advance. When the offset distance of any detection point is greater than the threshold, it is determined that a dynamic disturbance event has occurred in the area. For example, if the detection point The calculated offset distance is 60 meters, which immediately triggers the event response mechanism. Once a dynamic disturbance event is determined to have occurred, all coordinate data of the three detection points from 5 minutes before the event to the current moment are automatically extracted and arranged in chronological order to form a continuous coordinate sequence. These coordinate sequences are stored in a dedicated event data storage area and marked with the timestamp and unique identifier of the event. Based on the extracted coordinate sequence, the professional distance calculation algorithm of the geographic information system (GIS) is used to calculate the lengths of the three sides of the triangle at different times. For example, at time , the lengths of the three sides of the triangle are 、 、 ; at the moment , the lengths of the three sides become 、 、 , according to the formula Calculate side length The real-time rate of change of the other two edges is calculated similarly. and , the calculation results will be checked for rationality. If an abnormal change rate (such as more than 50%) occurs, it will be automatically recalculated or marked as suspicious data. Using the cosine theorem and inverse tangent function in trigonometric functions, the degree of the angle between the vertices of the triangle at different times can be accurately calculated. For example, at time ,vertex The angle is , at the moment , the angle becomes , through the formula Calculate vertices Dynamic offset angle of the included angle, similarly calculated for the vertex and vertices Angle offset and In order to improve the calculation accuracy, the coordinate data will be interpolated to ensure the accuracy of the angle calculation.

[0103] From the path data of the current benchmark solution, determine all path nodes within the triangular area and calculate the area of the triangular area using the spatial analysis function of GIS. , and count the number of nodes , thus obtaining the node distribution density Calculate the side length change rate, angle offset angle and node distribution density, set the side length change rate weight to 0.4, the angle offset angle weight to 0.3, and the node distribution density weight to 0.3. , and obtain the dynamic event impact intensity index According to the dynamic event impact intensity index and path node distribution density , set detailed priority calculation rules, when >0.6 and >10 nodes per square kilometer, the path redirection priority is set to high; when 0.3< ≤0.6 and 5 nodes / km²< When ≤10 nodes / km2, the priority is set to medium; when ≤0.3 and When the number of nodes is less than or equal to 5 per square kilometer, the priority is set to low. Perform quantitative calculations and accurately divide the priorities into three levels: high, medium, and low based on the calculation results. For example, if =0.7, =12 nodes / km², then = =16.8, the priority is determined to be high. Combined with the dynamic event impact intensity index , taking into account the vehicle speed (obtained in real time through on-board sensors) and road conditions (obtaining real-time congestion data from the traffic information platform), the detour compensation distance is calculated. For example, set the formula = ,in It is a coefficient that is dynamically adjusted according to the actual situation. When the traffic is smooth, =100 m / unit intensity; in traffic congestion, =150 m / unit strength, if the calculated =0.5, when traffic is smooth, detour compensation distance =100 0.5 = 50 meters. The calculated path redirection priority and detour compensation distance are combined to form a complete dynamic correction parameter. This parameter is stored in a structured data format and includes a priority identifier and a detour compensation distance value.

[0104] By collecting the position data of the detection points in real time and comparing it with the reference position, dynamic disturbance events within the triangle area can be captured quickly and accurately. The second-level data collection frequency and precise offset calculation method ensure the timeliness and sensitivity of event detection, avoiding the failure of path planning due to missed events. Extracting the coordinate sequence provides a more comprehensive understanding of the potential impact of dynamic events on the path. The dynamic event impact intensity index is calculated by comprehensively considering the change rate of the triangle side length, the vertex angle offset angle and the path node distribution density. , quantifying the impact of dynamic events on path planning from multiple dimensions. This multi-factor comprehensive evaluation method is more scientific and comprehensive than a single indicator judgment, and can accurately reflect the actual impact scope and severity of dynamic events. Dynamic correction parameters are generated according to the impact intensity of dynamic events and the distribution of path nodes, realizing the precision and dynamism of path adjustment strategies. The setting of path redirection priority can reasonably arrange the order of path adjustment according to the severity of the event, giving priority to paths with greater impact; the calculation of detour compensation distance provides a specific reference value for vehicle detours, avoiding resource waste and time delays caused by blind detours. The generation of dynamic correction parameters improves the adaptive ability and optimization effect of the path planning system in the face of dynamic events, ensuring the smooth progress of delivery tasks.

[0105] In a preferred embodiment of the present invention, step 5, calculating the geometric center of gravity offset of the affected path segment based on the dynamic correction parameter, performing directional adjustment on the offset path segment, and updating the global solution library to generate a path adjustment result covering the dynamic area, may include:

[0106] Step 550 , based on the path redirection priority in the dynamic correction parameters, extract the affected path segments within the triangular area and calculate the offset between the geometric center of gravity and the preset reference position;

[0107] Step 551 , performing a directional adjustment operation on the affected path segment according to the offset direction and the detour distance compensation value to obtain an adjusted path segment;

[0108] Step 552 , performing constraint penetration verification on the adjusted path segment to check whether the detour node is greater than the vehicle temperature zone adaptation range, and whether the path expansion violates the limit on the number of splits;

[0109] Step 553 : splice the verified adjusted path segments with the unaffected paths, update the global solution library, and generate a path adjustment result with dynamic area coverage.

[0110] In an embodiment of the present invention, based on the path redirection priority in the dynamic correction parameters, the spatial query function of the geographic information system (GIS) is used to screen the path segments within the triangular area from the current benchmark solution. If the priority is high, the buffer zone analysis tool of GIS is used to expand the triangular area outward by 500 meters (which can be adjusted according to actual needs) to extract all path segments that fall completely or partially within the expanded area. If the priority is medium or low, the Euclidean distance from each path segment node to the event occurrence point is calculated in combination with the coordinates of the dynamic event occurrence point, and only the critical path segments with a distance of less than 2 kilometers (which can be adjusted according to actual needs) are extracted. During the screening process, the path segments that have been marked as invalid or abandoned are automatically excluded. For each extracted path segment, it is regarded as a broken line consisting of a series of ordered nodes. Assume that the path segment contains nodes, nodes The coordinates of , ), the weighted average algorithm is used to calculate the geometric center of gravity coordinates. If the node represents an order delivery point, different weights are assigned according to the order weight, and the larger the weight, the higher the weight; if it is a transit point, the same weight is assigned, and the formula is used. , (in and Indicates the first The plane coordinates of the nodes, is the horizontal axis, is the vertical axis, Indicates the The importance weight of each node is proportional to the order weight. represents the total number of nodes in the path segment, and is the weighted geometric center of gravity of the path segment, that is, the balance point after considering the node weights). After the calculation is completed, the accuracy check is automatically performed. If the deviation between the center of gravity coordinates and the path segment node coordinates is too large (more than 10% of the path segment length), the node data is rechecked and manually corrected if necessary. The preset reference position coordinates of the geometric center of gravity of the path segment are retrieved from the reference solution database ( , ), using the distance measurement function of GIS, calculate the actual offset based on WGS84 , and the result is accurate to meters, and at the same time, through the inverse tangent function Calculate the offset angle with an accuracy of 0.1 degrees and record the offset amount and offset angle data to analyze the trend of the impact of dynamic events on the path. Combine the offset direction calculated in step 550 The adjustment direction is determined by combining the detour compensation distance in the dynamic correction parameters and real-time traffic flow data (obtained from the traffic management department's API). If the offset direction points outside the dynamic event area and the real-time traffic flow in that direction is less than a threshold (e.g., less than 500 vehicles per hour), the adjustment is made in that direction. If multiple feasible directions exist, a multi-objective optimization is constructed with the goals of avoiding the dynamic event, minimizing the increase in total driving distance, and minimizing the estimated driving time. The final adjustment direction is solved using a genetic algorithm. During the calculation process, factors such as road speed limits and turning restrictions are taken into account to ensure the feasibility of the adjustment direction.

[0111] When the detour distance needs to be increased, a new node is inserted at the appropriate position of the path segment according to the adjustment direction and detour compensation distance. First, the time required for the vehicle to travel the detour compensation distance at the current average speed (obtained in real time from the on-board sensor) is calculated. , and then at the end point of the original path segment A new node is inserted at the position corresponding to the time. The coordinates of the new node are obtained by taking the end point of the original path segment as the starting point, and calculating the displacement according to the travel time and speed along the adjustment direction. After the new node is inserted, the geometric center of gravity and the total length of the path segment are recalculated to ensure that the adjustment requirements are met. For cases where only the direction needs to be adjusted, the simulated annealing algorithm is used to move some nodes in the path segment. The maximum step size of the node movement (such as no more than 50 meters) and the temperature attenuation parameter are set to avoid excessive adjustment. After each node is moved, the distance between the adjusted path segment and the dynamic event area, as well as the change in the total length of the path segment are calculated. If the adjusted path avoids the dynamic event and minimizes the increase in the total length, the adjustment is retained; otherwise, the iterative adjustment is continued until the conditions are met. The temperature zone requirements of the orders associated with the detour nodes involved are obtained from the order information database (such as normal temperature 0 -30 , refrigerated-18 -0 , frozen<-18 At the same time, the temperature zone type and temperature zone adjustment range of the vehicle executing the route are obtained from the vehicle information database. The order temperature zone requirements and vehicle temperature zone types are compared one by one. If there is a mismatch, such as the order requires refrigerated transportation, but the vehicle only supports normal temperature transportation, then the adjusted route segment is determined to have failed the temperature zone adaptability verification. For cases where the temperature zone requirements are similar but not completely matched (such as the order requires refrigeration, and the vehicle temperature zone is -15 ), the damage risk probability of the goods in this temperature zone is assessed based on the data. If the risk probability is greater than the threshold (e.g., 5%), the verification is also determined to have failed. By analyzing the connection relationship between the nodes in the path segment, the number of path splits caused by the adjusted path segment in the entire global solution is counted. The preset limit for the number of path splits is 3. If the number of splits after the path expansion is greater than this limit, the impact of the split on the delivery efficiency is analyzed, and the estimated travel time and vehicle waiting time indicators of each sub-path after the split are calculated. If the sum of these indicators is greater than 120% of the original path's estimated total time (which can be adjusted according to actual needs), the adjusted path segment is determined to have failed the split limit verification.

[0112] Based on the unique identifiers and connection order of the path nodes, the verified adjusted path segments are spliced with the unaffected paths in the global solution library. A shortest path algorithm from graph theory (such as Dijkstra's algorithm) is used to ensure that the spliced path is final in terms of node connections and avoid unreasonable connections. During the splicing process, the access order and adjacency relationships of the path nodes are automatically updated, and the total length and estimated travel time of the spliced path are recalculated. The new spliced path overwrites the corresponding original path solution in the global solution library. All path-related information, including total length, estimated travel time, node access order, and vehicle allocation, is updated, and a detailed report on the path adjustment results is generated.

[0113] By extracting affected route segments based on path redirection priorities and spatial analysis techniques, and accurately calculating geometric center of gravity offsets, this approach can quickly and precisely identify the scope and extent of dynamic events' impacts on routes from massive amounts of route data. This avoids the waste of resources and inefficiencies associated with blind adjustments. By fully considering the importance of different nodes, the analysis results are more aligned with actual business needs, improving the relevance and accuracy of route adjustments and ensuring that companies focus optimization resources on the routes truly in need of improvement. Adjustment directions are determined based on offset direction, detour compensation distance, and real-time traffic data. Scientific node insertion and movement algorithms are employed to execute adjustments, ensuring clear objectives and rigorous planning for route adjustments. While avoiding dynamic events, multi-objective optimization and simulated annealing algorithms are used to rationally control route extension distances, effectively reducing transportation costs and delivery delays caused by unnecessary detours. Furthermore, the algorithm fully considers road conditions and vehicle driving characteristics, ensuring the geometric and logical rationality of the adjusted routes, improving their feasibility and practicality, and making the adjusted routes more suitable for real-world logistics and delivery scenarios. A strict constraint-penetration verification mechanism checks adjusted route segments based on two key dimensions: vehicle temperature zone compatibility and the limit on the number of route splits. This provides dual guarantees for the smooth operation of logistics and distribution operations. Temperature zone compatibility verification effectively avoids the risk of cargo damage caused by temperature zone mismatches between vehicles and orders, ensuring cargo quality and safety. Verification of the number of splits limits ensures that route adjustments are not overly complex, avoiding scheduling confusion and reduced delivery efficiency caused by excessive route splits. This ensures route feasibility and scheduling convenience, reducing the difficulty and risk of operational management.

[0114] In a preferred embodiment of the present invention, step 6, setting a solution survival threshold based on the updated global solution library, eliminating old solutions based on dynamic regional topology association features when the solution optimization period is greater than or equal to the threshold, and generating alternative optimization links based on dynamic correction parameters, may include:

[0115] Step 660 , dynamically calculate the solution survival threshold based on the frequency and path of dynamic events to adjust the result stability, and compare it with the current solution optimization period;

[0116] Step 661: If the optimization period is greater than or equal to the threshold, the matching degree between the path node and the offset of the center of gravity of the current region is extracted, and the solutions with matching degree less than the preset critical value are eliminated;

[0117] Step 662, based on the detour compensation distance and priority classification rules set in the dynamic correction parameters, dynamically generate an alternative optimized path, including giving priority to reusing path segments where there are verified detour path segments and achieving smooth connection with the current vehicle's real-time position; if there are no available detour segments, generate an alternative node set based on the detour compensation distance threshold to reconstruct the path.

[0118] In this embodiment of the present invention, dynamic event records within the triangular area formed by the order distribution center, vehicle dispatch center, and transportation hub over the past 12 months are retrieved from the event database. Each event record in the database contains information about the event type (traffic accident, road construction, weather disaster), occurrence time, duration, and impact range. Records are classified and counted by event type, for example:

[0119] Traffic accidents: By filtering the records with the event type field set to "traffic accident", a total of 30 accidents were found, including 12 during the morning rush hour and 18 during the evening rush hour.

[0120] Road construction: 20 "road construction" events were identified, involving five major transport roads, with an average construction period of 15 days;

[0121] Weather disasters: 10 weather disaster events were recorded, including 6 rainstorms and 4 heavy snowstorms.

[0122] Calculate the total frequency of dynamic events, add up the number of events of each type and divide by 12 months. At the same time, analyze the patterns of events in different time periods (such as weekdays, weekends, holidays) and different seasons. Obtain the actual execution data after each route adjustment from the route execution record database, including actual driving time, actual transportation cost, cargo delivery time, vehicle fuel consumption information, and compare it with the estimated driving time and estimated transportation cost data in the route adjustment plan. Taking a certain route adjustment as an example, a total of 20 transportation tasks were executed after the route adjustment. In terms of actual driving time, there were 5 times when the actual driving time deviation rate was greater than 15% compared to the estimated time. In terms of actual transportation cost, there were 4 times when the actual transportation cost deviation rate was greater than 10%. According to the frequency of dynamic events and path adjustment result stability index , using the formula Calculate the survival threshold of the solution ,when =5, =0.8, = ≈0.27 years. The numbers in the formula are dynamically adjusted according to the transportation characteristics and business needs of different regions. For example, for areas with complex traffic conditions and frequent dynamic events, the number is appropriately reduced to make the plan update more frequent; for areas with relatively stable traffic, it is increased to reduce unnecessary plan updates. The optimization cycle of the current plan is obtained from the plan management database, that is, the length of time from the timestamp of the plan completion to the current moment, and the length of time is compared with the calculated plan survival threshold. For comparison:

[0123] If the current solution optimization period ≥ threshold , then the scheme elimination mechanism is triggered and the process goes to step 661;

[0124] If the current solution optimization period is less than the threshold , then the solution is considered to be still in the valid use period and continue to use the solution for path planning.

[0125] The route adjustment record database stores detailed information about all route adjustments over the past period, including the reason for the adjustment, the adjusted route segment, and verification results (whether the adjustment was successful and how effective the adjustment was). During a search, the system identifies verified route segments that match the current situation based on the detour compensation distance and priority ranking rules set in the current dynamic adjustment parameters. For example, if the detour compensation distance in the current dynamic adjustment parameters is 800 meters, the database searches for route segments with a detour compensation distance between 700 and 900 meters and a successful verification result. If a matching path segment is found, the Dijkstra algorithm is used to calculate the shortest path from the vehicle's current location to the starting point of the path segment based on the vehicle's real-time location (obtained via onboard GPS with meter-level accuracy). Utilizing GIS path planning capabilities, a transition path is generated from the vehicle's current location to the starting point of the path segment, without violating traffic regulations (such as restrictions and prohibitions) and road restrictions (such as bridge load limits and tunnel height limits). This transition path is then connected to the verified path segment to form a complete alternative optimized path. During the connection process, the path is smoothed to avoid unreasonable situations such as sharp turns and sudden lane changes, ensuring safe and comfortable driving. If no detour path segment is available, a search for a matching alternative node within the current area is performed based on the detour compensation distance threshold set in the dynamic correction parameters. The search radius, centered on the vehicle's current location, is determined based on the detour compensation distance threshold (for example, if the detour compensation distance threshold is 800 meters, the search radius is set to 1000 meters, leaving a buffer area). Through the spatial analysis function of GIS, all locations that can serve as path nodes are determined within the search radius, including road intersections, parking lots, gas stations, etc., to form a set of alternative nodes.

[0126] By comprehensively analyzing the frequency of dynamic events and the stability of route adjustment results, a dynamic solution lifespan threshold is calculated. This allows the update cycle of routing solutions to better reflect changes in the actual transportation environment. This avoids the resource waste that can result from fixed-cycle updates, such as frequent solution updates during stable traffic conditions, which can lead to unnecessary computational and management costs. It also prevents solution lag and prevents the long-term use of outdated solutions in areas prone to dynamic events, which can lead to low transportation efficiency and increased costs. Flexible adjustment of the threshold based on actual conditions ensures the use of the latest and most effective routing solutions for dynamic events in different regions and time periods, improving the timeliness and adaptability of routing plans. Outdated solutions are eliminated based on the matching degree between the offsets of route nodes and regional centroids. Path solutions are rigorously screened based on spatial relevance, ensuring that the solutions retained in the global solution library are closely aligned with the current transportation area and eliminating those with weak regional relevance. This reduces the use of system resources, such as database storage space and computing resources, by ineffective solutions, improving the quality and retrieval efficiency of the solution library. Verified detour route segments are prioritized for reuse and seamlessly transitioned, leveraging successful experience. When faced with dynamic events, it can quickly reuse mature path fragments and generate alternative optimized paths in the shortest time.

[0127] The embodiment of the present invention further provides a computer-readable storage medium storing instructions, which, when executed on a computer, causes the computer to execute the above-described method. All implementations in the above-described method embodiment are applicable to this embodiment and can achieve the same technical effects.

[0128] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A vehicle path planning method based on an adaptive optimization algorithm, characterized in that: The method comprises: Step 1: Based on current order demand, vehicle resources, and pre-set constraints, a step-by-step progressive decision-making process is used to generate an initial set of routing solutions that meet time window, number of splits, vehicle type matching, and temperature zone requirements. Step 2: Based on the initial path solution set, multiple independent optimization links are started. Each link dynamically adjusts the local path nodes through a greedy strategy to generate a set of optimization candidate solutions. Step 3: Conduct real-time evaluation of the optimized candidate solution set, determine the final candidate solution based on cost savings rate, demand satisfaction rate, and soft service quality indicators, and update the final candidate solution to the current baseline solution; Step 4: According to the current benchmark solution, when a dynamic event is detected in the closed triangle area formed by the order distribution center detection point, the vehicle dispatch center detection point, and the transportation hub detection point, local link optimization is triggered, and dynamic correction parameters are generated based on the real-time position relationship between the triangle vertices. This includes real-time acquisition of the position data of the three detection points in the triangle area. If the offset between the real-time coordinates of any detection point and the benchmark position is greater than the preset tolerance threshold, it is determined to be a dynamic disturbance event and a coordinate sequence is extracted. Based on the coordinate sequence, the real-time change rate of the length of each side of the triangle and the dynamic offset angle of the vertex angle are calculated, and the dynamic event impact intensity index is generated in combination with the current path node distribution density. Based on the dynamic event impact intensity index and the path node distribution density of the area passed through in the current benchmark solution, the path redirection priority and detour compensation distance are dynamically calculated to generate dynamic correction parameters. Step 5: Calculate the geometric center of gravity offset of the affected path segment based on the dynamic correction parameters, perform directional adjustments on the offset path segment, and then update the global solution library to generate a path adjustment result covering the dynamic area. This includes extracting the affected path segment within the triangular area based on the path redirection priority in the dynamic correction parameters, calculating the offset between the geometric center of gravity and the preset reference position; performing directional adjustments on the affected path segment based on the offset direction and the detour distance compensation value to obtain the adjusted path segment; performing constraint penetration verification on the adjusted path segment to check whether the detour node is greater than the vehicle temperature zone adaptation range and whether the path expansion violates the split number limit; splicing the verified adjusted path segment with the unaffected path, updating the global solution library, and generating a path adjustment result covering the dynamic area; Step 6: Set a solution survival threshold based on the updated global solution library. When the solution optimization period is greater than or equal to the threshold, eliminate the old solution based on the dynamic regional topology association characteristics, and generate an alternative optimization link based on the dynamic correction parameters.

2. The vehicle path planning method based on the adaptive optimization algorithm according to claim 1 is characterized in that: Based on current order demand, vehicle resources, and pre-set constraints, a step-by-step, progressive decision-making process is used to generate a set of initial routing solutions that meet time window, number of splits, vehicle type matching, and temperature zone requirements. These solutions include: Based on the temperature zone attributes and time window restrictions of the orders, multi-dimensional clustering is performed on the current order demand to generate multiple order subsets with the same temperature zone and time window overlap ≥ the preset threshold; Based on the order subset, the vehicle resource library is traversed for models that meet the temperature zone adaptation conditions. The candidate vehicle set whose vehicle capacity covers the total order demand in the subset is determined. The order subset with a capacity greater than the single load capacity is split, ensuring that the number of splits is ≤ the preset threshold. The task set to be assigned is generated, including the split order units. Match the set of tasks to be assigned with the candidate vehicle set, and generate initial path segments that meet the vehicle model loading capacity constraints based on the distance priority between the vehicle's real-time location and the order delivery point; Perform time window conflict detection on the initial path segments. If a time window conflict exists, dynamically adjust the node order of the path segments based on the vehicle speed to generate a set of conflict-free feasible path segments. The paths in the same temperature zone and geographically adjacent in the set of feasible path segments are combined and optimized to generate a set of initial path solutions including multi-vehicle collaborative scheduling and multi-order consolidated delivery.

3. The vehicle path planning method based on the adaptive optimization algorithm according to claim 2, characterized in that: Based on the initial path solution set, multiple independent optimization links are started. Each link dynamically adjusts the local path nodes through a greedy strategy to generate a set of optimization candidate solutions, including: Nodes whose path cost ratio is greater than a preset ratio are extracted from the initial path solution set as highly sensitive adjustment objects to generate a candidate node set; For different optimization links, based on the distance deviation between the candidate node set and the vehicle's current position, differentiated node adjustment priorities are assigned to each link, generating a link-specific local adjustment range. Within the local adjustment range, greedy operations are performed on highly sensitive nodes. If the time window margin of the node's path is sufficient, an attempt is made to insert adjacent unassigned orders. If the time window is tight, an exchange is performed with adjacent path nodes to generate candidate sub-plans within the link. The candidate sub-schemes generated by each link are verified for conflicts, the schemes that conflict with the vehicle temperature zone adaptability are eliminated, and the remaining sub-schemes are merged into a set of optimized candidate schemes.

4. The vehicle path planning method based on the adaptive optimization algorithm according to claim 3 is characterized in that: The different optimization links include a first link that dynamically calculates deviation amplitude differences based on real-time distance deviations between candidate nodes and the current vehicle position to generate a distance-sensitive adjustment priority sequence; described The second link quantifies the path interaction strength based on the geographical overlap ratio between the candidate node's path and the adjacent paths, and generates an overlap-sensitive adjustment priority sequence.

5. The vehicle path planning method based on the adaptive optimization algorithm according to claim 4 is characterized in that: Conduct real-time evaluation of the optimization candidate set, determine the final candidate based on cost savings, demand fulfillment, and soft service quality indicators, and update the final candidate to the current baseline solution, including: For each plan in the set of optimized candidate plans, the difference between the total transportation cost and the historical benchmark plan is calculated, and the plan with a cost saving rate ≥ the preset threshold is determined to form a set of primary cost reduction plans; Verify the demand coverage of the primary fee reduction plan set to check whether each plan covers all key order nodes. If there are any urgent orders that are not covered, mark them as demand-missing plans and eliminate them, generating a secondary demand-compliant plan set. Evaluate the service quality of the secondary demand compliance solution set, calculate the comprehensive service quality score based on the order delivery on-time rate and the frequency of route adjustment, and generate a priority queue for the solutions based on the score; Determine the candidate solutions with the final ranking in comprehensive scores from the solution priority queue, and simultaneously activate the compensation mechanism for unmet demand. Re-inject the unallocated orders from the eliminated demand-missing solutions into the initial path solution set to generate a pool of orders to be reallocated. The final candidate solution is updated to the current baseline solution, and the path replanning link of the order pool to be reallocated is triggered.

6. The vehicle path planning method based on the adaptive optimization algorithm according to claim 5 is characterized in that: Evaluate the service quality of the secondary demand compliance solution set. Calculate a comprehensive service quality score based on the order delivery on-time rate and the frequency of route adjustments. Generate a priority queue of solutions based on the score, including: Extract the matching deviation between the actual delivery time of orders in each plan and the preset time window, use the ratio of the total deviation time to the total number of orders as the benchmark indicator, and generate the on-time rate ranking sequence of each plan; Based on the punctuality ranking sequence, the number of node order adjustments triggered by each solution during the path optimization process is counted to generate an adjustment frequency sequence associated with the punctuality. Based on the punctuality ranking sequence and the adjustment frequency sequence, determine the solutions where the total punctuality deviation is less than the preset upper limit and the adjustment frequency is less than the preset limit, forming a preliminary set of preferred solutions; For the solutions in the preliminary preferred solution set, they are first sorted in ascending order according to the total deviation of punctuality. If the total deviation is the same, they are sorted again in ascending order according to the adjustment frequency, and the number of unserviced orders in the path coverage area is extracted. The final solution priority queue is generated by sorting in ascending order according to the number of unserviced orders.

7. The vehicle path planning method based on the adaptive optimization algorithm according to claim 6 is characterized in that: A solution survival threshold is set based on the updated global solution library. When the solution optimization period exceeds the threshold, the old solution is eliminated based on the dynamic regional topology association characteristics, and an alternative optimization link is generated based on the dynamically modified parameters, including: Adjust the result stability based on the frequency and path of dynamic events, dynamically calculate the solution survival threshold, and compare it with the current solution optimization cycle; If the optimization period is greater than or equal to the threshold, the matching degree between the path node and the current area center offset is extracted, and the solutions with matching degree less than the preset critical value are eliminated; Based on the detour compensation distance and priority classification rules set in the dynamic correction parameters, alternative optimized paths are dynamically generated, including prioritizing the reuse of verified detour path segments and achieving smooth connection with the current vehicle's real-time position; if there are no available detour segments, a set of alternative nodes is generated according to the detour compensation distance threshold for path reconstruction.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a program, which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images