Community sharing logistics management method and system

Abstract

The application relates to the technical field of logistics management and control, and provides a community shared logistics management and control method and system. Through integration of mobile resource data, fixed resource data and environment data, a logistics resource graph and a logistics passing network are constructed, accurate matching of resources and intelligent planning of a distribution path are realized, a distribution process is dynamically monitored and timely adjusted, and overall operation efficiency and service quality of community logistics are significantly improved. Specifically, the logistics management and control system can optimize resource allocation, reduce resource idling and waste, and reduce logistics operation cost. Through early planning of a main path and a backup path, path adjustment is carried out in combination with real-time environment changes, traffic congestion and abnormal passing events and other problems are effectively coped with, logistics distribution time is shortened, distribution timeliness is improved, a logistics task distribution process is comprehensively monitored and dynamically adjusted, stability and reliability of logistics service are ensured, user satisfaction is enhanced, and sustainable development of community shared logistics is promoted.

Description

Technical Field

[0001] This application relates to the field of logistics management technology, and in particular to community-shared logistics management methods and systems. Background Technology

[0002] Currently, most community logistics resource management systems adopt a traditional static management model, managing mobile and fixed resources independently. This fails to fully consider the spatiotemporal relationships and dynamic changes between resources. This approach leads to inflexible resource scheduling, resulting in both idle and scarce resources, hindering efficient resource utilization. In the route planning stage, existing algorithms typically calculate based solely on road distances and fixed travel times, ignoring the impact of traffic congestion, weather changes, and road construction on delivery routes. This often results in planned routes failing to meet timeliness requirements during actual delivery, increasing costs and delay risks. Furthermore, most systems fail to address how to match dispersed resources and tasks to achieve dynamic optimization of logistics delivery routes. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this application provides a community-shared logistics management method and system.

[0004] In the first aspect, this application provides a community shared logistics management and control system, which includes: acquiring mobile resource data, fixed resource data and environmental data of the community by deploying edge nodes, identifying abnormal passage events in the community based on environmental data, and constructing a logistics resource map based on mobile resource data, fixed resource data and abnormal passage events. Resource nodes represent mobile resources and fixed resources, and node attributes include resource location and transport capacity. The edges between resource nodes represent the spatiotemporal constraint relationship between mobile resources and fixed resources.

[0005] A logistics network is generated by combining abnormal passage events in the community. The passage difficulty coefficient of each path segment in the logistics network is predicted by a spatiotemporal attention graph neural network. The resource matching degree is determined based on the resource location and transportation capacity in the logistics resource map and the passage difficulty coefficient of each path segment in the logistics network. The delivery path of the logistics task is generated by a genetic algorithm based on the resource matching degree. The delivery path includes at least the main path and the backup path.

[0006] The system receives and processes logistics tasks through a blind selection mechanism, delivers the tasks according to the delivery route, and monitors the delivery progress deviation, resource health, and environmental mutation index of the logistics tasks during the delivery process to determine whether to update the logistics resource map and delivery route.

[0007] As an optional implementation, the logic for constructing the logistics resource map includes:

[0008] Mobile and fixed resources are encoded using graph convolutional networks to form resource nodes, with resource location and capacity as node attributes, and node attributes are incrementally updated according to changes in resource nodes.

[0009] The spatial distance between each resource node is calculated based on the resource location. The temporal constraints between each resource node are determined by combining abnormal passage events. The spatial distance and temporal constraints are integrated to obtain the spatiotemporal constraint relationship, and the spatiotemporal constraint relationship is used as the edge between resource nodes.

[0010] The spatiotemporal correlation between mobile and fixed resources is analyzed through a spatiotemporal attention mechanism, and the weights of the edges are dynamically adjusted according to the spatiotemporal correlation to form a micro-layer.

[0011] The resource nodes at the micro level are aggregated, with each community area as a super node. The average location and comprehensive transportation capacity of all resource nodes in each community area are calculated as super node attributes. The edges between super nodes are determined based on the traffic distance and travel time between community areas to form a macro level. The macro and micro levels are then integrated to construct a logistics resource graph.

[0012] As an optional implementation, the sub-logic for identifying abnormal access events includes:

[0013] By deploying edge nodes, environmental data from the community is acquired and preprocessed. This environmental data includes traffic video streams, meteorological data, traffic flow data, and announcement texts. Features of the environmental data are extracted and fused using a feature alignment algorithm to generate fused features. At the same time, homomorphic encryption technology is used to encrypt the fused features.

[0014] The encrypted fused features are input into a spatiotemporal attention network. The multi-head attention mechanism captures the temporal dependencies of environmental data, and the spatial attention mechanism analyzes the spatial dependencies of environmental data, outputting a spatiotemporal feature vector.

[0015] Based on historical environmental data, the normal state boundary is adjusted by an adaptive dynamic threshold algorithm, and the spatiotemporal feature vector is processed by a long short-term memory network to predict the environmental state. The predicted environmental state is compared with the normal state boundary to identify abnormal passage events, and abnormal passage events are recorded and classified.

[0016] As an optional implementation, the generation logic of the logistics access network includes:

[0017] A basic road network is constructed based on community map data. Roads are abstracted as directed edges, and intersections are abstracted as road network nodes. The basic attributes of each road network node include road network location and road network type. The basic attributes of each directed edge include road length, road speed limit, number of lanes, and road type.

[0018] Map abnormal access events in the community to the underlying road network to adjust the directed edges and their underlying properties;

[0019] Based on the duration of the abnormal passage event, the basic road network is divided into multiple time slices. The directed edges and their basic attributes in each time slice are updated. The buffer zone in the basic road network is determined based on the impact area of ​​the abnormal passage event, and the basic attributes of the directed edges are updated incrementally within the buffer zone.

[0020] By combining the resource location and transport capacity of resource nodes in the logistics resource map, the set of road network nodes through the directed edges of the basic road network for each resource node is recorded to generate the logistics passage network.

[0021] As an optional implementation, the logic for generating the delivery route includes:

[0022] Resource nodes and road network nodes are alternately arranged into chromosomes. Each gene position includes location coordinates, access time window and allocated capacity. Based on resource matching degree, an initial resource node is selected for each logistics task, an initial path is generated, and the initial path is randomly perturbed to generate an initial population.

[0023] The fitness function is determined based on the total delivery time, total travel cost, and resource utilization rate. The initial path is iteratively optimized through selection, crossover, and mutation operations.

[0024] After the genetic algorithm iterative optimization terminates, the chromosome with the highest fitness is selected as the main path for decoding. The main path is then subjected to deterministic verification, and the path with the greatest difference from the main path is selected as the first backup path. At the same time, a second backup path is generated based on abnormal passage events to generate delivery paths, and trigger conditions are set for the first and second backup paths.

[0025] As an optional implementation, the prediction sub-logic for the travel difficulty coefficient of each path segment includes:

[0026] The system extracts the basic attributes of each road network node and each directed edge in the transportation network, receives environmental data and the resource location and capacity of mobile resources, and acquires historical traffic data to form multimodal data. At the same time, it extracts and fuses features from the multimodal data to form multimodal features.

[0027] Multimodal features are processed by a spatiotemporal attention graph neural network to output temporal attention vectors and spatial attention vectors. The temporal attention vectors and spatial attention vectors are then fused by tensor product to generate a spatiotemporal correlation matrix.

[0028] Based on the spatiotemporal correlation matrix, the physical passage cost, traffic accident cost, and resource competition cost of each path segment in the logistics passage network are determined to comprehensively predict the passage difficulty coefficient of each path segment.

[0029] As an optional implementation, the sub-logic for determining the resource matching degree includes:

[0030] Spatial matching degree is determined based on the resource location of resource nodes to the road network location of road network nodes; time matching degree is determined based on the time constraints between resource nodes and the physical passage cost of each path segment in the logistics network; and capacity matching degree is determined by comparing the transport capacity of resource nodes with the resource competition cost of each path segment in the logistics network.

[0031] The weights of spatial matching, temporal matching, and capacity matching dimensions are dynamically adjusted based on the type and severity of abnormal passage events.

[0032] The resource matching degree is determined by weighted summation based on the dimension weights.

[0033] As an optional implementation, the logic for updating the logistics resource map and delivery route includes:

[0034] The logistics tasks are delivered according to the delivery route, and the deviation between the actual delivery progress and the target progress of each logistics task is calculated to obtain the delivery progress deviation.

[0035] Real-time monitoring of the device status of mobile resources and the usage status of fixed resources, and assessment of resource health using fault tree analysis;

[0036] Integrate abnormal passage events and calculate the environmental mutation index using the entropy weight method;

[0037] The fuzzy control algorithm outputs update decision signals based on delivery progress deviation, resource health, and environmental mutation index to determine whether to update the logistics resource map and delivery route.

[0038] As an optional implementation, the processing sub-logic of the logistics task includes:

[0039] After receiving a logistics task, the task characteristics of the logistics task are extracted. The task characteristics include task type, weight and volume, time requirement and address information. The address information is segmented into community area and location coordinates and the information is encrypted in layers.

[0040] Based on the real-time status of mobile and fixed resources in the logistics resource map, a set of available resources is selected. Based on the set of available resources and the resource matching degree, a reinforcement learning algorithm is used to calculate the expected revenue of each resource node in undertaking logistics tasks, so as to determine the resource node that will execute the logistics task.

[0041] Secondly, this application provides a community-shared logistics management and control method, which includes: acquiring mobile resource data, fixed resource data and environmental data of the community by deploying edge nodes, and identifying abnormal access events in the community based on the environmental data;

[0042] A logistics resource graph is constructed based on mobile resource data, fixed resource data, and abnormal passage events. Resource nodes represent mobile resources and fixed resources. Node attributes include resource location and transport capacity. Edges between resource nodes represent the spatiotemporal constraint relationship between mobile resources and fixed resources.

[0043] By combining abnormal access events in the community, a logistics access network is generated, and the access difficulty coefficient of each path segment in the logistics access network is predicted by a spatiotemporal attention graph neural network.

[0044] The resource matching degree is determined based on the resource location and transportation capacity in the logistics resource map and the passage difficulty coefficient of each path segment in the logistics passage network.

[0045] Based on resource matching degree, a genetic algorithm is used to generate delivery routes for logistics tasks. The delivery route includes at least a main route and a backup route.

[0046] The system receives and processes logistics tasks through a blind selection mechanism, delivers the tasks according to the delivery route, and monitors the delivery progress deviation, resource health, and environmental mutation index of the logistics tasks during the delivery process to determine whether to update the logistics resource map and delivery route.

[0047] Compared with existing technologies, the beneficial effects of this application are as follows: By integrating mobile resource data, fixed resource data, and environmental data, a logistics resource map and logistics traffic network are constructed, enabling precise matching of resources and intelligent planning of delivery routes. Simultaneously, the delivery process is dynamically monitored and adjusted in a timely manner, significantly improving the overall operational efficiency and service quality of community logistics. Specifically, this logistics management system optimizes resource allocation, reduces resource idleness and waste, lowers logistics operating costs, effectively addresses traffic congestion and abnormal traffic events by pre-planning main and backup routes and adjusting routes in conjunction with real-time environmental changes, shortens delivery time, improves delivery efficiency, comprehensively monitors and dynamically adjusts the logistics task delivery process, ensures the stability and reliability of logistics services, enhances user satisfaction, and promotes the sustainable development of community shared logistics.

[0048] By deploying edge nodes to acquire mobile resource data, fixed resource data, and environmental data of the community, and identifying abnormal passage events based on environmental data, a logistics resource map is constructed. This enables the comprehensive and accurate acquisition of various types of community logistics-related data, transforming mobile and fixed resources into resource nodes. Resource location and transport capacity are used as node attributes, while abnormal passage events are considered to determine the spatiotemporal constraints between resource nodes. This allows the logistics resource map to truly reflect the distribution and interrelationships of community logistics resources, providing a solid data foundation for subsequent logistics planning and decision-making.

[0049] By combining abnormal traffic events to generate a logistics traffic network, and using a spatiotemporal attention graph neural network to predict the traffic difficulty coefficient of each path segment, the logistics traffic network can be dynamically adjusted according to real-time abnormal traffic events, making the logistics traffic network more in line with actual road traffic conditions. By predicting the traffic difficulty coefficient of path segments, quantitative reference indicators are provided for resource matching and route planning. For example, when allocating delivery tasks, resources and routes corresponding to path segments with low traffic difficulty coefficients can be prioritized to reduce delivery time and costs, provide a basis for formulating reasonable transportation plans, and improve the controllability of logistics transportation.

[0050] Based on the resource location and capacity in the logistics resource map, as well as the difficulty coefficient of the path segments in the logistics network, the resource matching degree is determined, and a delivery path is generated through a genetic algorithm. This achieves precise matching between resources and logistics tasks, fully considering factors such as resource location, capacity, and path difficulty to ensure that the most suitable resources are selected to execute the logistics task. The application of the genetic algorithm can find the optimal delivery path in a complex path search space. The setting of the main path and backup path improves the flexibility and reliability of logistics delivery. When the main path becomes impassable due to unforeseen circumstances, it can be quickly switched to the backup path to ensure that the logistics task is completed on time and reduce delivery delays and customer complaints caused by path problems.

[0051] The system receives and processes logistics tasks through a blind selection mechanism, delivers them according to the delivery route, and monitors delivery progress deviations, resource health, and environmental mutation indices during the delivery process to determine whether to update the logistics resource map and delivery route. This ensures the fairness and objectivity of logistics task allocation, avoids interference from human factors, and enables timely detection of problems that occur during the delivery process. Based on the monitoring results, the system determines whether to update the logistics resource map and delivery route, thus achieving dynamic adjustment of logistics delivery. Attached Figure Description

[0052] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0053] Figure 1 This is a system flowchart of the community shared logistics management and control system provided in the embodiments of this application;

[0054] Figure 2 This is a prediction sub-logic diagram of the passage difficulty coefficient for each path segment of the community shared logistics management system provided in this application embodiment;

[0055] Figure 3 A logic diagram for generating delivery routes in the community shared logistics management and control system provided in this application embodiment;

[0056] Figure 4 This is a flowchart illustrating the community shared logistics management method provided in this application embodiment. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of the embodiments of this application more apparent and understandable, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0058] Example 1:

[0059] like Figure 1 The diagram shown illustrates a system flowchart for a community shared logistics management and control system, which includes a resource perception module, a route generation module, and a flexible delivery module.

[0060] The resource awareness module is used to acquire mobile resource data, fixed resource data and environmental data of the community by deploying edge nodes. Based on the environmental data, it identifies abnormal passage events in the community and constructs a logistics resource graph based on the mobile resource data, fixed resource data and abnormal passage events. Resource nodes represent mobile resources and fixed resources. Node attributes include resource location and transport capacity. The edges between resource nodes represent the spatiotemporal constraint relationship between mobile resources and fixed resources.

[0061] Furthermore, the sub-logic for identifying abnormal access events includes:

[0062] By deploying edge nodes, environmental data from the community is acquired and preprocessed. This environmental data includes traffic video streams, meteorological data, traffic flow data, and announcement texts. Features of the environmental data are extracted and fused using a feature alignment algorithm to generate fused features. At the same time, homomorphic encryption technology is used to encrypt the fused features.

[0063] The encrypted fused features are input into a spatiotemporal attention network. The multi-head attention mechanism captures the temporal dependencies of environmental data, and the spatial attention mechanism analyzes the spatial dependencies of environmental data, outputting a spatiotemporal feature vector.

[0064] Based on historical environmental data, the normal state boundary is adjusted by an adaptive dynamic threshold algorithm, and the spatiotemporal feature vector is processed by a long short-term memory network to predict the environmental state. The predicted environmental state is compared with the normal state boundary to identify abnormal passage events, and abnormal passage events are recorded and classified.

[0065] The community environment is complex and ever-changing, and single data sources cannot fully reflect the actual traffic conditions. It is necessary to integrate multi-source information such as traffic, weather, and announcements, while ensuring data privacy and security. Edge nodes, including cameras, geomagnetic sensors, and meteorological monitoring equipment, are deployed at key locations such as community entrances and main roads. Cameras capture traffic video streams, geomagnetic sensors acquire traffic flow data, and meteorological monitoring equipment acquires meteorological data such as temperature, humidity, and rainfall. Municipal announcements are also captured in real time. After acquiring environmental data, a federated learning framework is used. Each edge node extracts features from the environmental data locally using feature alignment algorithms, including video image features, sensor values, and text semantics. These features are mapped to the same data structure and fused into a fused feature. Homomorphic encryption is then used to encrypt the fused feature, transmitting only the encrypted calculation results. This achieves multi-type data compatibility and privacy protection, providing a complete and secure data foundation for subsequent analysis. It avoids the risk of information loss or leakage due to inconsistent data formats and provides standardized, encrypted multimodal data input for the spatiotemporal attention network, ensuring the accuracy and security of feature extraction.

[0066] Anomalies possess spatiotemporal characteristics, requiring feature extraction from both temporal variation patterns and spatial impact ranges to improve identification accuracy. Encrypted, fused features are input into a spatiotemporal attention network—a Transformer-based spatiotemporal multimodal network. In the temporal dimension, a multi-head attention mechanism analyzes the temporal dependencies of environmental data across different time periods, including identifying traffic flow patterns during weekday morning and evening rush hours. In the spatial dimension, a spatial attention mechanism focuses on different areas within the community, such as analyzing road conditions around community gates and centralized delivery points. Ultimately, a spatiotemporal feature vector fusing spatiotemporal and resource-related information is output. This fully captures the spatiotemporal dependencies of environmental data, enabling a more accurate characterization of the community's traffic environment and providing rich information for anomaly identification. The output feature vector contains deep spatiotemporal and resource information, providing a reliable basis for Long Short-Term Memory (LSTM) networks to predict environmental conditions.

[0067] Community traffic conditions fluctuate, necessitating dynamic definition of the normal and abnormal boundaries to promptly detect sudden anomalies. Firstly, based on historical and real-time environmental data, an adaptive dynamic threshold algorithm dynamically adjusts the normal state boundary according to factors such as season and weather; for example, the standard road speed automatically decreases in rainy weather. Then, spatiotemporal feature vectors are input into a long short-term memory network to learn historical normal patterns and predict the current environmental state. The predicted environmental state is compared with the normal state boundary; if the predicted environmental state exceeds the normal state boundary, it is identified as an abnormal traffic event. The geographic system is used to locate the area where the abnormal traffic event occurs. Simultaneously, the abnormal traffic events are recorded and classified, annotating the event type, location, impact range, and duration, generating abnormal event description data. Event types include traffic accidents and road construction. This enables intelligent and dynamic identification of abnormal traffic events, rapid response to emergencies within the community, and reduced delays and errors from manual judgment. The generated abnormal event description data directly drives updates to the logistics resource map and adjustments to route planning, ensuring the logistics management system adapts promptly to environmental changes.

[0068] Specifically, the construction logic of the logistics resource map includes:

[0069] Mobile and fixed resources are encoded using graph convolutional networks to form resource nodes, with resource location and capacity as node attributes, and node attributes are incrementally updated according to changes in resource nodes.

[0070] The spatial distance between each resource node is calculated based on the resource location. The temporal constraints between each resource node are determined by combining abnormal passage events. The spatial distance and temporal constraints are integrated to obtain the spatiotemporal constraint relationship, and the spatiotemporal constraint relationship is used as the edge between resource nodes.

[0071] The spatiotemporal correlation between mobile and fixed resources is analyzed through a spatiotemporal attention mechanism, and the weights of the edges are dynamically adjusted according to the spatiotemporal correlation to form a micro-layer.

[0072] The resource nodes at the micro level are aggregated, with each community area as a super node. The average location and comprehensive transportation capacity of all resource nodes in each community area are calculated as super node attributes. The edges between super nodes are determined based on the traffic distance and travel time between community areas to form a macro level. The macro and micro levels are then integrated to construct a logistics resource graph.

[0073] The status of community logistics resources changes in real time, requiring dynamic updates to node attributes to accurately reflect resource availability. Resource location determines its spatial distribution, influencing transportation route planning, while capacity reflects its carrying capacity, crucial for resource scheduling and task allocation. Furthermore, the status of logistics resources is dynamic, meaning the location of mobile resources and warehouse storage levels change dynamically. Therefore, incremental updates to node attributes are necessary to ensure the resource graph reflects the real-time situation. A graph convolutional network is used to encode mobile and fixed resources, including delivery vehicles and autonomous vehicles, and fixed resources such as parcel lockers and delivery stations. Resource location and capacity are used as node attributes. China Mobile Resources updates the node attributes of mobile resource nodes by combining driving speed increments. At the same time, based on their historical trajectories and abnormal passage events, it predicts future location and capacity changes through gating loop units. For example, it predicts that a delivery vehicle will detour due to road construction ahead, and adjusts its arrival time to the next node and remaining capacity. Fixed resources update node attributes based on data such as usage frequency and maintenance status. When the express locker is used frequently, its available capacity is estimated to be reduced. This presents the dynamic status of resource nodes in real time and accurately, providing precise data support for resource scheduling and route planning. The updated node attributes provide an accurate resource state basis for the generation of spatiotemporal constraint edges, ensuring that the edge weight calculation is consistent with the actual situation.

[0074] Resource coordination is constrained by spatiotemporal factors, necessitating the construction of edges reflecting real-world constraints to optimize resource scheduling. Spatial distance determines the shortest path length for transportation, while temporal constraints affect transportation timeliness, such as increased travel time due to traffic congestion and road construction. This study constructs edges between resource nodes based on a spatiotemporal attention mechanism. Using Geographic Information System (GIS) technology, the spatial distance between each resource node is calculated based on its location. The Haversine formula is used to calculate the spherical distance between two latitude and longitude coordinates, yielding the spatial distance between resource nodes. Combining the impact range and severity of abnormal traffic events, the travel time of affected road segments is dynamically adjusted. For example, if a road is closed for 2 hours due to construction, the connection between mobile resources passing through that segment and the corresponding fixed resource node will be delayed by 2 hours, thus determining the temporal constraints between resource nodes. Integrating spatial distance and temporal constraints forms spatiotemporal constraint relationships, which are then used as edges to construct the edges between resource nodes. This ensures that the logistics resource map accurately reflects the actual situation during logistics transportation, considering both spatial and temporal dimensions. This provides accurate information for optimizing transportation routes and rationally scheduling transportation time, helping to reduce transportation costs and improve transportation efficiency.

[0075] In logistics management and control systems, the spatiotemporal correlation between different resource nodes varies. By analyzing the spatiotemporal correlation between resource nodes through a spatiotemporal attention mechanism and dynamically adjusting the edge weights, important resource connections can be highlighted, making the logistics resource graph more focused on key logistics activities and providing more accurate information for micro-level resource scheduling and optimization. The spatiotemporal correlation between mobile and fixed resources is analyzed using this mechanism. Spatial distance and time constraints are normalized to eliminate the influence of dimensions. Then, the spatiotemporal correlation between different resource node pairs is measured by calculating their attention scores at different time steps. The calculation of attention scores is based on a multi-head attention mechanism, such as mapping spatial distance and time constraints to multiple different feature spaces, calculating attention scores in each feature space, and then weighted summing of the multiple attention scores to obtain the final spatiotemporal correlation.

[0076] Based on the calculated spatiotemporal correlation, the weights of edges are dynamically adjusted. Specifically, the weights of edges are adjusted to reflect the spatiotemporal correlation between delivery vehicle capacity and parcel locker access needs. Edges between nodes with high capacity matching have increased weights, prioritizing resource scheduling. By continuously adjusting edge weights, a micro-layer reflecting the spatiotemporal correlation between resource nodes is formed. The spatiotemporal attention mechanism automatically identifies important resource connections in the logistics management system. By adjusting edge weights, it highlights key logistics activity paths. This facilitates more refined resource scheduling at the micro-level, prioritizing transportation tasks between resource nodes with high spatiotemporal correlation, improving logistics efficiency, and reducing operating costs. The resulting micro-layer provides the foundation for resource node aggregation and the construction of the macro-layer. The weighted edge and node relationships in the micro-layer allow for more reasonable division of community areas during resource node aggregation, and also influence the calculation of supernode attributes and the determination of edges between supernodes.

[0077] In large-scale logistics management and control systems, there are numerous resource nodes. Directly analyzing and processing all resource nodes at the micro level leads to excessive computational complexity and low efficiency. By aggregating resource nodes at the micro level to form super nodes at the macro level, the logistics management and control system can be abstracted and generalized at a higher level, simplifying the system structure and facilitating macro-level logistics planning and decision-making. Aggregating resource nodes at the micro level involves grouping resource nodes with similar characteristics and close connections into the same community area, with each community area serving as a super node. For each super node, the average position and comprehensive transport capacity of all resource nodes within it are calculated as super node attributes. The average position is calculated by weighting the resource positions of all nodes, with the weights determined according to the importance of the resource nodes. The comprehensive transport capacity is obtained by adding the transport capacities of all nodes.

[0078] Edges between supernodes are determined based on the traffic distance and travel time between community areas. The traffic distance is calculated using geographic information system (GIS) technology to determine the distance between the center locations of community areas, while the travel time is calculated by combining real-time and historical traffic data. For example, if the traffic distance between community area A and community area B is Z kilometers and the travel time is W hours, then an edge is constructed between supernode A and supernode B with the attributes of traffic distance Z kilometers and travel time W hours. Finally, the macro and micro layers are integrated to form a complete logistics resource graph. This is achieved by simultaneously retaining the resource nodes and edges of the micro layer and the supernodes and edges of the macro layer in the graph structure, and establishing a mapping relationship between micro layer nodes and macro layer supernodes.

[0079] Aggregating resource nodes to form a macro layer reduces the complexity of the logistics resource map and improves the efficiency of analyzing and processing large-scale logistics management systems. The macro layer provides an overall view of the logistics management system, facilitating macro-level logistics planning and decision-making. Integration with the micro layer enables the logistics resource map to take into account information at both macro and micro levels, providing more comprehensive and flexible support for logistics management. The constructed logistics resource map provides strong data support for various applications in the logistics management system, enabling hierarchical management of community logistics resources. The macro layer controls overall scheduling, while the micro layer ensures precise task execution, improving the management efficiency and response speed of the logistics management system. The completed logistics resource map provides comprehensive resource information for the path generation module, supporting its calculation of resource matching degree and planning of delivery routes.

[0080] The path generation module is used to generate a logistics network by combining abnormal passage events in the community. It predicts the passage difficulty coefficient of each path segment in the logistics network through a spatiotemporal attention graph neural network. Based on the resource location and transportation capacity in the logistics resource map and the passage difficulty coefficient of each path segment in the logistics network, it determines the resource matching degree. Based on the resource matching degree, it generates the delivery path of the logistics task through a genetic algorithm. The delivery path includes at least the main path and the backup path.

[0081] Specifically, the generation logic of the logistics access network includes:

[0082] A basic road network is constructed based on community map data. Roads are abstracted as directed edges, and intersections are abstracted as road network nodes. The basic attributes of each road network node include road network location and road network type. The basic attributes of each directed edge include road length, road speed limit, number of lanes, and road type.

[0083] Map abnormal access events in the community to the underlying road network to adjust the directed edges and their underlying properties;

[0084] Based on the duration of the abnormal passage event, the basic road network is divided into multiple time slices. The directed edges and their basic attributes in each time slice are updated. The buffer zone in the basic road network is determined based on the impact area of ​​the abnormal passage event, and the basic attributes of the directed edges are updated incrementally within the buffer zone.

[0085] By combining the resource location and transport capacity of resource nodes in the logistics resource map, the set of road network nodes through the directed edges of the basic road network for each resource node is recorded to generate the logistics passage network.

[0086] Community road networks form the foundation of logistics route planning. Abnormal traffic events can alter road conditions in real time, necessitating dynamic updates to network attributes to reflect actual road conditions. Utilizing community map data, a geographic information system (GIS) abstracts roads within the community as directed edges and intersections as road network nodes. Each road network node is assigned detailed attributes, including its location and type, such as the community area it belongs to and the presence of schools or hospitals nearby prone to congestion. Rich information is added to the directed edges, including road length, speed limits, number of lanes, and road type. When the resource perception module detects an abnormal traffic event, such as a minor vehicle collision at the entrance of a community, the logistics control system immediately matches spatial coordinates... The system accurately locates affected road sections, adjusts the speed limit of the affected road section to walking speed, sets the number of lanes to 1 (meaning only the emergency lane is retained), and adds an "Accident in Progress" label and estimated recovery time to the road network attributes of that road section. If the abnormal traffic event is municipal construction, in addition to adjusting the road attributes, the turning restrictions of surrounding roads will also be updated simultaneously to prevent vehicles from accidentally entering the construction area. This creates a dynamic road network that fits the actual situation of the community, can respond to abnormal events in a timely manner, provides accurate basic data for subsequent route planning, and reduces delivery delays caused by lagging road information. The updated basic road network provides accurate geographical information and traffic conditions for time slice division and buffer zone setting.

[0087] The impact of abnormal traffic events on roads changes over time and has spatial diffusion. Dividing roads into time slices and setting buffer zones allows for more refined management of road traffic status. Based on the duration of abnormal traffic events, the basic road network is divided into multiple time slices. For a road construction event expected to last 3 hours, it is divided into 6 time slices of 30 minutes each. Within each time slice, road attributes are dynamically updated according to the progress of event handling. For example, in the first time slice after construction begins, the construction section is completely closed, and the traffic speed on surrounding alternative roads is reduced by 20%. As construction progresses, some lanes of the construction section will be opened in subsequent time slices, and the attributes of surrounding roads will be adjusted accordingly.

[0088] Centered on the location of abnormal traffic events, the buffer radius is determined based on the event type and severity: a 200-meter buffer for traffic accidents and a 500-meter buffer for construction events. Incremental updates are performed on all road segments within the buffer zones. For example, temporary traffic guidance signs are added to remind vehicles to detour in advance, traffic light timing schemes are adjusted, and green light durations on surrounding main roads are increased to alleviate traffic pressure. For parcel lockers near construction areas, their attributes are marked as "increased delivery difficulty," providing a reference for subsequent resource scheduling and route planning. This achieves refined spatiotemporal management of road traffic conditions, enabling more accurate prediction of road traffic conditions at different times and in different areas, improving the timeliness and accuracy of route planning. The well-defined time slices and updated buffer road networks clarify the possibility and cost of resource nodes passing through various road network nodes at different times, providing temporal and spatial constraints for the integration of resource nodes and road network nodes.

[0089] Logistics resources need to operate on a road network. Clearly defining the relationships between resource nodes and road network nodes is crucial for constructing a complete logistics network and supporting subsequent path planning. For each mobile and fixed resource node in the logistics resource graph, a path search algorithm calculates the shortest path from the resource node to each road network node. During the calculation, the basic attributes of the road network and the characteristics of the resources themselves are fully considered. For trucks with heavy loads, roads with insufficient load capacity are avoided; for autonomous vehicles with limited range, shorter paths are prioritized. The set of road network nodes along the directed edges of the basic road network for each resource node is recorded, and a resource passage cost attribute is added to each associated path to generate the logistics network.

[0090] The cost attributes comprehensively consider factors such as travel distance, time consumption, and additional costs due to road congestion. For example, the time cost of a delivery vehicle's route from the delivery station to the community's parcel locker increases significantly if it passes through congested sections. Frequent starts and stops also increase fuel consumption costs. Furthermore, based on the capacity status of resource nodes and the urgency of tasks, associated paths are prioritized, ensuring smooth passage for urgent tasks. This establishes a close connection between resources and the road network, forming a logistics network that includes resource access characteristics. This provides a complete topology and cost information for subsequent calculations of path segment access difficulty coefficients, determination of resource matching, and generation of delivery routes. The integrated logistics network provides a specific set of path segments and basic attributes for predicting access difficulty coefficients, making the predictions more aligned with actual logistics scenario needs.

[0091] Furthermore, such as Figure 2 As shown, the prediction sub-logic for the difficulty coefficient of each path segment includes:

[0092] The system extracts the basic attributes of each road network node and each directed edge in the transportation network, receives environmental data and the resource location and capacity of mobile resources, and acquires historical traffic data to form multimodal data. At the same time, it extracts and fuses features from the multimodal data to form multimodal features.

[0093] Multimodal features are processed by a spatiotemporal attention graph neural network to output temporal attention vectors and spatial attention vectors. The temporal attention vectors and spatial attention vectors are then fused by tensor product to generate a spatiotemporal correlation matrix.

[0094] Based on the spatiotemporal correlation matrix, the physical passage cost, traffic accident cost, and resource competition cost of each path segment in the logistics passage network are determined to comprehensively predict the passage difficulty coefficient of each path segment.

[0095] Single data points cannot fully reflect the difficulty of navigating a route. Integrating multi-source data and extracting features can more accurately depict the traffic conditions of a route. This involves extracting basic attributes of road network nodes, as well as fundamental attributes such as the length of directed edges, speed limits, and the number of lanes, from the logistics network. Simultaneously, real-time environmental data is integrated, including video streams captured by traffic cameras, meteorological data from weather sensors, and municipal announcements, as well as real-time location and capacity data of mobile resources, and historical traffic data, specifically traffic speed statistics for the past week. The multimodal data is preprocessed, and the video streams are used to identify vehicle numbers and directions through object detection algorithms. Natural language processing techniques are employed to extract key events and time / location information from the text data. Then, a feature alignment algorithm is used to map the features of different types of data to the same dimensional space. Rainfall (numerical), traffic flow (numerical), and road type (categorical) in meteorological data are converted into fusionable feature vectors through encoding and normalization. Then, through feature cross-combination, multimodal features are formed to enhance the data's representation ability. This forms a feature system containing multidimensional information, which comprehensively reflects various factors affecting the difficulty of path passage. It provides rich and valuable input data for the subsequent spatiotemporal attention graph neural network. High-quality multimodal feature data directly affects the spatiotemporal feature extraction effect of the spatiotemporal attention graph neural network on path segments, and thus affects the prediction accuracy of the passage difficulty coefficient.

[0096] The difficulty of navigating a path is influenced by both temporal and spatial factors. A spatiotemporal correlation matrix can be used to uncover the spatiotemporal dependencies between path segments, providing a basis for cost analysis. The fused multimodal features are input into a spatiotemporal attention graph neural network. In the temporal dimension, a multi-head attention mechanism processes the features of path segments in different time slices, learning the patterns of traffic flow changes throughout the day, such as identifying which road segments are prone to congestion during the morning rush hour, as well as the duration and spread of congestion. In the spatial dimension, a spatial attention mechanism analyzes the mutual influence between adjacent path segments, such as how congestion on main roads affects the traffic conditions of surrounding side roads, or how construction in a certain area alters the delivery routes of nearby express delivery lockers. Tensor products are used to fuse temporal and spatial attention vectors to generate a spatiotemporal correlation matrix.

[0097] Each element in the spatiotemporal correlation matrix represents the correlation strength between two path segments within a specific time window. A matrix element (segment A, segment B, time slice t) = 0.7 indicates a strong correlation between the traffic conditions of segment A and segment B within time slice t. This is because they share a key intersection; if one segment experiences congestion, the other will also be affected. The logistics management system visualizes and analyzes the spatiotemporal correlation matrix, helping managers intuitively understand the spatiotemporal relationships between path segments and identify potential traffic bottlenecks and congestion propagation paths. This allows for in-depth analysis of the complex spatiotemporal dependencies between path segments, comprehensively depicting the path traffic conditions from both temporal and spatial dimensions. It provides deep feature support for accurately assessing the physical traffic costs, traffic accident costs, and resource competition costs of each path segment. The spatiotemporal correlation matrix determines the weighting of each cost factor in calculating the traffic difficulty coefficient, and its accuracy directly affects the reliability of the final traffic difficulty coefficient, thus influencing the determination of resource matching and the generation of delivery routes.

[0098] The difficulty of a route is composed of various cost factors, including physical accessibility, traffic accidents, and resource competition. Only by comprehensively analyzing the costs of each dimension can the accessibility of a route segment be accurately predicted. Based on the spatiotemporal correlation matrix, the physical accessibility cost, traffic accident cost, and resource competition cost of each route segment in the logistics network are analyzed. The physical accessibility cost is calculated based on the length of the route segment, the speed limit, and the real-time traffic flow. Preferably, for a route segment with a length of 2 kilometers and a speed limit of 40 kilometers per hour, if the current traffic flow is normal, the estimated travel time is 3 minutes. If congestion occurs and the actual travel speed drops to 20 kilometers per hour, the travel time will be extended to 6 minutes, and the corresponding physical accessibility cost will increase. The impact of factors such as road gradient and turning radius on travel time and energy consumption is also considered to further adjust the cost calculation.

[0099] Traffic accident costs are assessed by combining information on abnormal traffic events and historical accident data. If a route segment is located in an accident-prone area, or if there is ongoing construction or severe weather, traffic accident costs will increase. If a road segment has recently experienced frequent minor collisions and is currently undergoing road repairs, the traffic accident costs for that segment will increase significantly, manifesting as increased estimated travel time and a higher probability of cargo damage due to accident risks. Resource competition costs, on the other hand, are calculated based on the number of other logistics resources passing through the route segment at the same time and their capacity demand. If multiple delivery tasks are scheduled to pass through a route segment within a certain time slot, intense resource competition will lead to reduced travel speed and increased waiting time, thereby increasing resource competition costs.

[0100] In practice, when multiple delivery vehicles simultaneously head to the same residential area, queues form at the entrance, increasing delivery time and operating costs. This paper normalizes each cost to ensure they fall within the same numerical range. Then, based on weights determined by a spatiotemporal correlation matrix, the three costs are weighted and synthesized to obtain a traffic difficulty coefficient for each route segment. Difficulty levels are then categorized as low, medium, and high. For example, the weight of physical traffic cost for a route segment might be 0.4, traffic accident cost 0.3, and resource competition cost 0.3. After calculating each cost, a weighted traffic difficulty coefficient is obtained, and a threshold is used to determine if it falls into the high difficulty level. This comprehensively and systematically assesses various factors affecting route traffic difficulty. Through scientific cost analysis and coefficient synthesis methods, a quantitative indicator for route difficulty is provided, offering accurate decision-making basis for resource matching and route planning. The traffic difficulty coefficient, as a crucial parameter, directly participates in the calculation of resource matching, influencing the assessment of resource and task suitability, and ultimately determining the direction and quality of delivery route generation.

[0101] Furthermore, the sub-logic for determining resource matching degree includes:

[0102] Spatial matching degree is determined based on the resource location of resource nodes to the road network location of road network nodes; time matching degree is determined based on the time constraints between resource nodes and the physical passage cost of each path segment in the logistics network; and capacity matching degree is determined by comparing the transport capacity of resource nodes with the resource competition cost of each path segment in the logistics network.

[0103] The weights of spatial matching, temporal matching, and capacity matching dimensions are dynamically adjusted based on the type and severity of abnormal passage events.

[0104] The resource matching degree is determined by weighted summation based on the dimension weights.

[0105] The matching of resources and tasks needs to be considered comprehensively from three dimensions: space, time, and capacity. A single-dimensional assessment cannot meet the actual logistics needs and will lead to resource waste or task delays. Among them, spatial matching degree refers to calculating the actual travel distance between resource nodes and road network nodes, and making corrections based on the topology of the road network and real-time traffic conditions. For example, although the straight-line distance between two nodes is short, the actual travel distance will increase significantly due to the presence of closed communities or one-way roads in between. In this case, the spatial matching degree will decrease accordingly. At the same time, the travel direction of resources and the accessibility of road network nodes should be considered. If a delivery vehicle needs to make multiple U-turns or detours from its current location to the target node, the spatial matching degree will also decrease.

[0106] Time matching degree is based on the time constraints between resource nodes and the physical passage cost of the path segment. The time constraint is the timeliness requirement of the logistics task, and the physical passage cost needs to be combined with the passage difficulty coefficient to predict the passage time and assess the possibility of the resource completing the task within the specified time. For urgent tasks, a strict time window is set. If the resource cannot reach the target node within the required time, the time matching degree is 0. For ordinary tasks, a certain time buffer is allowed. The matching degree is calculated based on the difference between the actual passage time and the specified time. If a task requires delivery within 60 minutes, and a resource is calculated to take 70 minutes to pass through the relevant path segment, then the time matching degree is calculated according to the preset calculation formula (such as 1-(70-60) / 60) to obtain the specific value.

[0107] Capacity matching is obtained by comparing the remaining capacity of resource nodes with the resource competition cost of a route segment. It reflects the capacity demand of the route segment in the current time period. If the resource competition in a route segment is fierce and the resource capacity requirement is high, and the remaining capacity of a certain resource cannot meet the task demand, the capacity matching degree is low. Conversely, if the remaining capacity of a resource is sufficient and the demand of the route segment is small, the capacity matching degree is high. When a certain express locker is expected to receive a large number of packages in a certain time period, it requires a large amount of capacity for delivery. If the remaining capacity of a certain delivery vehicle can only meet part of the demand, then the capacity matching degree between the vehicle and the express locker is low. The degree of resource and task suitability is quantified from three key dimensions, providing basic data for a comprehensive assessment of resource matching. This avoids unreasonable resource allocation caused by a single-dimensional assessment. The calculation result of the three-dimensional matching degree is the basis for dynamic weight adjustment and comprehensive matching degree calculation. Its accuracy directly affects the reliability of the final resource matching degree, and thus affects the generation of delivery routes.

[0108] The importance of the three matching dimensions—space, time, and capacity—changes under different abnormal scenarios. Dynamically adjusting the weights allows resource matching strategies to better align with actual needs. A mapping relationship is established between the types of abnormal traffic events and the weight adjustment strategy. During heavy rain, slippery roads lead to a general decrease in driving speed, increasing the importance of the time dimension. The weight of the time matching degree is increased from 0.3 to 0.5, while the weights of the space and capacity matching degrees are correspondingly decreased. During holidays when express delivery volume surges, resource capacity becomes a key factor, increasing the capacity matching degree weight from 0.3 to 0.4, with appropriate adjustments to the time and space matching degree weights. Dynamic weight adjustment is achieved through a fuzzy inference algorithm. Factors such as the severity, duration, and impact range of abnormal traffic events are used as input, and a pre-defined fuzzy rule base is used. For example, if an abnormal event is severe and its duration... The longer the congestion, the greater the weight of the time matching degree, thus outputting the adjustment coefficients of the weights of each dimension. When a road segment is detected to be congested for a long time due to a serious traffic accident, the fuzzy inference algorithm will increase the weight of the time matching degree by 0.2, decrease the weight of the space matching degree by 0.1, and decrease the weight of the capacity matching degree by 0.1. The logistics management and control system will also continuously optimize the fuzzy rule base based on the historical weight adjustment effect and the actual resource matching situation, improving the accuracy and adaptability of weight adjustment. This allows the resource matching strategy to be flexibly adjusted according to different abnormal scenarios and actual needs, improving the rationality and effectiveness of resource allocation, and enhancing the logistics system's ability to cope with complex environmental changes. The adjusted weights are used to comprehensively calculate the resource matching degree, directly affecting the priority ranking of resources and the generation of the initial population, thereby affecting the process of the genetic algorithm finding the optimal delivery route.

[0109] The three-dimensional matching degree is integrated into a single comprehensive index, which facilitates resource screening and sorting, and quickly finds the most suitable resource for the task. Based on the dynamically adjusted weights of spatial matching degree, temporal matching degree, and capacity matching degree, the resource matching degree is calculated using a weighted summation formula. Preferably, for a resource node, the spatial matching degree is 0.8, the temporal matching degree is 0.6, and the capacity matching degree is 0.7, with corresponding weights of 0.3, 0.4, and 0.3 respectively. Therefore, the resource matching degree = 0.8 × 0.3 + 0.6 × 0.4 + 0.7 × 0.3 = 0.69. The resource matching degrees of all available resources are then sorted in descending order to generate candidate resources. In the sorting process, if resources have the same matching degree, a secondary sorting will be conducted based on factors such as the resource's historical delivery success rate and service quality evaluation. If the overall matching degree of resources A and B is 0.7, but resource A has a higher delivery success rate in similar tasks in the past, then resource A will be ranked ahead of resource B. At the same time, the logistics management system will mark each resource in the candidate resource list with detailed information, including matching degree score, main advantages, and potential risks. The main advantages include time advantage and capacity advantage, and the potential risks include the possibility of route congestion and the aging of resource equipment, which will facilitate subsequent selection and decision-making.

[0110] This establishes a unified and quantifiable resource evaluation standard. By combining the matching degree and detailed annotation information, the best-matched resources for logistics tasks can be quickly and accurately selected, providing high-quality resource selection for the genetic algorithm to generate the initial path, thus improving the efficiency and quality of path planning. The candidate resource list serves as an important basis for the generation of the initial population of the genetic algorithm, and the quality of its resources directly affects the quality of the initial path and the convergence speed of the algorithm, thereby affecting the final delivery path generation effect.

[0111] Specifically, such as Figure 3 As shown, the logic for generating the delivery route includes:

[0112] Resource nodes and road network nodes are alternately arranged into chromosomes. Each gene position includes location coordinates, access time window and allocated capacity. Based on resource matching degree, an initial resource node is selected for each logistics task, an initial path is generated, and the initial path is randomly perturbed to generate an initial population.

[0113] The fitness function is determined based on the total delivery time, total travel cost, and resource utilization rate. The initial path is iteratively optimized through selection, crossover, and mutation operations.

[0114] After the genetic algorithm iterative optimization terminates, the chromosome with the highest fitness is selected as the main path for decoding. The main path is then subjected to deterministic verification, and the path with the greatest difference from the main path is selected as the first backup path. At the same time, a second backup path is generated based on abnormal passage events to generate delivery paths, and trigger conditions are set for the first and second backup paths.

[0115] Genetic algorithms need to transform the delivery path problem into an encoding form for optimization. The quality and diversity of the initial population directly affect the search efficiency and the quality of the final solution. Therefore, it is necessary to design a reasonable encoding method and generate a diverse set of initial paths. In the encoding process, a method of alternating encoding of resource nodes and road network nodes is adopted. Taking a logistics task in a community as an example, if there is a task that starts from delivery station A and needs to deliver goods to express lockers B and C in sequence, its chromosome encoding is [delivery station A → intersection 1 → express locker B → road segment X → intersection 2 → express locker C]. Each gene position contains three key pieces of information: location coordinates, access time window, and allocated transportation capacity. For example, for the gene position corresponding to delivery station A, the location coordinates record its specific latitude and longitude on the community map, the access time window is determined according to the task's timeliness requirements and the current task arrangement of the delivery station, assuming it is [9:00-9:15], and the allocated transportation capacity is set according to the weight and volume of the goods and the remaining transportation capacity of the delivery station.

[0116] When generating an initial path for each logistics task based on resource matching degree, the system first selects the resource node with the highest matching degree from the candidate resource list output by the resource matching degree determination logic step as the starting point. If a fresh food delivery task has extremely high timeliness requirements, the logistics management system will prioritize the delivery vehicle that is close to the delivery destination and has sufficient current capacity as the initial resource node. After generating the initial path, in order to increase population diversity, the initial path is randomly perturbed. The perturbation methods include randomly exchanging the order of adjacent nodes, adjusting the access time window of the nodes (fluctuating within a reasonable range), and fine-tuning the allocated capacity values. The order of express cabinets B and C in the above initial path is randomly exchanged to generate a new path. After multiple perturbations, the initial population is formed.

[0117] This encoding method can accurately and intuitively represent delivery routes, transforming the actual logistics and delivery problem into a form that can be handled by genetic algorithms. By selecting initial resource nodes based on resource matching degree, the feasibility and a certain degree of optimization of the initial path are guaranteed. Random perturbation operation effectively increases the diversity of the initial population, enabling the algorithm to cover a wider solution space during the search process, reducing the risk of getting trapped in local optima, and increasing the probability of finding the global optimum. The generated initial population serves as the basis for the iterative optimization of the genetic algorithm, and its quality and diversity directly affect the effectiveness of subsequent fitness function evaluation and genetic operations. A high-quality and diverse initial population can provide a good starting point for the genetic algorithm, enabling the algorithm to search for the optimal delivery path more efficiently in subsequent iterations, providing rich path samples for fitness function evaluation, and ensuring that the evaluation results are representative and effective.

[0118] Delivery route planning needs to comprehensively consider multiple objectives, including total delivery time, total travel cost, and resource utilization. Optimizing a single objective can lead to a poor overall solution. The fitness function, as a standard for measuring the quality of a route, needs to be reasonably designed to guide the genetic algorithm to search in a direction that satisfies multiple objectives. A multi-objective fitness function that includes total delivery time, total travel cost, and resource utilization is designed. When calculating the total delivery time, the predicted travel difficulty coefficient of each route segment in the logistics network generation logic steps is used, combined with the actual length of the route segment and the travel speed of the resources, to calculate the time required to pass through each route segment. The total delivery time of the entire route is obtained by summing them up. If a route contains 3 route segments, each with a different travel difficulty coefficient corresponding to a different estimated travel time, with route 1 estimated at 5 minutes, route 2 at 8 minutes, and route 3 at 6 minutes, then the total delivery time is 19 minutes.

[0119] The calculation of total toll cost comprehensively considers physical toll cost, traffic accident cost, and resource competition cost. Physical toll cost is determined based on the length of the route segment and road conditions. Traffic accident cost is estimated based on the probability and impact of abnormal traffic events in the area where the route segment is located. Resource competition cost is assessed based on the number of other delivery tasks and resource occupancy on the route segment at the same time. The total toll cost is obtained by weighting and summing these three costs. If the physical toll cost of a certain route is 10, the traffic accident cost is assessed as 5 due to passing through congested road segments, and the resource competition cost is 3 due to the large number of tasks during that time period, the total toll cost is calculated after setting the corresponding weights.

[0120] Resource utilization is measured by calculating the actual efficiency of resource use during the delivery process. For mobile resources, such as delivery vehicles, the utilization rate is calculated based on the ratio of the actual weight of goods transported to the maximum carrying capacity. For fixed resources, such as parcel lockers, the utilization rate is calculated based on the ratio of the actual number of goods stored to the total capacity. For example, if a delivery vehicle has a maximum carrying capacity of 50kg and actually transports 30kg of goods, then the resource utilization rate is 60%. During the iteration process of the genetic algorithm, the fitness of each individual (i.e., each delivery path) is determined by the values ​​of three objective functions. A combination of non-dominated sorting and crowding distance is used to handle multi-objective problems. Individuals in the population are sorted and selected. Non-dominated sorting divides individuals into different levels, and individuals of the same level are non-dominated to each other. Crowding distance is used to measure the distribution of individuals in the objective space and maintain the diversity of the population. In this way, it is ensured that the algorithm can find a better solution during the search process while maintaining the diversity of the population and avoiding premature convergence.

[0121] Multi-objective fitness functions can comprehensively and holistically evaluate the merits of delivery routes, taking into account key factors in the delivery process. This allows the algorithm to balance multiple objectives during optimization, avoiding the pursuit of a single objective while neglecting other important aspects. The non-dominated sorting and crowding distance processing methods effectively solve the problems of solution selection and diversity preservation in multi-objective optimization, improving the algorithm's search efficiency and solution quality. This helps to find delivery route solutions that better meet actual needs. The evaluation result of the fitness function determines the probability of an individual being selected in the selection operation. Individuals with high fitness are more likely to be retained and inherited by the next generation, which directly guides the search direction of the genetic algorithm, causing it to evolve towards paths with higher fitness. At the same time, the evaluation result also provides a basis for subsequent judgment on whether the algorithm has converged. If the fitness of individuals in the population changes little over several generations, the algorithm can be considered close to convergence, and the iteration can be terminated to enter the generation stage of the main path and backup paths.

[0122] Genetic operations are the core means for genetic algorithms to explore the solution space and find the optimal solution. In delivery route planning, the path must meet practical conditions such as time window constraints and resource capacity constraints. Therefore, constraint maintenance is required during genetic operations to ensure that the generated path is feasible. The tournament selection method is adopted. Each time, a number of individuals (e.g., 5) are randomly selected from the population, their fitness is compared, and the individual with the highest fitness is selected to enter the next generation. During the selection process, in order to avoid the algorithm getting trapped in local optima too early, an elite retention strategy is introduced. A certain percentage (e.g., 10%) of the individuals with the highest fitness in the current population are directly retained to the next generation and do not participate in the selection operation, ensuring that good genes are not lost due to random selection. At the same time, the number of times each individual is selected is recorded. If an individual is selected too many times, it means that the individual has a great advantage in the current population, which will cause the algorithm to converge too quickly. At this time, the probability of the individual being selected in the next selection is reduced to increase the diversity of the population. If an individual is selected in the tournament selection three times in a row, its probability of being selected in the next selection is reduced from the original 20% to 10%.

[0123] Simultaneously, a path segment-based crossover operator is designed. First, two parent paths are randomly selected, and a crossover point is randomly chosen on each path. The portions of the two parent paths after the crossover point are swapped to generate two child paths. However, during the swapping process, path infeasibility may occur, such as violating time window constraints or resource capacity constraints. Therefore, after the crossover operation, the feasibility of the generated child paths is immediately checked. For example, if a resource node cannot reach the corresponding road network node within the specified time window after the swap, a path repair mechanism is initiated. The repair mechanism first attempts to adjust the departure time of the resource node. If the time window requirement is still not met, a new crossover point is selected for the crossover operation. If resource capacity constraints are involved, and the resource demand of a path segment exceeds the remaining capacity after the crossover, a suitable backup resource is selected from the available resource set output by the resource matching degree determination logic step to replace it, ensuring that the path meets the resource capacity requirements. In addition, to increase the diversity of crossover operations, a partial mapping crossover strategy is occasionally adopted. By establishing node mapping relationships, the problem of duplicate nodes that occur during the crossover process is solved, further improving the probability of generating high-quality child paths through crossover operations.

[0124] When performing mutation operations on individuals, one or more gene loci within an individual are randomly selected for mutation. Mutation methods include randomly changing the position coordinates of the gene loci, adjusting the access time window, and modifying the allocated transport capacity values. For example, for a gene locus representing the position coordinates of a delivery vehicle's travel path, it is randomly mutated from the current road segment to another adjacent feasible road segment. After the mutation operation, the feasibility of the path is checked. If the mutated path violates the constraints, it is repaired according to the specific situation. If the time constraint is violated, the travel order of some road segments is replanned or the speed is adjusted; if the resource constraint is violated, resources are reallocated or alternative resources are found. At the same time, in order to balance the impact of mutation operations on the algorithm's search efficiency and convergence, the mutation probability is dynamically adjusted according to the evolutionary state of the population. In the early stages of the algorithm, when the population diversity is high, the mutation probability is appropriately reduced to accelerate the convergence speed of the algorithm. In the later stages of the algorithm, when the population tends to converge, the mutation probability is increased to avoid the algorithm getting stuck in local optima, expand the search range, and find a better solution.

[0125] Throughout the genetic operation, the paths are checked in real time to ensure they meet the time window and resource capacity constraints. For individuals that violate the constraints, penalties are imposed to reduce their fitness. For example, for paths that exceed the time window, the fitness value is reduced by a certain percentage based on the length of the delay, i.e., fitness is reduced by 5% for every 1 minute of delay. Similarly, for paths that exceed the resource capacity limit, corresponding fitness penalties are imposed based on the degree of excess. At the same time, constraints are actively repaired to ensure the feasibility of individuals in subsequent iterations, enabling the genetic algorithm to perform efficient searches within the feasible solution space.

[0126] Diverse genetic operation strategies can effectively explore the solution space. Selection operations preserve superior genes, crossover operations combine advantageous genes, and mutation operations introduce new genes, continuously generating better delivery route solutions. Strict constraint maintenance mechanisms ensure that the generated routes are feasible in actual logistics delivery, avoiding the generation of invalid routes and improving the practicality and reliability of the algorithm. Strategies such as dynamically adjusting mutation probabilities balance the search efficiency and convergence of the algorithm, enabling the algorithm to find high-quality delivery routes within a reasonable time, meeting the requirements of efficiency and accuracy for route planning in community shared logistics management and control systems.

[0127] After genetic operations and constraint maintenance, the fitness of individuals in the population is improved, and the feasibility and optimization of the paths are higher. This provides a rich and high-quality set of candidate paths for subsequent selection of primary and backup paths, increasing the probability of selecting the optimal delivery path from the population. At the same time, the optimized population state also provides a basis for judging whether the algorithm has converged. If the differences between individuals in the population gradually decrease and the fitness tends to stabilize, the algorithm can be considered to be close to convergence, and the iteration can be terminated to enter the generation stage of primary and backup paths, providing a reliable path scheme for the elastic delivery module.

[0128] Community logistics delivery involves numerous uncertainties, including sudden road congestion and vehicle breakdowns. Relying solely on a main route cannot guarantee the smooth completion of delivery tasks; therefore, backup routes need to be generated to allow for timely switching when the main route encounters problems, improving the reliability and stability of the logistics management system. Simultaneously, the generated routes need to be validated to ensure their feasibility and effectiveness in practical applications. After the genetic algorithm's iterative optimization terminates, the chromosome with the highest fitness is selected from the population for decoding into the main route. During decoding, the gene locus information in the chromosome is transformed into the actual delivery path, including the scheduling order of resource nodes, the routes traversed by road network nodes, the access time of each node, and resource allocation. Deterministic verification of the main route is performed using a simulation execution method. The logistics management system uses real-time acquired logistics resource status and road conditions... The system considers the logistics resource status, including the current location and remaining capacity of delivery vehicles and the usage of parcel lockers; the road conditions, including real-time traffic information and abnormal traffic events obtained through the resource perception module; and the task requirements, including cargo weight, volume, and timeliness. The system simulates the execution of a route in a virtual environment. If, during the simulation, a sudden increase in real-time traffic flow on a certain road segment leads to an increase in the estimated travel time, exceeding the task's time window, the logistics control system automatically attempts to adjust the speed of that segment, select alternative routes, or negotiate with relevant resources to adjust the delivery time to eliminate potential conflicts. If, after multiple adjustments, the constraints still cannot be met, the system re-selects the chromosome with the second-highest fitness from the population to decode the main path and verifies it again until a main path that satisfies all constraints is obtained.

[0129] The path with the greatest difference from the main path is selected from elite individuals as the first backup path. Elite individuals are those with high fitness in the population. The calculation of difference considers multiple factors, including path overlap, road segment types, and dependent resources. Path overlap is calculated by determining the proportion of the length of the overlapping road segment between the two paths to the total path length; the lower the proportion, the higher the difference. Road segment types include main roads, secondary roads, and internal roads within the community, with different weights assigned to different road segment types. The difference score of the road segment types traversed by the two paths is calculated. Dependent resources include the number and importance of resource nodes such as delivery stations and parcel lockers that the two paths rely on. Preferably, if one path mainly passes through main roads, while the other path passes through more internal roads within the community, and the delivery stations and parcel lockers they rely on also differ significantly, then the path is considered a backup path. The two routes are highly different. Furthermore, for different types of abnormal traffic events, including road closures, vehicle breakdowns, and severe weather, dedicated backup routes are pre-calculated. When calculating these backup routes, historical abnormal event data and current logistics resource status are considered to analyze the optimal response strategy for different abnormal situations. For road closures, the route with the shortest detour and lowest difficulty is prioritized, while also considering whether the route passes through other high-risk areas. For vehicle breakdowns, the location of nearby available backup vehicles and routes that can quickly complete cargo transfers are considered to ensure timely delivery of goods to their destination. Trigger conditions are set for each backup route; when the main route is delayed by more than 15 minutes or when abnormal events such as road closures occur on sections of the main route, the corresponding second backup route is triggered.

[0130] By selecting the path with the highest adaptability as the primary path and conducting rigorous verification, it is ensured that the primary path can efficiently and reliably complete delivery tasks under normal circumstances, meeting the timeliness and cost requirements of logistics delivery. Diverse backup path generation strategies pre-calculate dedicated backup paths for different abnormal events and set reasonable trigger conditions, enabling the logistics management system to quickly switch to the appropriate path when facing various emergencies. This significantly improves the anti-interference and emergency handling capabilities of the logistics management system, reduces the risk of delivery delays or failures due to path problems, and ensures the quality of logistics services and customer satisfaction. The generated primary and backup paths are directly delivered to the elastic delivery module for execution.

[0131] The flexible delivery module determines whether to trigger backup routes based on real-time monitoring of logistics task execution and environmental changes. When a severe congestion occurs on a section of the main route, causing delivery delays exceeding the backup route trigger conditions, the flexible delivery module immediately calls the corresponding backup route, adjusts the delivery plan, and reschedules resources to ensure that logistics tasks can continue smoothly. At the same time, the delivery routes and resource scheduling generated by the route generation module also provide important basis for the flexible delivery module's task scheduling, resource allocation, and other operations, helping the flexible delivery module to better coordinate logistics resources, improve overall logistics efficiency, and achieve intelligent and efficient management and control of community shared logistics.

[0132] The flexible delivery module is used to receive and process logistics tasks through a blind selection mechanism, deliver the logistics tasks according to the delivery route, and monitor the delivery progress deviation, resource health and environmental mutation index of the logistics tasks during the delivery process to determine whether to update the logistics resource map and delivery route.

[0133] Furthermore, the processing sub-logic for logistics tasks includes:

[0134] After receiving a logistics task, the task characteristics of the logistics task are extracted. The task characteristics include task type, weight and volume, time requirement and address information. The address information is segmented into community area and location coordinates and the information is encrypted in layers.

[0135] Based on the real-time status of mobile and fixed resources in the logistics resource map, a set of available resources is selected. Based on the set of available resources and the resource matching degree, a reinforcement learning algorithm is used to calculate the expected revenue of each resource node in undertaking logistics tasks, so as to determine the resource node that will execute the logistics task.

[0136] Logistics tasks contain diverse information, requiring precise extraction of key features to match resources, while ensuring the security of sensitive information such as addresses during transmission and processing. When the flexible delivery module receives a logistics task, it first initiates an information parsing process. Based on the task type, it quickly determines whether the task belongs to the category of fresh produce delivery, regular parcel delivery, or urgent delivery through a preset keyword matching and classification model. It then uses weight sensors and volume measurement equipment to obtain the weight and volume data of the goods. The timeliness requirements are extracted from the task order to determine the latest delivery time. For address information, geocoding technology is used to segment it into community areas and precise location coordinates.

[0137] After information extraction, a layered encryption strategy is adopted. The community area portion of the address information is initially encrypted using a symmetric encryption algorithm, while the location coordinates are processed using an asymmetric encryption algorithm. Encryption keys are generated and stored in a secure key store. This layered encryption method ensures the security of information at different levels and facilitates flexible decryption and use of partial information according to different needs in subsequent resource scheduling and path planning. Complete and accurate extraction of task features provides a comprehensive data foundation for subsequent resource matching. Layered encryption effectively protects user privacy and logistics information security, avoids the risk of sensitive data leakage, and improves system security and user trust. The extracted and encrypted task feature data provides a basis for screening available resource sets. At the same time, secure information storage and transmission methods ensure the integrity and reliability of data during the resource matching process.

[0138] The status of logistics resources within the community changes in real time. It is necessary to screen out resources that meet the basic conditions of the task and determine the optimal execution resources through quantitative evaluation to improve delivery efficiency and resource utilization. Based on the real-time status information of mobile and fixed resources in the logistics resource map, including location, remaining capacity and equipment operating status, preliminary screening is carried out in combination with task characteristics. For the delivery of heavy goods, small delivery vehicles with insufficient remaining capacity are excluded. For tasks with specific time requirements, only resources whose current location is close to the task starting point and are expected to be completed on time are retained.

[0139] From the selected set of available resources, a reinforcement learning algorithm is used to calculate the expected benefit of each resource node undertaking logistics tasks. The algorithm takes as input the spatial, temporal, and capacity matching degrees between the resource node and the task, as well as the resource's historical delivery performance and current task priority. Historical delivery performance includes on-time rate and customer reviews. By simulating various scenarios of different resources performing tasks, the algorithm learns the benefit feedback of resources under different conditions, evaluating the overall benefits brought by the resource after completing the task, namely, improved delivery timeliness, reduced costs, and increased customer satisfaction. For an urgent delivery task, the algorithm will focus on evaluating whether the resource can... If a delivery vehicle is expected to complete the delivery ahead of schedule and has a lower route cost, its expected benefit will be higher than that of other resources. This allows for precise selection of available resources, narrowing the resource matching range, and improving the efficiency of resource and task matching. Based on reinforcement learning-based expected benefit calculation, multiple factors can be comprehensively considered to dynamically evaluate the merits of resource execution, achieve optimal resource allocation, avoid resource waste, and improve the overall logistics service quality. The identified resource nodes for executing logistics tasks provide a clear basis for subsequent task delivery according to the delivery route. At the same time, the resource selection results will affect the resource status monitoring and route execution during the delivery process.

[0140] To ensure the accurate execution of logistics tasks according to the planned route, and to monitor task progress and resource status in real time, so as to respond to emergencies in a timely manner and ensure the smooth completion of delivery tasks; based on the delivery route provided by the route generation module, the task execution instructions are sent to the selected resource nodes, including detailed information such as departure time, driving route, stop time at each node and cargo loading and unloading requirements. During the delivery process, through GPS positioning devices and sensors (including speed sensors and power sensors) installed on mobile resources and the usage status monitoring system of fixed resources, i.e., express cabinet storage and retrieval records, data such as the location, speed, remaining capacity and equipment operation status of resources, as well as the task execution progress, i.e., the completed delivery nodes and the remaining cargo volume, are obtained in real time.

[0141] The collected real-time data is transmitted to the monitoring center of the elastic delivery module. A visual interface displays the real-time progress and resource status of tasks, marking the location of delivery vehicles on an electronic map in real time. Different colors indicate vehicle operating status, with green representing normal and red representing abnormal. A progress bar shows the completion percentage of the task, clearly presenting the status of delivered and undelivered goods. Monitoring personnel can view task execution details in real time through the interface, promptly identifying potential problems such as vehicles deviating from the planned route, equipment malfunction warnings, and task progress delays. This achieves real-time monitoring of the entire logistics task delivery process, making the delivery process transparent, facilitating timely detection and problem-solving, ensuring efficient execution of tasks according to the planned path, improving delivery accuracy and reliability, reducing the probability of delivery delays and errors, and enhancing customer satisfaction. The data acquired through real-time monitoring provides the foundation for calculating delivery progress deviations and assessing resource health, serving as a crucial basis for determining whether to update the logistics resource map and delivery routes, and providing real-time information support for subsequent decision-making.

[0142] Specifically, the logic for updating the logistics resource map and delivery routes includes:

[0143] The logistics tasks are delivered according to the delivery route, and the deviation between the actual delivery progress and the target progress of each logistics task is calculated to obtain the delivery progress deviation.

[0144] Real-time monitoring of the device status of mobile resources and the usage status of fixed resources, and assessment of resource health using fault tree analysis;

[0145] Integrate abnormal passage events and calculate the environmental mutation index using the entropy weight method;

[0146] The fuzzy control algorithm outputs update decision signals based on delivery progress deviation, resource health, and environmental mutation index to determine whether to update the logistics resource map and delivery route.

[0147] In actual delivery, various factors can affect the delivery progress, causing deviations from the plan. Calculating the deviation value allows for timely monitoring of task execution and determination of whether delivery strategies need to be adjusted. Based on the delivery route and estimated completion time of the task, combined with real-time task execution progress data, the deviation between the actual delivery progress and the target progress of each logistics task is calculated. The delivery route is divided into multiple key nodes, including the starting point, intermediate delivery points, and the destination. An estimated arrival time is set for each node. When resources arrive at a node, the actual arrival time is recorded. By comparing the actual arrival time with the estimated arrival time, the time difference is calculated to measure the delivery progress deviation at that node. If a task is expected to arrive at intermediate delivery point A at 10:00, but actually arrives at 10:15, the delivery progress deviation at node A is 15 minutes.

[0148] The calculation of the overall delivery progress deviation for the task comprehensively considers the deviation of each node. If the task includes multiple delivery nodes, different weights are assigned to the nodes based on their importance and delivery order. The total deviation value is then calculated by weighting the nodes. For example, the starting point and the ending point have relatively high weights, while intermediate delivery points have lower weights. The overall delivery progress deviation of the task is obtained by weighted summation, which more accurately reflects whether the task is actually delayed or advanced. This quantifies the delivery progress deviation, intuitively reflecting whether the logistics task execution deviates from the plan. It helps managers to promptly identify delays in the delivery process, providing clear data support for subsequent decision-making. This allows for appropriate measures to adjust the delivery strategy and ensure the task is completed on time. The delivery progress deviation value, as an important parameter, is input into the fuzzy control algorithm along with resource health and environmental mutation index to participate in the generation of update decision signals, affecting whether to trigger the update operation of the logistics resource map and delivery path.

[0149] The equipment status and usage of logistics resources directly affect the smooth progress of delivery tasks. Regularly assessing resource health can identify potential faults and problems in advance, avoiding delivery interruptions due to resource anomalies. For mobile resources, various sensors installed on the equipment collect real-time operating data, including indicators such as equipment temperature, vibration frequency, and power / fuel consumption. The collected data is compared with the standard parameter range for normal equipment operation. If any indicator exceeds the normal range, the abnormality is recorded and classified according to the degree of abnormality. For example, if the vehicle engine temperature is too high, it is judged as a minor abnormality; if an engine fault alarm occurs, it is judged as a serious abnormality. At the same time, combined with the equipment's historical maintenance records and usage time, fault tree analysis is used to comprehensively assess the equipment's health status and predict the probability and potential impact of future equipment failures.

[0150] For fixed resources, such as parcel lockers, their usage status is assessed by monitoring data such as the number of times the locker door is opened and closed, the utilization rate of internal storage capacity, and equipment operation logs. If the locker frequently experiences problems such as the locker door failing to open and close properly or insufficient storage space, or if the equipment operation log records multiple error operation prompts, its health rating is lowered. By combining the assessment results of mobile and fixed resources, a health score is generated for each resource node to intuitively reflect the current status of the resource. This comprehensive and scientific assessment of the health status of logistics resources allows for the timely detection of potential problems and failure risks, enabling proactive maintenance, repair, or resource scheduling measures to ensure normal resource operation, reduce the probability of delivery task failure due to resource failure, and improve the stability and reliability of the logistics system. The resource health score, as one of the input parameters of the fuzzy control algorithm, together with the delivery progress deviation and environmental mutation index, affects the output of the update decision signal, determining whether the logistics resource map needs to be updated, the resource scheduling strategy adjusted, and the delivery route replanned.

[0151] The community logistics environment is complex and ever-changing. Sudden changes in environmental factors, such as abnormal traffic events, can have a significant impact on delivery. Calculating the environmental mutation index can quantify the degree of environmental change and provide a basis for adjusting delivery strategies. The system integrates abnormal traffic event information identified by the resource perception module, including the type, location, scope of impact, and duration of abnormal traffic events, while also collecting real-time environmental data such as meteorological data and traffic flow data.

[0152] Entropy weighting is used to analyze and process various environmental data. Based on the data's dispersion and its importance to delivery, different environmental factors are assigned corresponding weights. For example, road construction has a significant impact on delivery routes and is assigned a higher weight, while ordinary weather changes have a relatively smaller impact and a lower weight. By calculating the entropy value of each environmental factor, its contribution to environmental mutation is determined, and then an environmental mutation index is calculated. A higher index indicates a more drastic change in the current community logistics environment and a greater impact on delivery tasks. For instance, when a serious traffic accident occurs in a certain area, leading to the closure of multiple roads, accompanied by heavy rain, the environmental mutation index will increase significantly. This transforms complex and variable environmental factors into quantifiable indices, intuitively reflecting the degree of change in the community logistics environment. This helps managers quickly assess the potential impact of environmental changes on delivery tasks, providing a scientific basis for decision-making. This allows for timely adjustments to delivery strategies to mitigate the adverse effects of environmental changes. The environmental mutation index, along with delivery progress deviation and resource health, serves as input to the fuzzy control algorithm, participating in the generation of update decision signals to determine whether the logistics resource map and delivery routes need to be updated to adapt to the changed environmental conditions.

[0153] Based on changes in the actual situation during the delivery process, timely judgments are made as to whether the logistics resource map and delivery route need to be updated to ensure that the logistics management system can flexibly respond to various emergencies and guarantee the efficient completion of delivery tasks. The calculated delivery progress deviation, resource health score, and environmental mutation index are input into the fuzzy control algorithm. In the fuzzy control algorithm, a fuzzy rule base is pre-set. Different decision results are corresponding to different parameter combinations and value ranges. For example, when the delivery progress deviation is large, the resource health is low, and the environmental mutation index is high, a strong update decision signal is output. When all indicators are within the normal range, a no-update signal is output.

[0154] The input precise values ​​are transformed into fuzzy sets through a fuzzy reasoning process. Logical reasoning is performed based on the fuzzy rule base to obtain an update decision signal. If an update decision signal is received, it is first determined whether the logistics resource map, the delivery route, or both need to be updated. For updating the logistics resource map, the attributes of resource nodes are adjusted (e.g., updating the status of faulty resources to unavailable) and the weights of edges (e.g., adjusting the connection weights between resource nodes due to road closures), as well as the structure of the macro and micro layers, based on the resource health assessment results and environmental changes. For updating the delivery route, the route generation module is called. Based on the updated logistics resource map and the current task requirements, the resource matching degree is recalculated, and a new delivery route, including the main route and backup routes, is generated through a genetic algorithm. The new route plan is then sent to the elastic delivery module for execution.

[0155] Based on a fuzzy control algorithm that integrates multiple factors, the system can make flexible and accurate update decisions, enabling the logistics system to adapt quickly to environmental changes and anomalies during task execution. It updates the logistics resource map and delivery routes in a timely manner, ensuring that resource scheduling and route planning are always in the optimal state, improving the efficiency and reliability of logistics and distribution, and reducing operating costs. The updated logistics resource map and delivery routes provide a new basis for the processing of subsequent logistics tasks. The flexible delivery module will execute tasks according to the new routes and continue to monitor task execution and environmental changes, forming a closed-loop management system and continuously optimizing the operation of the community shared logistics management and control system.

[0156] Example 2:

[0157] like Figure 4 The diagram shows a flowchart of a community shared logistics management method provided in this application embodiment. The method includes: acquiring mobile resource data, fixed resource data, and environmental data of the community by deploying edge nodes, and identifying abnormal access events in the community based on the environmental data.

[0158] A logistics resource graph is constructed based on mobile resource data, fixed resource data, and abnormal passage events. Resource nodes represent mobile resources and fixed resources. Node attributes include resource location and transport capacity. Edges between resource nodes represent the spatiotemporal constraint relationship between mobile resources and fixed resources.

[0159] By combining abnormal access events in the community, a logistics access network is generated, and the access difficulty coefficient of each path segment in the logistics access network is predicted by a spatiotemporal attention graph neural network.

[0160] The resource matching degree is determined based on the resource location and transportation capacity in the logistics resource map and the passage difficulty coefficient of each path segment in the logistics passage network.

[0161] Based on resource matching degree, a genetic algorithm is used to generate delivery routes for logistics tasks. The delivery route includes at least a main route and a backup route.

[0162] The system receives and processes logistics tasks through a blind selection mechanism, delivers the tasks according to the delivery route, and monitors the delivery progress deviation, resource health, and environmental mutation index of the logistics tasks during the delivery process to determine whether to update the logistics resource map and delivery route.

[0163] Since the principle of the method in this application embodiment is similar to that of the system described in this application embodiment, the implementation of the method is the same as that of the system, and the repeated parts will not be described again.

Claims

1. A community-shared logistics management and control system, characterized in that: include: By deploying edge nodes, mobile resource data, fixed resource data, and environmental data of the community are obtained. Based on the environmental data, abnormal passage events in the community are identified. A logistics resource graph is constructed based on the mobile resource data, fixed resource data, and abnormal passage events. Resource nodes represent mobile resources and fixed resources. Node attributes include resource location and transport capacity. The edges between resource nodes represent the spatiotemporal constraint relationship between mobile resources and fixed resources. A logistics network is generated by combining abnormal passage events in the community. The passage difficulty coefficient of each path segment in the logistics network is predicted by a spatiotemporal attention graph neural network. The resource matching degree is determined based on the resource location and transportation capacity in the logistics resource map and the passage difficulty coefficient of each path segment in the logistics network. The delivery path of the logistics task is generated by a genetic algorithm based on the resource matching degree. The delivery path includes at least the main path and the backup path. The system receives and processes logistics tasks through a blind selection mechanism, delivers the tasks according to the delivery route, and monitors the delivery progress deviation, resource health, and environmental mutation index of the logistics tasks during the delivery process to determine whether to update the logistics resource map and delivery route.

2. The community shared logistics management and control system as described in claim 1, characterized in that, The logic for constructing the logistics resource map includes: Mobile and fixed resources are encoded using graph convolutional networks to form resource nodes, with resource location and capacity as node attributes, and node attributes are incrementally updated according to changes in resource nodes. The spatial distance between each resource node is calculated based on the resource location. The temporal constraints between each resource node are determined by combining abnormal passage events. The spatial distance and temporal constraints are integrated to obtain the spatiotemporal constraint relationship, and the spatiotemporal constraint relationship is used as the edge between resource nodes. The spatiotemporal correlation between mobile and fixed resources is analyzed through a spatiotemporal attention mechanism, and the weights of the edges are dynamically adjusted according to the spatiotemporal correlation to form a micro-layer. The resource nodes at the micro level are aggregated, with each community area as a super node. The average location and comprehensive transportation capacity of all resource nodes in each community area are calculated as super node attributes. The edges between super nodes are determined based on the traffic distance and travel time between community areas to form a macro level. The macro and micro levels are then integrated to construct a logistics resource graph.

3. The community shared logistics management and control system as described in claim 2, characterized in that, The sub-logic for identifying abnormal access events includes: By deploying edge nodes, environmental data from the community is acquired and preprocessed. This environmental data includes traffic video streams, meteorological data, traffic flow data, and announcement texts. Features of the environmental data are extracted and fused using a feature alignment algorithm to generate fused features. At the same time, homomorphic encryption technology is used to encrypt the fused features. The encrypted fused features are input into a spatiotemporal attention network. The multi-head attention mechanism captures the temporal dependencies of environmental data, and the spatial attention mechanism analyzes the spatial dependencies of environmental data, outputting a spatiotemporal feature vector. Based on historical environmental data, the normal state boundary is adjusted by an adaptive dynamic threshold algorithm, and the spatiotemporal feature vector is processed by a long short-term memory network to predict the environmental state. The predicted environmental state is compared with the normal state boundary to identify abnormal passage events, and abnormal passage events are recorded and classified.

4. The community shared logistics management and control system as described in claim 3, characterized in that, The generation logic of the logistics access network includes: A basic road network is constructed based on community map data. Roads are abstracted as directed edges, and intersections are abstracted as road network nodes. The basic attributes of each road network node include road network location and road network type. The basic attributes of each directed edge include road length, road speed limit, number of lanes, and road type. Map abnormal access events in the community to the underlying road network to adjust the directed edges and their underlying properties; Based on the duration of the abnormal passage event, the basic road network is divided into multiple time slices. The directed edges and their basic attributes in each time slice are updated. The buffer zone in the basic road network is determined based on the impact area of ​​the abnormal passage event, and the basic attributes of the directed edges are updated incrementally within the buffer zone. By combining the resource location and transport capacity of resource nodes in the logistics resource map, the set of road network nodes through the directed edges of the basic road network for each resource node is recorded to generate the logistics passage network.

5. The community shared logistics management and control system as described in claim 4, characterized in that, The logic for generating the delivery route includes: Resource nodes and road network nodes are alternately arranged into chromosomes. Each gene position includes location coordinates, access time window and allocated capacity. Based on resource matching degree, an initial resource node is selected for each logistics task, an initial path is generated, and the initial path is randomly perturbed to generate an initial population. The fitness function is determined based on the total delivery time, total travel cost, and resource utilization rate. The initial path is iteratively optimized through selection, crossover, and mutation operations. After the genetic algorithm iterative optimization terminates, the chromosome with the highest fitness is selected as the main path for decoding. The main path is then subjected to deterministic verification, and the path with the greatest difference from the main path is selected as the first backup path. At the same time, a second backup path is generated based on abnormal passage events to generate delivery paths, and trigger conditions are set for the first and second backup paths.

6. The community shared logistics management and control system as described in claim 5, characterized in that, The prediction sub-logic for the travel difficulty coefficient of each path segment includes: The system extracts the basic attributes of each road network node and each directed edge in the transportation network, receives environmental data and the resource location and capacity of mobile resources, and acquires historical traffic data to form multimodal data. At the same time, it extracts and fuses features from the multimodal data to form multimodal features. Multimodal features are processed by a spatiotemporal attention graph neural network to output temporal attention vectors and spatial attention vectors. The temporal attention vectors and spatial attention vectors are then fused by tensor product to generate a spatiotemporal correlation matrix. Based on the spatiotemporal correlation matrix, the physical passage cost, traffic accident cost, and resource competition cost of each path segment in the logistics passage network are determined to comprehensively predict the passage difficulty coefficient of each path segment.

7. The community shared logistics management and control system as described in claim 6, characterized in that, The sub-logic for determining the resource matching degree includes: Spatial matching degree is determined based on the resource location of resource nodes to the road network location of road network nodes; time matching degree is determined based on the time constraints between resource nodes and the physical passage cost of each path segment in the logistics network; and capacity matching degree is determined by comparing the transport capacity of resource nodes with the resource competition cost of each path segment in the logistics network. The weights of spatial matching, temporal matching, and capacity matching dimensions are dynamically adjusted based on the type and severity of abnormal passage events. The resource matching degree is determined by weighted summation based on the dimension weights.

8. The community shared logistics management and control system as described in claim 7, characterized in that, The update logic for the logistics resource map and delivery route includes: The logistics tasks are delivered according to the delivery route, and the deviation between the actual delivery progress and the target progress of each logistics task is calculated to obtain the delivery progress deviation. Real-time monitoring of the device status of mobile resources and the usage status of fixed resources, and assessment of resource health using fault tree analysis; Integrate abnormal passage events and calculate the environmental mutation index using the entropy weight method; The fuzzy control algorithm outputs update decision signals based on delivery progress deviation, resource health, and environmental mutation index to determine whether to update the logistics resource map and delivery route.

9. The community shared logistics management and control system as described in claim 8, characterized in that, The processing sub-logic for the logistics task includes: After receiving a logistics task, the task characteristics of the logistics task are extracted. The task characteristics include task type, weight and volume, time requirement and address information. The address information is segmented into community area and location coordinates and the information is encrypted in layers. Based on the real-time status of mobile and fixed resources in the logistics resource map, a set of available resources is selected. Based on the set of available resources and the resource matching degree, a reinforcement learning algorithm is used to calculate the expected revenue of each resource node in undertaking logistics tasks, so as to determine the resource node that will execute the logistics task.

10. A community-shared logistics management and control method, implemented based on the community-shared logistics management and control system according to any one of claims 1-9, characterized in that, include: By deploying edge nodes, mobile resource data, fixed resource data, and environmental data of the community are obtained, and abnormal access events in the community are identified based on the environmental data; A logistics resource graph is constructed based on mobile resource data, fixed resource data, and abnormal passage events. Resource nodes represent mobile resources and fixed resources. Node attributes include resource location and transport capacity. Edges between resource nodes represent the spatiotemporal constraint relationship between mobile resources and fixed resources. By combining abnormal access events in the community, a logistics access network is generated, and the access difficulty coefficient of each path segment in the logistics access network is predicted by a spatiotemporal attention graph neural network. The resource matching degree is determined based on the resource location and transportation capacity in the logistics resource map and the passage difficulty coefficient of each path segment in the logistics passage network. Based on resource matching degree, a genetic algorithm is used to generate delivery routes for logistics tasks. The delivery route includes at least a main route and a backup route. The system receives and processes logistics tasks through a blind selection mechanism, delivers the tasks according to the delivery route, and monitors the delivery progress deviation, resource health, and environmental mutation index of the logistics tasks during the delivery process to determine whether to update the logistics resource map and delivery route.

Images