Emergency material scheduling system and material scheduling method based on ant colony algorithm

Abstract

The invention discloses an emergency material intelligent scheduling system and method based on an ant colony algorithm. A demand prediction module of the emergency material intelligent scheduling system dynamically predicts material demands by using a time-space diagram neural network, and constructs a hierarchical network model integrating multi-modal transportation of unmanned aerial vehicles, ground vehicles and the like. And the optimization calculation module adopts an improved ant colony algorithm, introduces a demand urgency degree weight factor and a green weight factor to construct a multi-objective fitness function, and optimizes a transportation path in combination with a dynamic pheromone updating mechanism and a multi-ant colony collaborative strategy. The system is equipped with an edge computing driven dynamic adjustment module to realize 30-second fast path re-planning, a psychological assistance priority model is innovatively integrated, and a psychological crisis index optimization scheduling strategy is extracted through sentiment analysis. According to the method, the problems of response lag and insufficient multi-objective optimization of a traditional scheduling system are effectively solved, the transportation efficiency is remarkably improved by 25%-35%, carbon emission is reduced, and high efficiency, fairness and humanity care of emergency scheduling are guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of emergency material dispatching and intelligent optimization, and in particular to an emergency material intelligent dispatching system based on an ant colony algorithm. Background Art

[0002] Current emergency material dispatch systems often use traditional path planning algorithms (such as Dijkstra and A* algorithms) or rely on manual decision-making based on empirical experience. These systems suffer from the following technical flaws: insufficient dynamic response capabilities; static models are unable to adapt to complex scenarios such as post-disaster road damage and sudden changes in demand; poor path planning timeliness, and an inability to update optimal solutions in real time; a lack of multi-objective optimization; existing solutions often focus on a single objective (such as the shortest path); and are weak in co-optimizing transportation costs, time efficiency, resource allocation fairness, and environmental factors. In particular, they lack mechanisms for quantifying carbon emissions and guiding low-carbon routes. A single mode of transportation, relying on a single means of transport (such as vehicles), results in inefficient material delivery in complex terrain, and newer transport tools such as drones and robots lack coordinated dispatching rules. Psychological needs are neglected; existing technologies fail to incorporate the psychological crisis index of disaster-stricken areas into dispatching priorities, and the allocation of medical supplies and psychological support resources lacks humanistic considerations. Decision delays are significant, and centralized cloud-based computing architectures result in long response times, with local route adjustments taking over three minutes, making them unable to meet the real-time requirements of emergency scenarios.

[0003] For example, patent CN201810321384.5 proposes an emergency path planning method based on a genetic algorithm. While this method incorporates dynamic road condition updates, it fails to address the issues of multi-transportation coordination and carbon emission optimization. The paper "Research on Fairness Target Assessment and Site Allocation Model in Disaster Emergency Operation Management" (Management Science and Engineering, 2020) takes minimizing total social cost (the sum of logistics costs and psychological costs of suffering) as the emergency rescue goal and constructs an integrated optimization model that considers disaster evacuation and resettlement, temporary shelter site selection, and material distribution. However, the need for psychological assistance remains unquantified. Therefore, there is an urgent need for an intelligent scheduling solution that integrates multimodal transportation, low-carbon sensing, real-time decision-making, and psychological assistance prioritization to improve the overall effectiveness of emergency response. Summary of the Invention

[0004] The present invention proposes an intelligent dispatching system and method for emergency supplies based on ant colony algorithm, which realizes efficient, green and humane emergency supply dispatching by integrating multimodal transport collaboration, carbon footprint perception, edge computing real-time decision-making and psychological assistance priority model.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0006] The present invention first provides an emergency material dispatching system based on ant colony algorithm, comprising:

[0007] A path modeling module, which is used to build a multi-modal hierarchical transportation network model including road traffic capacity constraints. The multi-modal hierarchical transportation network model integrates drones, ground transportation vehicles, and unmanned delivery robots, and describes the load, speed, energy consumption, and reachability cooperation relationships of different transportation tools through a mixed integer programming model;

[0008] The multi-modal hierarchical transportation network model includes:

[0009] The ground layer network G ground =(V g , E g , W g ) and the air layer network G air =(V a , E a , W a ), where:

[0010] V g is the set of ground layer network nodes, V a is the set of air layer network nodes. The node set includes warehouses, demand points, and transfer stations; E g is the set of edges of the passable paths in the ground layer network, E a is the set of edges of the passable paths in the air layer network; W g is the weight matrix of the ground layer network, W a is the weight matrix of the air layer network. The weight matrix includes time, cost, and carbon emission attributes;

[0011] Cross-layer connection points

[0012] Transport tool performance constraints, including:

[0013] Load constraint: Σ m∈路径 q m ≤Q k

[0014] where Q k is the maximum load of tool k, and q m is the material demand of path node m;

[0015] Range constraint:

[0016] Σ e∈路径 d e ·ξ k ≤D k

[0017] where D k is the maximum endurance of tool k, d e is the length of edge e, and ξ k is the unit distance energy consumption coefficient of transport tool k;

[0018] Terrain constraint:

[0019] The maximum slope max of the robot path e∈路径 θ e ≤15°;

[0020] The mixed-integer programming model includes:

[0021] Objective function:

[0022]

[0023] Where, T e,k is the transportation time of vehicle k on edge e; C e,k is the transportation cost of vehicle k on edge e; E e,k The carbon emission of vehicle k on edge e; x e,k is a decision variable, x e,k ∈{0,1}, taking values 0 or 1, indicating whether to select vehicle k for transportation on edge e; w T , w C , w E are weight coefficients, representing the optimization priorities of time, cost, and carbon emission respectively, w T +w C +w E =1;

[0024] Cross-layer collaboration constraint:

[0025] yv,k indicates whether node v switches the transportation vehicle K, y v,k ∈{0,1}, when taking the value 0, the original vehicle is maintained, and when taking the value 1, it switches to vehicle k at node v;

[0026] The optimization calculation module uses an improved ant colony algorithm for multi-objective path planning. The improved ant colony algorithm includes a dynamic pheromone update mechanism, a multi-objective fitness function, a tabu search hybrid strategy, and a multi-ant colony cooperation strategy, where:

[0027] The dynamic pheromone update mechanism introduces a demand urgency weight factor U r :

[0028]

[0029] Where, P r is the regional population density obtained from the government population database, M r is the casualty growth rate statistically obtained from the hospital emergency system in real time, R rThe material satisfaction rate is the ratio of the delivered quantity to the demand forecast value, where α1 + α2 + α3 = 1, and α1, α2, and α3 represent the weight coefficients of regional population density, casualty growth rate, and material satisfaction rate respectively, P max , M crit , R req are the preset thresholds of regional population density, casualty growth rate, and material satisfaction rate respectively;

[0030] The multi-objective fitness function synchronously optimizes the transportation time, cost, fairness, and carbon emission indicators, and introduces a green weight factor G into the path selection probability ij ;

[0031] The dynamic adjustment module, based on the edge computing node and the federated learning framework, responds to road interruptions or traffic anomalies in real time, and completes local path replanning within a set time through a lightweight ant colony algorithm.

[0032] The path selection probability formula is:

[0033]

[0034] where is the probability that vehicle k selects path (i, j) at time t; τ ij (t) is the pheromone concentration of path (i, j) at time t; η ij is the heuristic factor of path (i, j); τ is (t) is the pheromone concentration of path (i, s) at time t, where s is the allowed next node; G ij is the green weight factor of path (i, j); U r is the urgency weight factor of node r; α, β, γ, δ are weight adjustment parameters respectively;

[0035] For the denominator part of the path selection probability, calculate the total attraction for all allowed nodes s ∈ allowed k where allowed k is the set of next nodes that vehicle k is allowed to access when it is at the current location.

[0036] allowed k is dynamically generated according to the load, endurance, and road interruption status of vehicle k, and the generation process includes the following steps:

[0037] (a) Obtain all adjacent nodes of the current node i to form an initial candidate set;

[0038] (b) Based on the real-time load ∑qm of vehicle k and the maximum load Qk, filter out over-limit

[0039] nodes;

[0040] (c) Calculate the remaining endurance Dkremaining = Dk-∑de·∈k, excluding nodes with excessive distance;

[0042] (d) Through the road access score Sroad(e) and the urgency threshold Umin, the secondary filtering is low

[0043] Priority nodes;

[0044] (e) Output the final allowed k Set, used for path selection probability calculation.

[0045] Road access score threshold Sroad = 0.3;

[0046] Threshold source: According to historical disaster data statistics, when Sroad < 0.3, the road failure rate exceeds 80%;

[0047] Dynamic updates: Get the latest traffic and damage data every 5 minutes through the traffic management department API.

[0048] Minimum urgency threshold Umin=0.2:

[0049] Basis for setting: The demand forecast model shows that the material shortage rate in areas with Ur < 0.2 is less than 10%, and processing can be postponed;

[0050] Adjustment rule: If the global average urgency Uˉ>0.6, Umin is automatically increased to 0.4.

[0051] τ ij Initial value of (t)

[0052] τ ij (t+1)=(1-ρ)τ ij (t)+Δτ ij

[0053] Where ρ is the pheromone volatility, ρ = 0.05; Δτ ij The increment of pheromone released by all ants passing through path (i, j) in this round;

[0054] Δτ ij The calculation formula is:

[0055]

[0056] Where Z = 100 represents the pheromone intensity coefficient; L k represents the total cost of the path of the ath ant; is a binary indicator function. When ant a passes through path (i, j), otherwise is 0;

[0057] The heuristic factor η of path (i, j) ij is:

[0058]

[0059] where d ij is the distance of path (i, j), obtained from the map database, and v k is the speed of vehicle k;

[0060] The green weight factor G of path (i, j) ij is:

[0061]

[0062] where E ij is the total carbon emission of path (i, j).

[0063] The multi-ant colony cooperation strategy includes:

[0064] Allocating independent ant populations to drones, ground vehicles, and unmanned delivery robots respectively;

[0065] Cross-population pheromone interaction mechanism, realizing global resource allocation through pheromone sharing;

[0066] Transport vehicle exclusive constraint conditions, including drone flight range limit, vehicle load capacity limit, and robot terrain adaptation rules.

[0067] The path modeling module further includes:

[0068] Dynamic obstacle processing sub-module, which fuses satellite remote sensing and traffic monitoring data in real time to perform topological reconstruction on road interruption areas;

[0069] Multi-modal cooperation sub-module, which dynamically allocates transport vehicles according to disaster scenarios and preferentially enables drones to deliver to high-urgency areas.

[0070] In the psychological assistance priority module, the psychological assistance priority module processes social media and distress call text data through sentiment analysis methods to generate a psychological crisis index Q r Obtained by analyzing the sentiment of social media text, and weighted into the demand urgency weight factor; the demand urgency weight factor is extended to:

[0071]

[0072] where α1 + α2 + α3 + α4 = 1, Q j , α4 are the psychological crisis index and weight coefficients respectively.

[0073] The emergency material scheduling system based on the ant colony algorithm further includes:

[0074] A visual interaction terminal, which uses 3D and AR technologies to superimpose and display the scheduling plan and road conditions in real time, and provides a multi-strategy comparison view;

[0075] A digital twin and blockchain module, which records the entire process of scheduling decisions to ensure traceability and data integrity.

[0076] The present invention also provides an emergency material scheduling method, including the following steps:

[0077] Step S1: Obtain historical disaster situations, meteorological conditions, public opinions, and real-time traffic data through an external interface, and use a spatio-temporal graph neural network to predict the material demand in the next 6 - 72 hours;

[0078] Step S2: Construct a multi-modal hierarchical transportation network model, and define road traffic constraints, transportation tool performance parameters, and multi-modal cooperation rules;

[0079] Step S3: Use an improved ant colony algorithm to iteratively optimize the path. When initializing pheromones, they are differentially distributed according to road grades. After each round of iteration, update the pheromones through an elite strategy, and perform cross-ant colony resource coordination;

[0080] Step S4: Display the scheduling plan through an AR visualization terminal, supporting manual adjustment and comparison of multi-objective plans;

[0081] Step S5: During transportation, use edge computing nodes to monitor road conditions in real time. When the road is abnormal, trigger local path replanning, and the response time ≤ 30 seconds;

[0082] Step S6: Record the scheduling decision data through the blockchain and synchronously update it with the digital twin model.

[0083] In the above step S3, the multi-objective function for path optimization is:

[0084] Minimize(ω1·T + ω2·C + ω3·(1 - F) + ω4·E)

[0085] Wherein, T is the transportation time; C is the cost; F is the fairness index; E is the carbon emission; ω1 to ω4 are weight coefficients and Σω1 + ω2 + ω3 + ω4 = 1; ω1 to ω4 are set according to manual experience, ω1 = 0.5, ω2 = 0.2, ω3 = 0.2, ω4 = 0.1 (the golden 72 hours of rescue), ω1 = 0.3, ω2 = 0.3, ω3 = 0.2, ω4 = 0.2 (mid-term recovery stage), ω1 = 0.2, ω2 = 0.2, ω3 = 0.3, ω4 = 0.3 (late reconstruction stage).

[0086]

[0087] d e is the length of edge e, and v k is the speed of vehicle k;

[0088]

[0089] μ k is the fuel consumption rate of vehicle k; p is the unit price of fuel;

[0090]

[0091] R i is the material satisfaction rate of the c-th demand order, which is the ratio of the received quantity to the demanded quantity; n is the total number of demand points;

[0092]

[0093] is the unit-distance carbon emission coefficient of vehicle k.

[0094] The local path replanning includes:

[0095] Deploying a lightweight ant colony algorithm on the vehicle side through a federated learning framework;

[0096] Only performing local search on adjacent nodes of the affected path, restricting the number of iterations ≤ 50 times;

[0097] Asynchronously synchronizing the replanning result with the global model in the cloud.

[0098] The digital twin model includes:

[0099] Real-time mapping of the vehicle status, including position, load capacity, and remaining energy;

[0100] Dynamically updating the road traffic capacity score S road , when Sroad < 0.3, triggering topological reconstruction; removing the interrupted edge e and adding a detour path, or forcibly switching to the drone layer for transportation;

[0101]

[0102] where λ1 + λ2 = 1.

[0103] The present invention has the following beneficial effects:

[0104] The dynamic adjustment time consumption is compressed from the traditional 3 minutes to 30 seconds, thanks to the lightweight ant colony algorithm and federated learning framework driven by edge computing. By performing local path search at the transportation end side (within a 3-layer range around the affected nodes, with ≤ 50 iterations), the high latency of global computing in the cloud is avoided; the global optimization efficiency is increased by 40% due to the combination of the multi-ant colony cooperation strategy (parallel optimization of independent ant colonies of drones and vehicles) and the tabu search hybrid strategy, reducing ineffective iterations and accelerating convergence; the fairness index of material distribution is increased by 25% - 35% relying on the demand urgency weight factor (integrating population density, casualty rate, material satisfaction rate, and psychological crisis index) and the multi-objective fitness function, forcing a balance in resource allocation between high / low priority areas; the carbon emissions are reduced by 15% - 20% through the green weight factor to guide low-carbon path selection (such as preferentially delivering supplies to mountain disaster areas by drones) and optimizing tool scheduling with the multi-modal transportation cooperation model; the accuracy rate of psychological assistance delivery is ≥ 90% based on the sentiment analysis technology of the BERT model to accurately identify psychological crisis signals in social media and dynamically weight them into the scheduling decision; the success rate of road interruption topology reconstruction is > 95% by the dynamic obstacle processing sub-module that fuses satellite remote sensing and sensor data in real time and triggers multi-modal path switching within 5 seconds (such as vehicle rerouting + drone overobstacle); the audit traceability efficiency is increased by 50% relying on the blockchain distributed ledger and the digital twin model to achieve second-level data uploading and fast retrieval of the entire scheduling process. The above innovations form a four-layer collaborative architecture of "algorithm - data - transportation - system", systematically solving the problems of response speed, resource fairness, and sustainability in disaster emergency scenarios by reducing latency through edge computing, controlling costs through multi-objective optimization, and increasing humanity through the psychological model. BRIEF DESCRIPTION OF THE DRAWINGS

[0105] Figure 1 is a block diagram of the emergency material scheduling system based on the ant colony algorithm of the present invention;

[0106] Figure 2 is a flowchart of the emergency material scheduling method based on the ant colony algorithm of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0107] The following further elaborates on the present invention in detail with reference to the drawings and embodiments, so that those skilled in the art can implement it according to the description in the specification.

[0108] The composition of the emergency material scheduling system based on the ant colony algorithm is as Figure 1 shown, including:

[0109] 1. Demand perception module

[0110] Collects environmental data (such as meteorological and geological information) and material demand information at the disaster-stricken point in real time, and generates the demand urgency weight factor U r through the analysis of multi-source data (social media, distress signals). Its calculation formula is:

[0111]

[0112] Among them, P r is the regional population density, M r is the casualty growth rate, R r is the material satisfaction rate, and α1 + α2 + α3 = 1

[0113] , P max , M crit , R req are the preset thresholds respectively.

[0114] Preferably, a psychological crisis index is further introduced into the demand urgency weight factor, and the extended formula is:

[0115]

[0116] Among them, Q r uses the BERT model for text sentiment analysis, and the output range is 0 - 1 (1 represents the highest crisis level), and α1 + α2 + α3 + α4 = 1;

[0117] 2. Path Modeling Module

[0118] The path modeling module includes a multimodal collaboration sub-module. The multimodal collaboration sub-module constructs a multi-dimensional transportation network model including road traffic capacity constraints, and integrates drones, ground vehicles, and unmanned delivery robots;

[0119] Preferably, the path modeling module also includes a road traffic capacity evaluation sub-module, which fuses historical data and real-time traffic flow monitoring to dynamically calculate the road traffic score S road :

[0120]

[0121] Preferably, the path modeling module also includes a dynamic obstacle handling sub-module, which uses satellite remote sensing data to identify road interruption areas in real time and automatically reconstructs the topology of the transportation network.

[0122] 3. Optimization Calculation Module

[0123] An improved ant colony algorithm is used for multi-objective path planning. The specific improvements include:

[0124] Dynamic pheromone update mechanism: Introduce the demand urgency weight factor U j , and adjust the pheromone concentration to give priority to serving high-urgency areas;

[0125] Multi-objective fitness function: Synchronously optimize the transportation time T, cost C, fairness F, and carbon emissions E. The objective function is:

[0126] Minimize(0.35·T + 0.35·C + 0.15·(1 - F) + 0.15·E)

[0127] Preferably, the weight coefficients can be dynamically adjusted according to the disaster situation stage (for example, the time weight is increased to 0.5 in the initial stage of rescue);

[0128] Path selection probability formula:

[0129]

[0130] Preferably, the influence coefficient γ of demand urgency ranges from 0.5 to 1.2, and 0.9 is preferably adopted to balance efficiency and fairness.

[0131] 4. Dynamic adjustment module

[0132] Periodic global path replanning: Update the global plan every 30 - 60 minutes. Preferably, cloud-edge collaborative computing is adopted, and the global optimization time consumption ≤ 5 minutes;

[0133] Quick response to sudden road interruptions: Execute the lightweight ant colony algorithm through the edge computing node. Preferably, only local search is performed on the adjacent three layers of nodes of the affected path, the number of iterations ≤ 50 times, and the adjustment time consumption is compressed within 30 seconds.

[0134] 5. Visualization interaction terminal

[0135] Adopt 3D and AR technologies to overlay and display the real-time road conditions and dispatching plans. Preferably, distinguish the road traffic status by color coding (red: interrupted, yellow: congested, green: unobstructed);

[0136] Preferably, provide a multi-strategy comparison view, supporting quick switching of preset plans according to three groups of weights of time, cost, and fairness.

[0137] Digital twin and blockchain module

[0138] The digital twin model maps the position, cargo volume, and remaining energy of the transportation vehicle in real time, and the update frequency is 10 seconds / time;

[0139] The blockchain records the key operations of dispatching decisions (such as path changes, manual interventions), and the data upload delay to the blockchain ≤ 2 seconds.

[0140] An emergency material dispatching method, as Figure 2 shown, includes:

[0141] Step S1: Disaster situation data collection and demand urgency calculation

[0142] Synchronize meteorological warning data every hour through the API interface, and extract keywords in social media through natural language processing technology (such as "urgently needed drugs", "water cut");

[0143] Psychological crisis index Q j is calculated using the BERT model for text sentiment analysis, and the output range is 0 - 1 (1 represents the highest crisis level).

[0144] Step S2: Construction of transportation network model and loading of constraints

[0145] The delivery range of drones is set to a radius of 50 km, the load limit of ground vehicles is 5 tons, and the limited slope of unmanned delivery robots is ≤ 15°;

[0146] The carbon emission constraint is mapped through the type of transportation tool (drone: 200 g CO2 / km·kg, vehicle: 150 g CO2 / km·kg).

[0147] Step S3: Iterative optimization of improved ant colony algorithm

[0148] Differentiated initialization of pheromone: Set the initial value according to the road grade. The initial pheromone concentration of better highways is 1.5 times that of ordinary roads;

[0149] After each round of iteration, the elite ant strategy is adopted to add 20% pheromone to the optimal path additionally.

[0150] Step S4: Visual presentation of the plan and manual confirmation

[0151] The AR display terminal supports gesture operations to adjust path nodes, and the corrected plan is automatically synchronized to the blockchain;

[0152] The plan comparison view provides side-by-side display of indicators such as the fairness Gini coefficient and total carbon emissions.

[0153] Step S5: Dynamic monitoring of transportation process and path adjustment

[0154] When a sudden road interruption occurs, the system automatically sends a re-planned path to the affected vehicles, which is executed after the driver confirms;

[0155] The edge computing node adopts a federated learning framework, and the local models of each vehicle are synchronized with the cloud every 10 minutes.

[0156] Embodiment

[0157] 1. Data collection and demand prediction

[0158] Step 1.1: Multi-source data access

[0159] Obtain real-time disaster warning data (such as typhoon path, earthquake intensity) through the meteorological bureau API, and the update frequency is once per minute;

[0160] Access the public data streams of social media platforms (Douyin, Weibo), and use the Flume tool to collect real-time distress posts with geographical tags;

[0161] Call the road condition sensor data (traffic flow, road damage status) of the traffic management department and transmit it to the edge computing node.

[0162] Step 1.2: Calculation of demand urgency

[0163] Population density P r : Extract the gridded data of the affected area from the government population database and divide it into units of 1km×1km;

[0164] Casualty growth rate M r : Real-time count the number of newly added wounded through the hospital emergency system and calculate the hourly growth rate;

[0165] Material satisfaction rate R r : Dynamically update based on the ratio of the inventory data in the material warehouse to the demand forecast value;

[0166] Psychological crisis index Q r : Use a pre-trained BERT model to perform sentiment analysis on social media texts and output a crisis score ranging from 0 to 1 (for example, keywords such as "desperate" and "suffocating" trigger Q r >0.8).

[0167] Calculation formula:

[0168] 2. Multimodal transportation network modeling

[0169] Step 2.1: Hierarchical network construction

[0170] Ground layer: Based on the OpenStreetMap road network data, construct a traffic network for vehicles and robots with a node spacing ≤ 500m;

[0171] Air layer: Design 3D flight routes for drones, avoiding no-fly zones and high-rise buildings, with a node spacing ≤ 2km;

[0172] Constraint loading:

[0173] Vehicle load ≤ 5 tons, speed 80km / h, and the energy consumption model is Ecar = 0.15 · load (tons) · distance (km);

[0174] The maximum flight range of the drone is 50km, the load is 20kg, and the carbon emission coefficient is 200g CO2 / km·kg;

[0175] The limited slope of the robot ≤ 15°, and the speed is 5 - 15km / h.

[0176] Step 2.2: Dynamic obstacle handling

[0177] Identify the road collapse area through satellite remote sensing images (resolution 0.5m), and use the YOLOv5 model to detect the damage points;

[0178] Automatically trigger topological reconstruction: If the traffic capacity score Sroad of a certain section < 0.3 (the calculation formula is shown in the invention content), then mark this section as interrupted and reconnect the adjacent available nodes.

[0179] 3. Improved ant colony algorithm path optimization

[0180] Step 3.1: Algorithm initialization

[0181] Pheromone setting: The initial concentration of the highway is τ0 = 1.5, and that of the ordinary road is τ0 = 1.0;

[0182] Parameter configuration: α = 1.2, β = 0.8, γ = 0.9, δ = 0.6 (green weight factor);

[0183] Ant colony distribution:

[0184] Vehicle ant colony: 100 ants, each representing the feasible path of a vehicle;

[0185] UAV ant colony: 50 ants, each representing a UAV flight path.

[0186] Step 3.2: Iterative optimization process

[0187] Path selection: Each ant selects the next node according to the probability formula:

[0188]

[0189] (Gij is calculated from the transportation tool type and the path slope. For example, for a vehicle on a flat road, Gij = 1.0, and when the slope > 10%, Gij = 0.7)

[0190] Pheromone update:

[0191] After each round of iteration, the elite ants (the top 10% of the optimal paths) release a pheromone increment Δτ = 0.2;

[0192] Ordinary ants release Δτ = 0.1, and the evaporation rate ρ = 0.05.

[0193] Tabu search: Conduct a local search on the 5 adjacent nodes of the current optimal path and replace the sub-paths that can be optimized.

[0194] Step 3.3: Multi-objective fitness evaluation

[0195] Calculate the comprehensive score of each path:

[0196] S core =(0.35·T + 0.35·C + 0.15·(1 - F) + 0.15·E)

[0197] T: Travel time (hours);

[0198] C: Cost (yuan) = fuel consumption cost + labor cost;

[0199] F: Fairness index (Gini coefficient, 0 - 1);

[0200] E: Carbon emissions (kg CO2).

[0201] 4. Dynamic adjustment driven by edge computing

[0202] Step 4.1: Global replanning (every 60 minutes)

[0203] The cloud server performs full network path optimization, and after generating a new plan, synchronizes it to the edge nodes;

[0204] Time consumption control: Through Spark distributed computing, the optimization time consumption for a scale of 1000 nodes ≤ 5 minutes.

[0205] Step 4.2: Local rapid adjustment (sudden exception)

[0206] Trigger condition: The road condition sensor detects that the traffic capacity score of a certain section Sroad < 0.2;

[0207] Response process:

[0208] The edge node extracts the current positions of the affected vehicles and the surrounding 3 - layer nodes (within a range of about 5 km);

[0209] Start the lightweight ant colony algorithm (iteration ≤ 50 times), and generate a new path within 30 seconds;

[0210] Send the new path to the vehicle navigation terminal through the MQTT protocol, and execute it after the driver confirms.

[0211] 5. Visualization and blockchain record

[0212] Step 5.1: Use the Unity engine to render the scheduling plan for AR three - dimensional display, and the AR glasses overlay and display:

[0213] Red: Interrupted road;

[0214] Green: Optimal path;

[0215] Blue: Drone delivery route;

[0216] Gesture operation support: Finger sliding to adjust path nodes, double - click to view material details.

[0217] Step 5.2: Blockchain Evidence Preservation

[0218] Using the Hyperledger Fabric framework, each operation (such as route change, manual intervention) generates a transaction and is uploaded to the blockchain;

[0219] The data fields include: timestamp, operator ID, pre / post-change route hash values;

[0220] During auditing, it can be retrieved through the block explorer according to the time range, and the average query latency ≤ 2 seconds.

[0221] 6. Test Results of the Embodiment

[0222] Taking the disaster-stricken areas in the Yangtze River Basin (1000 demand points, 200 transportation tools) as an example:

[0223] Response time: The average time taken for local route adjustment is 28 seconds (182 seconds for the traditional solution);

[0224] Carbon emissions: Carbon emissions per ton-kilometer are reduced from 120 kg to 96 kg (a 20% reduction);

[0225] Fairness: The Gini coefficient is optimized from 0.48 to 0.31 (a 35% improvement);

[0226] Delivery accuracy: Psychological assistance supplies cover 98% of the high-risk areas (Qj > 0.8Qj > 0.8).

[0227] Although the embodiments of the present invention have been disclosed as above, it is not limited to the applications listed in the specification and embodiments. It can be fully applied to various fields suitable for the present invention. For those familiar with the field, additional modifications can be easily made. Therefore, without departing from the general concept defined by the claims and the equivalent scope, the present invention is not limited to the specific details and the examples shown and described herein.

Claims

1. An emergency material scheduling system based on the ant colony algorithm, characterized in that, Including: A path modeling module for constructing a multi-modal hierarchical transportation network model that includes road traffic capacity constraints. The multi-modal hierarchical transportation network model integrates drones, ground transportation vehicles, and unmanned delivery robots, and describes the load, speed, energy consumption, and reachability cooperation relationships of different transportation tools through a mixed-integer programming model; The multi-modal hierarchical transportation network model includes: Ground layer network G ground =(V g , E g , W g ) and the air layer network G air =(V a , E a , W a ), where: V g is the set of ground - layer network nodes, V a is the set of air - layer network nodes, and the node set includes warehouses, demand points, and transfer stations; E g is the edge set of the passable paths in the ground - layer network, E a is the edge set of the passable paths in the air - layer network; W g is the weight matrix of the ground - layer network, W a is the weight matrix of the air - layer network, and the weight matrix includes time, cost, and carbon - emission attributes; cross - layer connection points Transportation tool performance constraints, including: Load constraint: ∑ m∈路径 q m ≤Q k Among them, Q k is the maximum load of the vehicle k, and q m is the material demand of the path node m; Range constraint: ∑ e∈路径 d e ·ξ k ≤D k Among them, D k is the maximum endurance of the vehicle k, d e is the length of the edge e, and ξ k is the energy consumption coefficient per unit distance of the vehicle k; Terrain constraint: Maximum slope max of the robot path e∈路径 θ e ≤ 15°; The mixed-integer programming model includes: Objective function: Among them, T e,k is the transportation time of vehicle k on edge e; C e,k is the transportation cost of vehicle k on edge e; E e,k is the carbon emission of vehicle k on edge e; x e,k is a decision variable, x e,k ∈{0,1}, and its value is 0 or 1, indicating whether to select vehicle k for transportation on edge e; w T , w C , w E are weight coefficients, representing the optimization priorities of time, cost, and carbon emission respectively, w T + w C + w E = 1; Cross-layer cooperation constraint: where yv,k indicates whether node v switches the transportation vehicle K, and y v,k ∈ {0, 1}. When the value is 0, the original vehicle is maintained, and when the value is 1, it switches to transportation vehicle k at node v; An optimization calculation module that performs multi-objective path planning using an improved ant colony algorithm. The improved ant colony algorithm includes a dynamic pheromone update mechanism, a multi-objective fitness function, a tabu search hybrid strategy, and a multi-ant colony cooperation strategy, where: The dynamic pheromone update mechanism introduces a demand urgency weight factor U r : Among them, P r is the regional population density obtained from the government population database, M r is the casualty growth rate statistically obtained in real time from the hospital emergency system, R r is the material satisfaction rate, which is the ratio of the delivered quantity to the demand forecast value. α1 + α2 + α3 = 1, and α1, α2, and α3 respectively represent the weight coefficients of the regional population density, casualty growth rate, and material satisfaction rate. P max , M crit , R req are respectively the preset thresholds of the regional population density, casualty growth rate, and material satisfaction rate; The multi-objective fitness function synchronously optimizes the transportation time, cost, fairness, and carbon emission indicators, and introduces a green weight factor G into the path selection probability ij ; A dynamic adjustment module, based on edge computing nodes and a federated learning framework, responds to road interruptions or traffic anomalies in real time, and completes local path replanning within a set time through a lightweight ant colony algorithm.

2. The system according to claim 1, wherein In the optimization calculation module, the path selection probability formula is: wherein, is the probability that tool k selects path (i, j) at time t; τ ij (t) is the pheromone concentration of path (i, j) at time t; η ij is the heuristic factor of path (i, j); τ is (t) is the pheromone concentration of path (i, s) at time t, where s is the allowed next node; G ij is the green weight factor of path (i, j); U r is the urgency weight factor of node r; α, β, γ, δ are weight adjustment parameters respectively; For the denominator part of the path selection probability, for all allowed nodes s ∈ allowed k calculate the total attraction, allowed k is the set of next nodes that are allowed to be accessed when the vehicle k is at the current location.

3. The system according to claim 2, wherein allowed k Dynamically generated according to the load capacity, endurance and road interruption status of the vehicle k. The generation process includes the following steps: (a) Obtain all adjacent nodes of the current node i to form an initial candidate set; (b) Based on the real-time load ∑qm of transportation tool k and the maximum load Qk, filter out over-limit nodes; (c) Calculate the remaining endurance Dkremaining = Dk - ∑de·∈k and exclude nodes with over-limit distances; (d) Secondarily filter out low-priority nodes through the road traffic score Sroad(e) and the urgency threshold Umin; (e) Output the final allowed k Set, used for path selection probability calculation.

4. The system according to claim 2, wherein: τ ij The initial value τ of ij (t) is: Pheromone concentration τ of path (i, j) at time t+1 ij (t+1)=is: τ ij (t + 1) = (1 - ρ)τ ij (t) + Δτ ij Among them, ρ is the pheromone evaporation rate, ρ = 0.05; Δτ ij is the increment of pheromone released by all ants passing through the path (i, j) in this round; Δτ ij The calculation formula for Among them, Z = 100 represents the pheromone intensity coefficient; L k represents the total path cost of the a-th ant; is a binary indicator function. When ant a passes through path (i, j), otherwise is 0; The heuristic factor η of the path (i, j) ij is as follows: Among them, d ij is the distance of the path (i, j), obtained from the map database, and v k is the speed of the vehicle k; The green weight factor G of the path (i, j) ij is as follows: Among them, E ij is the total carbon emissions of the path (i, j).

5. The system according to claim 1, wherein: The multi-ant colony cooperation strategy includes: Assign independent ant populations to drones, ground vehicles, and unmanned delivery robots respectively; A cross-population pheromone interaction mechanism to achieve global resource allocation through pheromone sharing; Transportation tool-specific constraint conditions, including drone range limits, vehicle load limits, and robot terrain adaptation rules.

6. The system according to claim 1, wherein The path modeling module further includes: A dynamic obstacle handling sub-module that fuses satellite remote sensing and traffic monitoring data in real time to reconstruct the topology of the road interruption area; A multi-modal cooperation sub-module that dynamically allocates transportation tools according to disaster scenarios and preferentially enables drones to deliver to high-urgency areas.

7. The system according to claim 1, wherein In the psychological assistance priority module, the psychological assistance priority module processes the text data of social media and distress calls through sentiment analysis methods to generate a psychological crisis index Q r Obtained by analyzing the sentiment of social media text and weighted into the weight factor of demand urgency; the weight factor of demand urgency is extended to: where α1 + α2 + α3 + α4 = 1, Q j , and α4 are the psychological crisis index and weight coefficients respectively.

8. The system according to any one of claims 1-7, characterized in that, It also includes: A visualization interaction terminal that uses 3D and AR technologies to overlay and display the scheduling plan and road conditions in real time, and provides a multi-strategy comparison view; A digital twin and blockchain module that records the entire process of scheduling decisions to ensure traceability and data integrity.

9. An emergency material scheduling method based on the system according to any one of claims 1-8, characterized in that, Including the following steps: Step S1: Obtain historical disaster situations, meteorology, public opinion, and real-time traffic data through an external interface, and use a spatio-temporal graph neural network to predict the material demand in the next 6 - 72 hours; Step S2: Construct a multi-modal hierarchical transportation network model, define road traffic constraints, transportation tool performance parameters, and multi-modal cooperation rules; Step S3: Use an improved ant colony algorithm to iteratively optimize the path. When initializing pheromones, allocate them differently according to road grades. After each iteration, update the pheromones through an elite strategy and perform cross-ant colony resource coordination; Step S4: Display the scheduling plan through an AR visualization terminal, supporting manual adjustment and multi-objective plan comparison; Step S5: During transportation, use edge computing nodes to monitor road conditions in real time. When the road is abnormal, trigger local path replanning, and the response time ≤ 30 seconds; Step S6: Record dispatching decision data through the blockchain and synchronously update it with the digital twin model.

10. The method according to claim 9, wherein In the said Step S3, the multi-objective function for path optimization is: Minimize(ω1·T + ω2·C + ω3·(1 - F) + ω4·E) where T is the transportation time; C is the cost; F is the fairness index; E is the carbon emission; ω1 to ω4 are weight coefficients and Σω1 + ω2 + ω3 + ω4 = 1; d e is the length of edge e, v k is the speed of vehicle k; μ k is the fuel consumption rate of tool k; p is the unit price of fuel; R i is the material satisfaction rate of the c-th requirement order, which is the ratio of the received quantity to the required quantity; n is the total number of demand points; ζ k is the carbon emission coefficient per unit distance of tool k; In the said Step S5, the local path replanning includes: Deploy a lightweight ant colony algorithm on the vehicle side through the federated learning framework; Only perform local search on adjacent nodes of the affected path, and limit the number of iterations ≤ 50 times; Asynchronously synchronize the replanning result with the cloud global model.

Images