A two-stage collaborative scheduling method for emergency material distribution of forest fire considering flexible power exchange of unmanned aerial vehicle

Abstract

This invention relates to a two-stage collaborative scheduling method for forest fire emergency material delivery that considers flexible battery swapping for drones. The method includes constructing an attention-weighted priority evaluation system based on fire point accessibility and urgency; building a truck-drone two-stage collaborative delivery model that combines offline global planning with online dynamic adjustment; designing an attention-weighted ant colony-tabo search hybrid algorithm to solve for the optimal path in the model; establishing an adaptive two-layer emergency allocation and rescue mechanism to dynamically allocate tasks to newly added demand points; using trucks as mobile charging stations for drones to achieve flexible battery swapping for drones and seamless vehicle-drone takeoff and landing coordination; and simultaneously updating vehicle and demand point status in real time through a rolling update mechanism to respond to abnormal situations and continuously execute the scheduling plan until delivery is completed. This invention can adapt to complex forest fire emergency material delivery scenarios with road blockages and dynamically changing demands, effectively reducing operating costs and shortening task completion and material delivery response times.

Description

Technical Field

[0001] This invention relates to the field of forest fire emergency management and material distribution collaborative scheduling technology, and in particular to a two-stage collaborative scheduling method for forest fire emergency material distribution that takes into account the flexible battery swapping of drones. Background Technology

[0002] Forest fires pose a serious threat to forest resources, ecosystems, and human safety, and reliable emergency supply delivery is a key element in efficient firefighting. Fires are often accompanied by road blockages and traffic congestion. Traditional single-truck delivery methods suffer from high route failure rates and inability to reach remote fire sites. While drone delivery offers advantages such as strong terrain adaptability and fast response times, its limited battery capacity leads to insufficient range, becoming a core issue hindering the efficient delivery of emergency supplies.

[0003] In existing truck-drone collaborative emergency supply delivery technologies, drones typically need to return to fixed truck delivery nodes for charging / battery swapping, while trucks must wait at these nodes for the drones to complete resupply, resulting in significant time waste. Furthermore, drones are prone to ineffective detours, increasing power and labor costs. Simultaneously, existing scheduling models often employ static planning, making it difficult to handle dynamic changes in demand points at fire scenes. They lack differentiated priority design based on the urgency and accessibility of fire points, easily leading to delays in the delivery of supplies to critical fire locations. In addition, traditional solution algorithms, such as single ant colony algorithms, suffer from premature convergence and parameter rigidity issues, while machine learning methods are limited by training data and exhibit poor generalization ability in cold-start fire emergency scenarios, making it difficult to simultaneously meet the dual requirements of offline global optimization and online real-time adjustment.

[0004] In summary, existing forest fire emergency material delivery technologies suffer from problems such as rigid drone battery swapping modes, low vehicle-machine collaboration efficiency, insufficient dynamic demand response capabilities, and poor adaptability of scheduling algorithms. These issues fail to meet the requirements of high efficiency, flexibility, and real-time delivery of emergency materials in complex fire scenarios. There is an urgent need for a collaborative scheduling method that takes into account flexible drone battery swapping, fire point priority scheduling, and offline-online two-stage planning. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a two-stage collaborative scheduling method for forest fire emergency material delivery that considers flexible battery swapping for drones. This method solves the technical problems in existing forest fire emergency material delivery, such as reliance on fixed truck nodes for drone battery swapping, low vehicle-drone collaboration efficiency, delayed response to dynamic new demands at the fire site, unreasonable resource allocation due to the lack of differentiated priority for fire points, and the difficulty in balancing offline global optimization with online real-time adaptability in scheduling algorithms. The invention achieves efficient truck-drone collaboration, flexible drone battery swapping, rapid response to dynamic demands, and optimized resource allocation.

[0006] To achieve the above technical objectives, the present invention provides the following technical solution: a two-stage collaborative scheduling method for the delivery of emergency supplies for forest fires, considering the flexible battery swapping of drones, comprising the following steps: Determine the accessibility type and urgency of each demand point, and construct a priority matrix for paths between each demand point; the accessibility type includes truck-only accessibility, drone-only accessibility, and accessibility by both trucks and drones; the urgency level includes low urgency and high urgency. When all requirements are known, the system enters the offline global planning stage; when new requirements are added, the system enters the online dynamic adjustment stage. A two-stage collaborative delivery model of truck and drone is constructed with the goal of minimizing the total distance cost between truck and drone. An objective function is constructed and constraints are set. We designed an attention-weighted ant colony-taboo hybrid algorithm to solve the two-stage collaborative delivery model of trucks and drones. At the same time, we established an adaptive two-layer emergency allocation and rescue mechanism to plan the main road of trucks and the task sequence of drones, and obtained the offline initial optimal path in the offline global planning stage and the online adjusted optimal path in the online dynamic adjustment stage. The adaptive two-layer emergency allocation and rescue mechanism allocates truck and drone tasks to newly added demand points based on the accessibility type of the newly added demand points during the online dynamic adjustment phase. Using trucks as mobile charging stations for drones enables flexible battery swapping and coordinated take-off and landing between drones and trucks. The system dynamically monitors the status of trucks and drones, and re-plans routes if any abnormalities occur.

[0007] Optionally, the construction of the priority matrix for paths between each demand point includes: Based on the accessibility type and urgency of each demand point, attention weights for accessibility type and urgency are set, and then weighted and normalized using the Sigmoid activation function to obtain the priority matrix of each demand point; the internal elements of the priority matrix are represented as follows: ; in, The first in the priority matrix Line number Column elements, representing the number of columns starting from the first column. The demand point goes to the first Path priority for each requirement point; This represents the Sigmoid activation function; , These represent the attention weights for reachability type and urgency, respectively. , They represent the first The availability type and urgency of each demand point.

[0008] Optionally, in the offline global planning phase, the objective function of the truck-drone two-stage collaborative delivery model is... The definition is as follows: ; in, This indicates a minimize operation; , These represent the distance cost coefficients for trucks and drones, respectively. Indicates the first The truck arrived at the The straight-line distance between the demand points; Indicates the first The drone to the The straight-line distance between the demand points; , The instructions are respectively the first truck, the Does the drone serve the first Decision variables for each demand point; Total number of trucks; The total number of drones; This represents the total number of demand points. During the online dynamic adjustment phase, the objective function of the truck-drone two-stage collaborative delivery model is... The definition is as follows: ; in, This represents the total number of new demand points; , The first truck, the The drone to the first The service allocation adjustment amount for each newly added demand point; the service allocation adjustment amount indicates whether the newly added demand point will be reassigned to a truck or drone for service. The total distance cost is defined as follows: ; in, This represents the total distance cost.

[0009] Optionally, the constraints include: demand point service matching constraints during the offline global planning phase, drone flight distance and remaining battery power constraints, truck path closure constraints, spatiotemporal coordinated take-off and landing constraints between drones and trucks, and vehicle return to the delivery center after completing the task. The demand point service matching constraint is defined as follows: each demand point can only be served once by a truck or a drone, and the type of the service vehicle must match the reachability type of the demand point. The constraints on the flight distance and remaining battery power of the drone are defined as follows: the total distance of all flight missions performed by each drone shall not exceed the maximum flight distance corresponding to its battery capacity; The truck route closure constraint is defined as follows: the driving route of each truck must form a closed loop starting from the distribution center, visiting all the demand points it is responsible for in sequence, and finally returning to the distribution center, and no sub-loops can appear in the driving route; The spatiotemporal coordinated take-off and landing constraint between the UAV and the truck is defined as follows: the take-off and landing positions of the UAV must be precisely matched with the position of the truck, that is, the UAV must be on the truck at the time of take-off and must also find the current position of the truck when returning after completing the mission. The constraints also include: service matching constraints for new demand points in the online dynamic adjustment phase, drone remaining battery power constraints, vehicle route adjustment continuity constraints, and the constraint that vehicles must return to the delivery center after completing their tasks, which must be followed in both the offline global planning phase and the online dynamic adjustment phase. The new demand point service matching constraint is defined as follows: each new demand point can only be served by one truck or one drone once, and the type of the service vehicle must match the reachability type of the new demand point. The remaining battery power constraint for the drone is defined as follows: the total distance of all flight missions performed by each drone for the newly added demand point shall not exceed the remaining flight distance calculated based on its remaining battery power. The vehicle route adjustment continuity constraint is defined as follows: when a new demand point is assigned to any truck, the original driving route of the truck must still remain a closed loop starting from the distribution center, visiting all the demand points it is responsible for in sequence, and finally returning to the distribution center after the new demand point is inserted, and there cannot be sub-loops or breaks in the driving route. The constraint that vehicles must return to the distribution center after completing their mission is defined as follows: all trucks and drones must return to the distribution center after completing all material delivery missions.

[0010] Optionally, the design employs a hybrid attention-weighted ant colony-taboo search algorithm to solve the two-stage truck-drone collaborative delivery model. Simultaneously, it establishes an adaptive two-layer emergency allocation and rescue mechanism, plans the truck backbone route and drone mission sequence, and obtains the offline initial optimal path in the offline global planning stage and the online adjusted optimal path in the online dynamic adjustment stage, including: During the offline global planning phase, truck trunk routes are generated based on the tabu search algorithm and the urgency of the demand points. An attention-weighted ant colony algorithm is designed to update pheromone concentration based on a priority matrix, thereby optimizing the drone mission sequence that takes off from each truck stop along the main truck route. During the online dynamic adjustment phase, based on the established adaptive two-layer emergency allocation and rescue mechanism, the allocation of truck and drone tasks for newly added demand points is optimized. When the attention-weighted ant colony-taboo hybrid algorithm reaches the preset maximum number of iterations, or when the total distance cost tends to stabilize, it outputs the offline initial optimal path and the online adjusted optimal path.

[0011] Optionally, the generation of truck trunk routes based on the tabu search algorithm combined with the urgency of the demand points includes: Set the dynamic escape probability of the tabu search algorithm as follows: ; in Indicates the first The urgency of each demand point; The aforementioned attention-weighted ant colony algorithm updates pheromone concentration based on a priority matrix to optimize the drone mission sequence taking off from each truck stop along the main truck route, including: Treating trucks and drones as ants, the formula for calculating the transfer probability is: ; in This represents the ant index. Indicates the first The ant in the 1st The set of accessible neighboring demand points for each demand point. For heuristic functions, Indicates the number of iterations starting from the 1st iteration. The demand point to the first The path to each demand point pheromone concentration, , These are the weighting coefficients for pheromones and heuristic functions, respectively. For the first Index of accessible neighboring demand points for each demand point; transition probability Indicates the first Only one ant is currently located at the th When selecting the first requirement, choose to visit the next one. The probability of each demand point; Indicates from the first The demand point goes to the first The path priority of each demand point represents the path priority weight from the current demand point to accessible neighboring demand points. For the first Index of accessible neighboring demand points for each demand point; The set of accessible neighborhood demand points is the set of all demand points that have not yet been visited by the current ant and that match the reachability type of the vehicle represented by the current ant. The pheromone concentration is updated as follows: ; in Indicates the next iteration from the th The demand point to the first Path of each demand point pheromone concentration; The first in the priority matrix Line number Column elements, representing the number of columns starting from the first column. The demand point goes to the first The path priority of each requirement point.

[0012] Optionally, during the online dynamic adjustment phase, based on the established adaptive two-layer emergency allocation and rescue mechanism, the allocation of truck and drone tasks for newly added demand points is optimized, including: Determine the reachability type of the new demand point. If the new demand point is reachable only by drones, determine whether the distance between the new demand point and the current truck location exceeds a preset distance threshold. If it does not exceed the threshold, execute a restricted ant colony search to quickly return a feasible path. If it does exceed the threshold, execute the full attention-weighted ant colony-taboo search hybrid algorithm. The restricted ant colony search calculates the round-trip distance from each idle drone to the new demand point from its current location. First, it filters out drones whose remaining range meets the requirements for round-trip to the new demand point, and then selects the drone closest to the demand point to serve the new demand point. If there are no idle drones whose remaining range meets the requirements for round-trip to the new demand point, it evaluates the remaining range of the drones currently performing tasks after completing their current tasks, first filters out drones that can take on the service task of the new demand point, and then selects the drone whose current task was completed earliest. If the newly added demand point is accessible only by trucks, then the newly added demand point is added to the set of unvisited demand points for trucks, and a new truck trunk route is generated based on the tabu search algorithm; If the new demand point is reachable by both trucks and drones, the system first searches for available drones with sufficient remaining range to travel to and from the new demand point. If no available drones have sufficient remaining range, the system searches for drones that are currently returning and have sufficient remaining range to travel to and from the new demand point. From the searched drones, the system selects the drone closest to the new demand point to serve that new demand point. If there are neither available drones with sufficient remaining range nor drones currently returning and have sufficient remaining range, the new demand point is added to the set of unvisited demand points for trucks. A new truck trunk route is then generated based on the tabu search algorithm to serve the new demand point.

[0013] Optionally, using the truck as a mobile charging station for the drone, enabling flexible battery swapping and coordinated takeoff and landing between the drone and the truck, includes: The take-off and landing locations of the drones are linked to the real-time locations of the trucks. The drones do not need to return to the trucks' fixed delivery nodes and can meet up with the trucks in real time at any location on the trucks' travel path. The drone can quickly resume flight after battery swapping, and the truck can complete the drone battery swapping operation while in motion, eliminating the need to wait for the drone at fixed delivery nodes; Record the remaining flight distance and battery swapping status of the drone, update the drone's endurance in real time, and provide a basis for power allocation for online task assignment.

[0014] Optionally, the dynamic detection of the truck and drone's status, and the re-planning of the route if an anomaly occurs, includes: The location, remaining tasks, battery status, accessibility type, and urgency of the demand points for trucks and drones are refreshed at fixed intervals. If an abnormal situation occurs, the attention-weighted ant colony-taboo search hybrid algorithm and the adaptive two-layer emergency allocation and rescue mechanism will be re-triggered to re-plan the path; The anomalies include at least drone malfunctions, truck route disruptions, and changes in demand point priorities.

[0015] Optionally, if an abnormal situation occurs, the attention-weighted ant colony-taboo search hybrid algorithm and the adaptive two-layer emergency allocation and rescue mechanism will be re-triggered to re-plan the path, including: When a drone malfunctions, the unfulfilled service requests of the malfunctioning drone are treated as new requests, and the unfulfilled service requests are reallocated to trucks or drones based on an adaptive two-layer emergency allocation and rescue mechanism. When a truck's path is blocked, the affected trucks are replanned locally based on the attention-weighted ant colony-taboo search hybrid algorithm. If the truck has already entered the restricted section, its unfinished demand points are reassigned to other unaffected drones or trucks based on the adaptive two-layer emergency allocation and rescue mechanism. When the priority of a demand point changes, the demand point whose priority is increased is regarded as a new demand point, and the truck or drone service is reallocated to the demand point based on the adaptive two-level emergency allocation and rescue mechanism.

[0016] By employing the above technical solution, the present invention provides a two-stage collaborative scheduling method for the delivery of emergency supplies for forest fires, taking into account the flexible battery swapping of drones, which has at least the following beneficial effects: (1) By using a truck as a mobile charging station for drones, the drones can meet and exchange batteries at any location on the truck's driving path without having to return to a fixed node. The trucks do not have to wait for the drones, which alleviates the problems of rigid drone battery exchange and mutual waiting between vehicles and drones in the traditional model, reduces ineffective travel and resource waste, and reduces the total operating cost compared to the traditional model. It realizes flexible battery exchange for drones and improves the efficiency of vehicle-drone collaboration. (2) This invention constructs a two-stage collaborative delivery model of truck-drone. In the offline stage, the global path optimization of the initial demand point is realized to ensure the economy of the scheduling scheme. In the online stage, the new temporary demand point is quickly responded to and the path adjustment is completed under the premise of minimizing the interference with the original plan, so as to realize dynamic scheduling. Compared with the traditional static planning model, the task completion time is shortened, and the global optimization and dynamic response are taken into account. (3) Based on the accessibility type and urgency of the demand points, the present invention designs differentiated priorities, and prioritizes the allocation of scarce drone resources to key fire points that are geographically isolated and spread rapidly, effectively shortening the material delivery response time of key fire points, improving the timeliness of fire fighting, and realizing the optimal allocation of resources. (4) In view of the characteristics of forest fire scenarios such as "short response time", "urgent demand" and "fast speed", this invention designs an attention-weighted ant colony-tabo search hybrid algorithm. By integrating the attention-weighted ant colony algorithm and the tabu search algorithm, the algorithm adaptability problem is solved. The positive feedback mechanism of the ant colony algorithm realizes parallel search, and the tabu search avoids premature convergence and cyclic search. At the same time, the lightweight design of the algorithm ensures that the online scheduling returns a feasible path within milliseconds, which meets the real-time requirements of emergency scenarios. Meanwhile, the positive feedback mechanism and the local search capability of the tabu search are deeply coupled with the unique constraints of forest fires through fire point priority, real-time rolling in offline and online two stages, etc., to achieve a balance between global optimization and dynamic response. (5) This invention establishes an adaptive two-layer emergency allocation and rescue mechanism, and designs differentiated allocation rules for three types of demand points: those that can only be reached by trucks, those that can only be reached by drones, and those that can be reached by both. This enables the rapid and reasonable allocation of new demand points, ensures full coverage and efficiency of emergency material distribution, and solves the problem of algorithm adaptability. (6) The present invention refreshes the status of vehicles and demand points in real time through a rolling update mechanism. If abnormal situations such as vehicle failure or path blockage occur, the scheduling algorithm and allocation mechanism can be quickly re-triggered to realize the re-optimization and allocation of remaining resources. It adapts to complex forest fire emergency scenarios such as road blockage and dynamic changes in demand, making the algorithm of the present invention robust and fault-tolerant, and adaptable to complex fire scenarios. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart of a two-stage collaborative scheduling method for the delivery of emergency supplies for forest fires, which takes into account the flexible battery swapping of drones, according to the present invention. Figure 2 A schematic diagram of the attention-weighted ant colony-tacit search hybrid algorithm designed for this invention; Figure 3 This is a schematic diagram of the two-stage cooperative path optimization process of truck-drone in an embodiment of the present invention; Figure 4 This is a flowchart of the truck-drone collaborative dynamic scheduling method for handling temporary demand points in an embodiment of the present invention; Figure 5 This is a schematic diagram of the travel paths of the truck and the drone during the offline global planning stage in an embodiment of the present invention; Figure 6 This is a schematic diagram of the travel paths of trucks and drones during the online dynamic adjustment phase in an embodiment of the present invention; Figure 7 This is a visualization of path analysis data for different vehicles in the two-stage collaborative path optimization process of truck-drone in an embodiment of the present invention. Detailed Implementation

[0018] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. This will allow for a full understanding of how the present application uses technical means to solve technical problems and achieve technical effects, and to facilitate its implementation.

[0019] Those skilled in the art will understand that all or part of the steps in the implementation of the methods of the embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0020] Please refer to Figures 1-7 This illustration demonstrates a specific implementation of this embodiment. This embodiment constructs an attention-weighted priority evaluation system based on fire point accessibility and urgency, then builds a two-stage collaborative delivery model of truck-drone, combining offline global planning and online dynamic adjustment. An attention-weighted ant colony-taboo hybrid algorithm is designed to solve for the optimal path in the model. An adaptive two-layer emergency allocation and rescue mechanism is established to achieve dynamic task allocation for newly added demand points. Trucks are used as mobile charging stations for drones, enabling flexible drone battery swapping and seamless vehicle-drone take-off and landing collaboration. Simultaneously, a rolling update mechanism refreshes the vehicle and demand point status in real time, responding to abnormal situations and continuously executing the scheduling plan until delivery is completed. This solves the problems of rigid drone battery swapping modes, low vehicle-drone collaboration efficiency, delayed dynamic demand response, and unreasonable resource allocation in existing technologies. It achieves drone battery swapping without fixed demand points, improves vehicle-drone collaboration efficiency, balances global optimization and dynamic responsiveness in scheduling, optimizes the allocation of emergency resources, and possesses good robustness and fault tolerance. It can adapt to complex forest fire emergency material delivery scenarios with road blockages and dynamically changing demands, effectively reducing operating costs and shortening task completion time and material delivery response time.

[0021] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All experiments were implemented using MATLAB R2024a programming, running on Windows 11 Professional Edition, with hardware configuration of Intel i7-13900H 3.8GHz, 64GB RAM, and RTX 4060 graphics card. The core algorithm code has been encapsulated as a scheduling middleware, which can be used plug-and-play at edge computing demand points. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Please refer to Figure 1 This embodiment proposes a two-stage collaborative scheduling method for the distribution of emergency supplies for forest fires, taking into account the flexible battery swapping of drones. The method includes: (1) Scene parameter settings Twenty emergency supply delivery points were selected, of which eight can be reached by trucks alone, five by drones alone, and seven by both trucks and drones. Equipped with 2 trucks and 2 drones, each truck carries 1 drone. The drones have a maximum range of 50km. The distance cost coefficient between the trucks and drones is... , The distribution center is located at coordinates The demand points are distributed within a two-dimensional area of ​​[-10,10]×[-10,10]km. There are no initial road obstructions at the fire scene. During the online phase, the coordinates are... Add one temporary demand point that can only be reached by drones (i.e., the newly added demand point).

[0023] (2) Method execution steps S1. Determine the accessibility type and urgency of each demand point, and construct a priority matrix of paths between each demand point; the accessibility type includes truck-only accessibility, drone-only accessibility, and accessibility by both trucks and drones; the urgency level includes low urgency and high urgency.

[0024] As a preferred embodiment of step S1, it specifically includes: S11. The demand points for emergency supplies for forest fires are divided into three categories according to their accessibility: those accessible only by trucks, those accessible only by drones, and those accessible by both trucks and drones. The urgency of the demand points is determined according to the speed of fire spread and the distribution of people.

[0025] In this embodiment, the specific method for classifying the urgency of the demand points is as follows: S111, Calculate fire spread risk factors , This is the demand point index, reflecting the immediate threat posed by the fire point (i.e., the demand point) to the surrounding area. It is based on the rate of fire spread. (Unit: meters per minute) Quantified: ; S112, Calculation of personnel safety risk factors This factor reflects the degree of threat posed to life safety by a fire. The number of people around the fire point is taken into account. The closest distance between personnel and the fire (Unit: meters): First, score based on the number of people: If Personnel score = 1.0; if Personnel score = 0.6; if Personnel score = 0.2; Secondly, based on distance scoring: if Distance score = 1.0; if Distance score = 0.6; if Distance score = 0.3; Ultimately, the personnel safety risk factor is the maximum of the two: , This indicates the operation of retrieving the maximum value; S113, Comprehensive calculation of urgency level : A weighted summation method is used, and the weights can be adjusted according to the emphasis of the scenario: .

[0026] S12. Based on the accessibility type and urgency of each demand point, set the accessibility type attention weight and urgency attention weight for each demand point, and then use the Sigmoid activation function to perform weighted normalization to obtain the priority matrix of the paths between each demand point; the internal elements of the priority matrix are represented as follows: ; in, Represents the first in the priority matrix Line number Column elements represent the demand points. (Departure point) Head to the demand point The path priority of (target requirement point) is entirely determined by the requirement point. The availability and urgency of the target demand point determine its adequacy. This represents the Sigmoid activation function, which normalizes data and amplifies the distinguishability of key fire points. , These represent the attention weights for reachability type and urgency, respectively. , Representing the demand points respectively The accessibility type and urgency level are considered. In this embodiment, the accessibility type attention weight for each demand point is set only when trucks can reach it. Set only when drones can reach it. Both trucks and drones can be set upon arrival. The urgency level of each demand point, and the low urgency level ( When setting Medium urgency ( When setting High urgency level ( When setting The priority matrix is ​​used to quantify the accessibility type and urgency of a requirement point. Head to the demand point path The combined effects of priorities.

[0027] This invention designs differentiated priorities based on the accessibility type and urgency of demand points, prioritizing the allocation of scarce drone resources to geographically isolated, rapidly spreading critical fire points. This effectively shortens the material delivery response time to critical fire points, improves the timeliness of fire suppression, and achieves optimized resource allocation.

[0028] S2. When all requirements are known, the system enters the offline global planning stage. When new requirements are added, the system enters the online dynamic adjustment stage.

[0029] A two-stage collaborative delivery model of truck and drone is constructed with the goal of minimizing the total distance cost between truck and drone. An objective function is constructed and constraints are set.

[0030] As a preferred embodiment of step S2, it specifically includes: The objective function of the truck-drone two-stage collaborative delivery model during the offline global planning phase. The definition is as follows: ; in, This indicates a minimize operation; , These represent the distance cost coefficients for trucks and drones, respectively. Indicates the first The truck arrived at the The straight-line distance between the demand points; Indicates the first The drone to the The straight-line distance between the demand points; , The instructions are respectively the first truck, the Does the drone serve the first Decision variables for each demand point (with values ​​of 0 or 1); Total number of trucks; The total number of drones; This represents the total number of demand points. The objective function of the truck-drone two-stage collaborative delivery model during the online dynamic adjustment phase. The definition is as follows: ; in, This represents the total number of new demand points; , The first truck, the The drone to the first The service allocation adjustment amount for each new demand point (a 0-1 decision variable, indicating whether to reassign the new demand point to trucks or drones for material delivery compared to the offline global planning stage; a value of 1 represents adding the material delivery task, and a value of 0 represents not adding it).

[0031] The total distance cost is defined as follows: ; in, This represents the total distance cost.

[0032] This invention constructs a two-stage collaborative delivery model of trucks and drones. In the offline stage, global path optimization of the initial demand points is achieved to ensure the economy of the scheduling scheme. In the online stage, new temporary demand points are quickly responded to, and path adjustment is completed while minimizing interference with the original plan, thus realizing dynamic scheduling. Compared with the traditional static planning model, the task completion time is shortened, and global optimization and dynamic response are taken into account.

[0033] The constraints in the offline global planning phase include: demand point service matching constraints, drone flight distance and remaining battery power constraints, truck path closure constraints, spatiotemporal coordinated take-off and landing constraints of drones and trucks, and vehicle return to the distribution center after completing the task constraints.

[0034] The demand point service matching constraint is defined as follows: each demand point can only be served once by a truck or a drone, and the type of the service vehicle must match the reachability type of the demand point. The constraints on the flight distance and remaining battery power of the drone are defined as follows: the total distance of all flight missions performed by each drone shall not exceed the maximum flight distance corresponding to its battery capacity; The truck route closure constraint is defined as follows: the driving route of each truck must form a closed loop starting from the distribution center, visiting all the demand points it is responsible for in sequence, and finally returning to the distribution center, and no sub-loops can appear in the driving route; The spatiotemporal coordinated take-off and landing constraint between the UAV and the truck is defined as follows: the take-off and landing positions of the UAV must be precisely matched with the position of the truck, that is, the UAV must be on the truck at the time of take-off and must also find the current position of the truck when returning after completing the mission. The constraints in the online dynamic adjustment phase include: service matching constraints for new demand points, remaining power constraints for drones, continuity constraints for vehicle route adjustments, and the constraint that vehicles must return to the delivery center after completing their tasks, which must be followed in both the offline global planning phase and the online dynamic adjustment phase. The new demand point service matching constraint is defined as follows: each new demand point can only be served by one truck or one drone once, and the type of the service vehicle must match the reachability type of the new demand point. The remaining battery power constraint for the drone is defined as follows: the total distance of all flight missions performed by each drone for the newly added demand point shall not exceed the remaining flight distance calculated based on its remaining battery power. The vehicle route adjustment continuity constraint is defined as follows: when a new demand point is assigned to any truck, the original driving route of the truck must still remain a closed loop starting from the distribution center, visiting all the demand points it is responsible for in sequence, and finally returning to the distribution center after the new demand point is inserted, and there cannot be sub-loops or breaks in the driving route. The constraint that vehicles must return to the distribution center after completing their mission is defined as follows: all trucks and drones must return to the distribution center after completing all material delivery missions.

[0035] S3. Design an attention-weighted ant colony-taboo hybrid algorithm to solve the two-stage collaborative delivery model of trucks and drones. At the same time, establish an adaptive two-layer emergency allocation and rescue mechanism to plan the main road of trucks and the task sequence of drones, and obtain the offline initial optimal path in the offline global planning stage and the online adjusted optimal path in the online dynamic adjustment stage.

[0036] The flowchart of the designed attention-weighted ant colony-tabu search hybrid algorithm can be found in [reference needed]. Figure 2 , Figure 2 It includes three main modules: initialization, offline global planning, and online phase dynamic adjustment. Figure 2 In the algorithm, the initialization module is the initial preparation stage, performing parameter initialization, including setting parameters, initializing pheromones, establishing the attention matrix W (i.e., the priority matrix of paths between demand points), and initializing the population. Pheromones initialization and population initialization are the steps for setting path pheromones and the initial solution population of the algorithm, respectively. In the offline global planning module, the initial solution (i.e., the offline initial optimal path) is generated, and the attention-weighted ant colony algorithm is applied to calculate the transition probability. The attention-weighted mechanism is implemented by combining a priority matrix; simultaneously, adjacency rules, tabu verification, and incentive rules of the tabu search algorithm are combined to achieve tabu search reinforcement. This forms the pheromone update criterion for the attention-weighted ant colony-tabu search hybrid algorithm in the offline global planning module: global optimization and optimal path reinforcement. In the online phase dynamic adjustment module, "the appearance of a temporary demand point P*" indicates a new demand point that triggers dynamic scheduling in the online phase; "determining the nature of the temporary demand point P*" is the step of identifying the reachability type of the new demand point; then, scheduling vehicles can be allocated for material delivery based on the nature of the temporary demand point; the online phase dynamic adjustment module performs local path replanning, where "adding the temporary demand point P* to the tabu list" and "dynamically adjusting the tabu list capacity" are the core constraints and adaptive adjustment mechanisms of the tabu search algorithm; the attention-weighted ant colony (AWACO) algorithm is executed to update probabilities and achieve dynamic pheromone adjustment. The adjustment results are combined with the results of the tabu search algorithm to achieve incremental adjustment (triggering the appearance of the temporary demand point P*) and boundary adjustment; "truck-related" and "drone-related" represent the algorithm optimization steps for truck paths and drone task sequences, respectively. Set a stopping criterion: When the maximum number of iterations is reached or the objective function converges, stop the attention-weighted ant colony-tautology hybrid algorithm and output the optimal cooperative solution. In the diagram, arrow 2 indicates the execution order and data flow between the algorithm's steps.

[0037] The two-stage collaborative path optimization process for truck-drone can be referenced. Figure 3 ,like Figure 3 The yellow dots represent points reachable by both trucks and drones ("both reachable"); blue dots represent points reachable only by trucks; and purple dots represent points reachable only by drones. The small house icon in the diagram represents the starting point (distribution center). Truck 1 and Truck 2 depart from the starting point, carrying Drone 1 and Drone 2 respectively, to deliver supplies (Truck 1 and Drone 1 form group 1, and Truck 2 and Drone 2 form group 2). Based on the nature of temporary demand points (new demand points), an optimization algorithm is used to schedule vehicle delivery. Black lines represent truck paths, black arrows indicate truck directions, green dashed arrows represent drone flight paths, and blue dashed lines represent possible truck paths after a new demand point appears.

[0038] The routes for trucks and drones during the offline global planning phase can be referenced. Figure 5 , Figure 5 The diagram illustrates the travel routes of trucks 1 and 2 and the flight paths of drones 1 and 2 during the offline global planning phase. A Cartesian coordinate system is established in the diagram, with the horizontal axis labeled "X (km)" and the vertical axis labeled "Y (km)," using kilometers (km) to represent the planar positions of the vehicle paths at each demand point. Figure 5The diagram on the right shows the following: a black five-pointed star icon corresponds to "Distribution Center," which is the starting and ending point of this collaborative delivery; an orange square icon corresponds to "Truck-only," used to identify demand points accessible only by truck-type vehicles; a blue triangle icon corresponds to "Drone-only," used to identify demand points accessible only by drone-type vehicles; a purple circle icon corresponds to "Both accessible," used to identify demand points accessible by both truck-type and drone-type vehicles; a solid blue line corresponds to "Truck 1 Route," the path of truck number 1; a solid orange line corresponds to "Truck 2 Route," the path of truck number 2; a dashed blue line corresponds to "Drone 1 Flight Route," the flight path of drone number 1; a dashed orange line corresponds to "Drone 2 Flight Route," the flight path of drone number 2; and the numbers (1 to 20) next to each demand point are their respective point numbers.

[0039] The routes of trucks and drones during the online dynamic adjustment phase can be referenced. Figure 6 . Figure 6 The diagram illustrates the routes of trucks 1 and 2 and the flight paths of drones 1 and 2 during the online dynamic adjustment phase. A Cartesian coordinate system is established in the diagram, with the horizontal axis labeled "X (km)" and the vertical axis labeled "Y (km)," using kilometers (km) to represent the planar positions of each node and vehicle path. Figure 6 The diagram on the right shows the following: a black five-pointed star icon corresponds to the "Distribution Center," representing the start and end points of this collaborative delivery; a red solid line corresponds to "Truck 1 Route," the route of truck number 1; a blue solid line corresponds to "Truck 2 Route," the route of truck number 2; a yellow dashed line corresponds to "Drone 1 Flight Path," the flight path of drone number 1; a green dashed line corresponds to "Drone 2 Flight Path," the flight path of drone number 2; square icons correspond to "Truck Demand Points," indicating the demand points visited by the trucks; and triangles... The icon points correspond to "Drone Task Points," indicating the demand points that drones are responsible for accessing; the circular icon points correspond to "Charging Locations," indicating nodes where drones can dock and charge; the pentagram icon corresponds to "New Demand Points," which are temporary service demand nodes that appear in this dynamic scheduling; among the numbers marked next to each demand point, T1-1 to T1-7 are the demand point numbers for truck 1, T2-1 to T2-8 are the demand point numbers for truck 2, D1-1 and D1-2 are the demand point numbers for drone 1, and D2-1 to D2-3 are the demand point numbers for drone 2.

[0040] Visual data on the path analysis of different vehicles in the two-stage truck-drone collaborative path optimization process can be found in [reference]. Figure 7 . Figure 7 The diagram shows data analysis of different vehicle paths, in which... Figure 7(a) is a bar chart of total vehicle travel distance, used to count the total travel distance of each independent vehicle to complete the corresponding task. Its horizontal axis is truck 1, truck 2, drone 1, drone 2, and the vertical axis "distance (km)" represents the total travel distance of the vehicle in kilometers (km). Figure 7 (b) is a bar chart of the number of driving segments for each vehicle, used to count the number of independent driving segments when each independent vehicle performs a task; its horizontal axis represents truck 1, truck 2, drone 1, drone 2, and the vertical axis "count" represents the number of driving segments corresponding to each vehicle; Figure 7 (c) is a pie chart comparing the travel distances of different vehicles, used to compare the proportion of the total travel distance of the two types of vehicles; its sectors are "trucks" and "drones", which refer to the overall truck type vehicles and the overall drone type vehicles, respectively, and are used to show the proportion of the total travel distance of the two types of vehicles in the total travel distance of all vehicles. Figure 7 (d) is a bar chart of the number of service points, used to count the total number of service points allocated to the two types of vehicles; its horizontal axis represents trucks and drones, referring to the overall number of truck-type vehicles and the overall number of drone-type vehicles, and the vertical axis "count" represents the number of service points.

[0041] As a preferred embodiment of step S3, it specifically includes: S31. First, the K-Means algorithm is used to cluster the initial demand points into two groups, which are then assigned to two trucks. S32. During the offline global planning phase, the truck trunk route is generated based on the tabu search algorithm and the urgency of the demand points. Specifically, the dynamic escape probability of the tabu search algorithm is set as follows: ; in Indicates the first The urgency of each demand point; S33. Design an attention-weighted ant colony algorithm to update pheromone concentration based on a priority matrix and optimize the drone mission sequence that takes off from each truck stop on the main truck route. Specifically, the truck trunk path is constructed through neighbor operations of insertion, swapping, and 2-opt inversion; Treating trucks and drones as ants, the formula for calculating the transfer probability is: ; in This represents the ant index. Indicates the first The ant in the 1st The set of accessible neighboring demand points for each demand point. For heuristic functions, Indicates the number of iterations starting from the 1st iteration. The demand point to the first The path to each demand point pheromone concentration, , These are the weighting coefficients for pheromones and heuristic functions, respectively. For the first Index of accessible neighboring demand points for each demand point; Indicates the first Only one ant is currently located at the th When selecting the first requirement, choose to visit the next one. The probability of each demand point; Indicates from the first The demand point goes to the first The path priority of each demand point represents the path priority weight from the current demand point to accessible neighboring demand points. For the first Index of accessible neighboring demand points for each demand point; The pheromone concentration is updated as follows: ; in Indicates the next iteration from the th The demand point to the first Path of each demand point pheromone concentration; The first in the priority matrix Line number Column elements, representing the number of columns starting from the first column. The demand point goes to the first Path priority for each requirement point; The set of accessible neighborhood demand points refers to: for the first... For each ant, at the current demand point, it is the set of all demand points that have not been visited by the ant in this path and that match the reachability type of the vehicle represented by the current ant; when the vehicle is a truck, the matching demand points are those whose reachability type is "truck-only" or "truck and drone-only"; when the means of transport is a drone, the matching demand points are those whose reachability type is "drone-only" or "truck and drone-only".

[0042] Through the above process, the ants construct a "takeoff-supply delivery-return" task chain, with pheromones simultaneously encoding travel costs and the intensity of the remaining fire. This allows for the generation of an initial optimal offline path, which in this embodiment is: Truck 1 serves 7 demand points, Truck 2 serves 8 demand points, Drone 1 serves 4 demand points, and Drone 2 serves 1 demand point.

[0043] This invention, based on a priority matrix, directly embeds the reachability and urgency of fire points into the algorithm's transition probability and pheromone enhancement rules. This allows the algorithm to have prior knowledge of the fire situation from the beginning, avoiding indiscriminate search and enabling the proactive allocation of scarce drone resources to isolated, high-urgency fire points.

[0044] S34. During the online dynamic adjustment phase, new demand points are added. Online scheduling is triggered. It is determined that only drones can reach the point. The distance between drones 1 and 2 and the point is calculated. Drone 1 is 3.2km away from the point (not exceeding the preset distance threshold, which is set to 5km in this embodiment). The remaining range is 42km, which meets the round-trip requirement. Drone 1 is assigned to undertake the task. The flight path of drone 1 is optimized by restricted ant colony search without adjusting the truck trunk route, thereby generating an updated drone task sequence. The restricted ant colony search calculates the round-trip distance from each idle drone to the new demand point from its current location. First, it filters out drones whose remaining range meets the requirements for round-trip to the new demand point, and then selects the drone closest to the demand point to serve the new demand point. If there are no idle drones whose remaining range meets the requirements for round-trip to the new demand point, it evaluates the remaining range of the drones currently performing tasks after completing their current tasks, first filters out drones that can take on the service task of the new demand point, and then selects the drone whose current task was completed earliest. When the attention-weighted ant colony-taboo hybrid algorithm reaches the preset maximum number of iterations, or when the path cost and completion time tend to stabilize, it outputs the offline initial optimal path and the online adjusted optimal path.

[0045] This invention employs tabu search in the offline phase, utilizing greedy initialization and 2-opt neighborhood operations to rapidly generate the truck backbone path. An attention-weighted ant colony-tabu search hybrid algorithm further refines the UAV task chain. In the online phase, for newly added requirement points, a restricted ant colony search is invoked, using the same pheromone matrix and priority weights to return feasible paths within milliseconds. Both approaches achieve seamless integration of offline global knowledge and rapid online response by sharing a priority matrix and dynamic pheromones.

[0046] On the other hand, facing the urgent and rapid demands of emergency supplies delivery in forest fires, traditional methods for path planning at temporary demand points during the online phase may simply limit the number of iterations to achieve a quick solution. However, the method of this invention combines scene characteristics and performs restricted ant colony search within a small space to optimize the process, thereby improving speed and reducing computation time without sacrificing the accuracy of priority determination. Simultaneously, the dynamic escape probability set in the tabu search module allows the algorithm proposed in this invention to automatically increase the neighborhood perturbation intensity near high-urgency fire points, avoiding premature convergence. Furthermore, this parameter is directly driven by the urgency of the fire and requires no additional parameter tuning.

[0047] This invention establishes an adaptive two-layer emergency allocation and rescue mechanism, designing differentiated allocation rules for three types of demand points: those accessible only by trucks, those accessible only by drones, and those accessible by both. This enables rapid and reasonable allocation of new demand points, ensuring full coverage and efficiency in emergency material delivery, and solving the algorithm adaptability problem.

[0048] S4. Use trucks as mobile charging stations for drones to enable flexible battery swapping and coordinated take-off and landing between drones and trucks.

[0049] As a preferred embodiment of step S4, it specifically includes: The take-off and landing locations of the drone are linked to the real-time location of the truck. The drone does not need to return to the truck's fixed delivery node and can rendezvous with the truck in real time at any location along the truck's travel path. Specifically, the take-off and landing locations of the drone are defined to be linked to the real-time location of the truck, satisfying the following constraints: ; ; in, , They represent the first The takeoff and landing locations of the drone; Indicates time No. The location of the truck; , The first The takeoff and return times of the drone.

[0050] After the drone and truck meet and complete the battery swap, the drone's battery is restored to 100%, and it can quickly resume flight. The truck completes the drone's battery swap while in motion, eliminating the need to wait for the drone at fixed delivery nodes.

[0051] In this embodiment, after completing its initial task, the drone 1 travels along the path of the truck 1. The truck meets at the designated location for battery swapping. After the swap, the battery is fully charged and the truck flies to the new demand point to complete the delivery of supplies. Truck 1 can continue driving normally without waiting. After the battery swap, it continues to the next demand point.

[0052] Record the remaining flight distance and battery swapping status of the drone, update the drone's endurance in real time, and provide a basis for power allocation for online task assignment.

[0053] This invention uses trucks as mobile charging stations for drones, allowing drones to meet and swap batteries at any location along the truck's route without returning to a fixed node. Trucks do not need to wait for drones, alleviating the problems of rigid drone battery swapping and mutual waiting between vehicles and drones in the traditional model. This reduces ineffective travel and resource waste, resulting in a lower overall operating cost compared to the traditional model. It enables flexible drone battery swapping and improves vehicle-drone collaboration efficiency.

[0054] In this embodiment, the transportation dispatch process after the emergence of temporary demand points (new demand points) can be referred to as follows: Figure 4 .

[0055] Figure 4 This is a flowchart of the truck-drone collaborative dynamic scheduling method for dealing with temporary demand points in an embodiment of the present invention. Figure 4 Temporary demand points appear in step 1 , indicating the triggering condition for the scheduling process, where the parameter " "Refers to newly added demand points; Step 2, "Record the remaining routes of trucks and the remaining task sets of drones in the current state," is the step of recording the current execution status of each vehicle (truck 1, truck 2, drone 1, drone 2); Step 3, determine the current temporary demand points." The nature of this process involves classifying and judging the accessibility of temporary demand points, specifically determining the temporary demand points. Which of the following three scenarios applies: "Truck-only reachable", "Drone-only reachable", or "Both truck and drone reachable"? Step 3 is divided into three scheduling scenarios: Scenario 1, where only trucks can reach the destination, adds the temporary demand point to the set of unvisited demand points for trucks and generates a new truck trunk route based on the tabu search algorithm; Scenario 2, where only drones can reach the destination, prioritizes using the idle drone with remaining range sufficient for round-trip travel and closest to the demand point; if no suitable idle drone is available, selects the drone that has completed the current task, has sufficient range, and is expected to be available earliest. "Remaining range" refers to the maximum flight distance that the drone can currently support, and "round-trip travel demand" refers to the distance from the drone's current location to the temporary demand point. The returned flight mileage requirement is as follows: "Idle drone" refers to a standby drone that is not performing a task, and "drone on the way" refers to a drone that is performing a planned task. Scenario 3 is a scenario where both trucks and drones can reach the destination. First, search for idle drones with sufficient remaining range to travel to and from the temporary demand point and that are not currently in operation. If none are found, search for drones returning home and select the closest one to perform the task. Otherwise, compare the travel costs of the two trucks, merge the unvisited points by the trucks with the temporary point, and use the AWACO algorithm for optimization, selecting the truck with the shortest total path to subsequently visit the temporary point. Here, "idle drone" refers to an idle drone, "drone on the way" refers to a schedulable drone performing a return mission, "travel cost" refers to the comprehensive cost of the truck's travel path, "AWACO algorithm" is the attention-weighted ant colony algorithm for truck path optimization designed in this invention, and "temporary point" refers to a temporary demand point. Step 4 updates the execution status based on the truck route or drone mission settings. This step updates the execution status of each vehicle after the scheduling plan is determined. The route examples "Truck 1:0→4→2→8", "Truck 2:0→5→1", "Drone 1:0→15→13", and "Drone 2:0→Temporary Demand Point→12" are the vehicle routes after the scheduling is updated. The numbers on the route are the route node numbers, "0" represents the distribution center node, and the remaining numbers are the demand point numbers to be visited. The arrow symbol is used to indicate the execution order of the route nodes.

[0056] S5 dynamically detects the status of trucks and drones, and re-plans the route if any abnormality occurs.

[0057] As a preferred embodiment of step S5, it specifically includes: The location, remaining tasks, battery status, and accessibility and urgency of the required points are refreshed at fixed intervals (which can be flexibly set according to the complexity of the forest fire emergency scenario, with a setting range of 5s-30s). If any abnormal situation occurs, the execution of the original scheduling plan will be stopped immediately, and the attention-weighted ant colony-taboo search hybrid algorithm and adaptive two-layer emergency allocation and rescue mechanism will be re-triggered to re-plan the path. The anomalies include at least drone malfunctions, truck route disruptions, and changes in demand point priorities.

[0058] More specifically, when a drone malfunctions, the unfulfilled service requests of the currently malfunctioning drone are treated as new requests, and the unfulfilled service requests are reallocated to trucks or drones based on an adaptive two-layer emergency allocation and rescue mechanism. When a truck's path is blocked, the system uses a hybrid algorithm of attention-weighted ant colony and tabu search to replan the local routes of the affected trucks, avoiding the restricted area. If the truck has already entered the restricted section, the system uses an adaptive two-layer emergency allocation and rescue mechanism to reassign its unserved demand points to other unaffected vehicles (drones or trucks). When the priority of a demand point changes, the demand point whose priority is increased is regarded as a new demand point, and the vehicle (drone or truck) is reassigned to serve the demand point based on the adaptive two-level emergency allocation and rescue mechanism.

[0059] This invention refreshes the status of vehicles and demand points in real time through a rolling update mechanism. If abnormal situations such as vehicle failure or route blockage occur, the scheduling algorithm and allocation mechanism can be quickly re-triggered to achieve the re-optimization and allocation of remaining resources. This adapts to complex forest fire emergency scenarios with road blockage and dynamic changes in demand, making the algorithm of this invention robust and fault-tolerant, and adaptable to complex fire scenarios.

[0060] In this embodiment, the vehicle's position and status are refreshed every 10 seconds. After completing the new task, the drone 1 returns to the truck 1. Once the truck and drone are in position, they will be ready to be deployed again after another battery swap. After all deliveries to the required points are completed, both the truck and the drone will return to the delivery center, and the operation will be completed.

[0061] (3) Implementation results In this embodiment, the total operating cost using the method of the present invention is 125.25 minutes, which is 4.5% lower than the traditional truck-drone joint delivery model (total cost of 131.21 minutes); the offline phase task planning time is 2.3 seconds, and the online phase path adjustment time is 0.4 seconds, meeting the real-time requirements of emergency scenarios; the material delivery response time for newly added demand points is 8 minutes, which is 35% shorter than the traditional model, achieving efficient dynamic response and optimized resource allocation.

[0062] Table 1. Comparison of specific transportation costs between the traditional truck-drone joint material delivery model and the method of this invention.

[0063] Table 1 compares the specific transportation costs of the traditional truck-drone joint material delivery model and the two-stage collaborative delivery model for emergency supplies that considers flexible battery swapping of drones proposed in this invention. It covers the number of services, travel distance, and per-vehicle cost for trucks (Truck 1, Truck 2) and drones (Drone 1, Drone 2) in both models. The total cost of the traditional model is 131.21, while the total cost of the proposed truck-drone two-stage collaborative delivery model is 125.25. The data shows that the proposed model achieves lower costs. By optimizing the number of drone services, shortening the travel distance, and combining a flexible battery swapping mechanism, drones are freed from strict dependence on truck charging at demand points, enabling independent operation of trucks and drones without waiting for each other, reducing unnecessary travel and resource waste, and verifying the effectiveness of the proposed model in improving the efficiency of emergency supply delivery and reducing costs.

[0064] This invention enables unmanned aerial vehicles (UAVs) to swap batteries without fixed demand points, improves vehicle-machine collaboration efficiency, balances global scheduling optimization and dynamic responsiveness, achieves optimized allocation of emergency resources, and has good robustness and fault tolerance. It can adapt to complex forest fire emergency material delivery scenarios such as road blockages and dynamic changes in demand, effectively reducing operating costs and shortening task completion time and material delivery response time.

[0065] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0066] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus or device (such as a computer-based system, a processor-included system or other system that can fetch and execute instructions from, an instruction execution system, apparatus or device).

[0067] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A two-stage collaborative scheduling method for forest fire emergency material distribution considering flexible UAV power switching, characterized in that, include: Determine the reachability and urgency of each demand point, and construct a priority matrix for the paths between each demand point; The reachability types include truck-only reachable, drone-only reachable, and reachable by both trucks and drones; The level of urgency includes low urgency and high urgency; When all requirements are known, the system enters the offline global planning stage; when new requirements are added, the system enters the online dynamic adjustment stage. A two-stage collaborative delivery model of truck and drone is constructed with the goal of minimizing the total distance cost between truck and drone. An objective function is constructed and constraints are set. We designed an attention-weighted ant colony-taboo hybrid algorithm to solve the two-stage collaborative delivery model of trucks and drones. At the same time, we established an adaptive two-layer emergency allocation and rescue mechanism to plan the main road of trucks and the task sequence of drones, and obtained the offline initial optimal path in the offline global planning stage and the online adjusted optimal path in the online dynamic adjustment stage. The adaptive two-layer emergency allocation and rescue mechanism allocates truck and drone tasks to newly added demand points based on the accessibility type of the newly added demand points during the online dynamic adjustment phase. Using trucks as mobile charging stations for drones enables flexible battery swapping and coordinated take-off and landing between drones and trucks. The system dynamically monitors the status of trucks and drones, and re-plans routes if any abnormalities occur.

2. The two-stage collaborative scheduling method for forest fire emergency material delivery considering flexible battery swapping of UAVs, as described in claim 1, is characterized in that: The priority matrix for constructing paths between various demand points includes: Based on the accessibility type and urgency of each demand point, attention weights for accessibility type and urgency are set, and then weighted and normalized using the Sigmoid activation function to obtain the priority matrix of each demand point; the internal elements of the priority matrix are represented as follows: ； in, The first in the priority matrix Line number Column elements, representing the number of columns starting from the first column. The demand point goes to the first Path priority for each requirement point; This represents the Sigmoid activation function; , These represent the attention weights for reachability type and urgency, respectively. , They represent the first The availability type and urgency of each demand point.

3. The two-stage collaborative scheduling method for forest fire emergency material delivery considering flexible battery swapping of UAVs, as described in claim 1, is characterized in that: The objective function of the truck-drone two-stage collaborative delivery model during the offline global planning phase. The definition is as follows: ； in, This indicates a minimize operation; , These represent the distance cost coefficients for trucks and drones, respectively. Indicates the first The truck arrived at the The straight-line distance between the demand points; Indicates the first The drone to the The straight-line distance between the demand points; , The instructions are respectively the first truck, the Does the drone serve the first Decision variables for each demand point; Total number of trucks; The total number of drones; This represents the total number of demand points; During the online dynamic adjustment phase, the objective function of the truck-drone two-stage collaborative delivery model is... The definition is as follows: ； in, This represents the total number of new demand points; , The first truck, the The drone was used to attack the first The service allocation adjustment amount for each newly added demand point; the service allocation adjustment amount indicates whether the newly added demand point will be reassigned to a truck or drone for service. The total distance cost is defined as follows: ； in, This represents the total distance cost.

4. The two-stage collaborative scheduling method for forest fire emergency material delivery considering flexible battery swapping of UAVs, as described in claim 1, is characterized in that: The constraints include: demand point service matching constraints in the offline global planning phase, drone flight distance and remaining battery power constraints, truck route closure constraints, spatiotemporal coordinated take-off and landing constraints of drones and trucks, and vehicle return to the delivery center after completing the task. The demand point service matching constraint is defined as follows: each demand point can only be served once by a truck or a drone, and the type of the service vehicle must match the reachability type of the demand point. The constraints on the flight distance and remaining battery power of the drone are defined as follows: the total distance of all flight missions performed by each drone shall not exceed the maximum flight distance corresponding to its battery capacity; The truck route closure constraint is defined as follows: the driving route of each truck must form a closed loop starting from the distribution center, visiting all the demand points it is responsible for in sequence, and finally returning to the distribution center, and no sub-loops can appear in the driving route; The spatiotemporal coordinated take-off and landing constraint between the UAV and the truck is defined as follows: the take-off and landing positions of the UAV must be precisely matched with the position of the truck, that is, the UAV must be on the truck at the time of take-off and must also find the current position of the truck when returning after completing the mission. The constraints also include: service matching constraints for new demand points in the online dynamic adjustment phase, drone remaining battery power constraints, vehicle route adjustment continuity constraints, and the constraint that vehicles must return to the delivery center after completing their tasks, which must be followed in both the offline global planning phase and the online dynamic adjustment phase. The new demand point service matching constraint is defined as follows: each new demand point can only be served by one truck or one drone once, and the type of the service vehicle must match the reachability type of the new demand point. The remaining battery power constraint for the drone is defined as follows: the total distance of all flight missions performed by each drone for the newly added demand point shall not exceed the remaining flight distance calculated based on its remaining battery power. The vehicle route adjustment continuity constraint is defined as follows: when a new demand point is assigned to any truck, the original driving route of the truck must still remain a closed loop starting from the distribution center, visiting all the demand points it is responsible for in sequence, and finally returning to the distribution center after the new demand point is inserted, and there cannot be sub-loops or breaks in the driving route. The constraint that vehicles must return to the distribution center after completing their mission is defined as follows: all trucks and drones must return to the distribution center after completing all material delivery missions.

5. The two-stage collaborative scheduling method for forest fire emergency material delivery considering flexible battery swapping of UAVs, as described in claim 1, is characterized in that: The proposed attention-weighted ant colony-taboo hybrid algorithm solves the two-stage truck-drone collaborative delivery model. Simultaneously, it establishes an adaptive two-layer emergency allocation and rescue mechanism, plans the truck backbone route and drone mission sequence, and obtains the offline initial optimal path in the offline global planning stage and the online adjusted optimal path in the online dynamic adjustment stage, including: During the offline global planning phase, truck trunk routes are generated based on the tabu search algorithm and the urgency of the demand points. An attention-weighted ant colony algorithm is designed to update pheromone concentration based on a priority matrix, thereby optimizing the drone mission sequence that takes off from each truck stop along the main truck route. During the online dynamic adjustment phase, based on the established adaptive two-layer emergency allocation and rescue mechanism, the allocation of truck and drone tasks for newly added demand points is optimized. When the attention-weighted ant colony-taboo hybrid algorithm reaches the preset maximum number of iterations, or when the total distance cost tends to stabilize, it outputs the offline initial optimal path and the online adjusted optimal path.

6. The two-stage collaborative scheduling method for forest fire emergency material delivery considering flexible battery swapping of UAVs, as described in claim 5, is characterized in that: The generation of truck trunk routes based on the tabu search algorithm and the urgency of demand points includes: Set the dynamic escape probability of the tabu search algorithm as follows: ； in Indicates the first The urgency of each demand point; The aforementioned attention-weighted ant colony algorithm updates pheromone concentration based on a priority matrix to optimize the drone mission sequence taking off from each truck stop along the main truck route, including: Treating trucks and drones as ants, the formula for calculating the transfer probability is: ； in This represents the ant index. Indicates the first The ant in the 1st The set of accessible neighboring demand points for each demand point. For heuristic functions, Indicates the number of iterations starting from the 1st iteration. The demand point to the first The path to each demand point pheromone concentration, , These are the weighting coefficients for pheromones and heuristic functions, respectively. For the first Index of accessible neighboring demand points for each demand point; transition probability Indicates the first Only one ant is currently located at the th When selecting the first requirement, choose to visit the next one. The probability of each demand point; Indicates from the first The demand point goes to the first The path priority of each demand point represents the path priority weight from the current demand point to accessible neighboring demand points. For the first Index of accessible neighboring demand points for each demand point; The set of accessible neighborhood demand points is the set of all demand points that have not yet been visited by the current ant and that match the reachability type of the vehicle represented by the current ant. The pheromone concentration is updated as follows: ； in Indicates the next iteration from the th The demand point to the first Path of each demand point pheromone concentration; The first in the priority matrix Line number Column elements, representing the number of columns starting from the first column. The demand point goes to the first The path priority of each requirement point.

7. A two-stage collaborative scheduling method for forest fire emergency material delivery considering flexible battery swapping of UAVs, as described in claim 5, is characterized in that: During the online dynamic adjustment phase, based on the established adaptive two-layer emergency allocation and rescue mechanism, the allocation of truck and drone tasks for newly added demand points is optimized, including: Determine the reachability type of the new demand point. If the new demand point is reachable only by drones, determine whether the distance between the new demand point and the current truck location exceeds a preset distance threshold. If it does not exceed the threshold, execute a restricted ant colony search to quickly return a feasible path. If it does exceed the threshold, execute the full attention-weighted ant colony-taboo search hybrid algorithm. The restricted ant colony search calculates the round-trip distance from each idle drone to the new demand point from its current location. First, it filters out drones whose remaining range meets the requirements for round-trip to the new demand point, and then selects the drone closest to the demand point to serve the new demand point. If there are no idle drones whose remaining range meets the requirements for round-trip to the new demand point, it evaluates the remaining range of the drones currently performing tasks after completing their current tasks, first filters out drones that can take on the service task of the new demand point, and then selects the drone whose current task was completed earliest. If the newly added demand point is accessible only by trucks, then the newly added demand point is added to the set of unvisited demand points for trucks, and a new truck trunk route is generated based on the tabu search algorithm; If the new demand point is reachable by both trucks and drones, the system first searches for available drones with sufficient remaining range to travel to and from the new demand point. If no available drones have sufficient remaining range, the system searches for drones that are currently returning and have sufficient remaining range to travel to and from the new demand point. From the searched drones, the system selects the drone closest to the new demand point to serve that new demand point. If there are neither available drones with sufficient remaining range nor drones currently returning and have sufficient remaining range, the new demand point is added to the set of unvisited demand points for trucks. A new truck trunk route is then generated based on the tabu search algorithm to serve the new demand point.

8. The two-stage collaborative scheduling method for forest fire emergency material delivery considering flexible battery swapping of UAVs, as described in claim 1, is characterized in that: The method of using trucks as mobile charging stations for drones, enabling flexible battery swapping and coordinated takeoff and landing between drones and trucks, includes: The take-off and landing locations of the drones are linked to the real-time locations of the trucks. The drones do not need to return to the trucks' fixed delivery nodes and can meet up with the trucks in real time at any location on the trucks' travel path. The drone can quickly resume flight after battery swapping, and the truck can complete the drone battery swapping operation while in motion, eliminating the need to wait for the drone at fixed delivery nodes; Record the remaining flight distance and battery swapping status of the drone, update the drone's endurance in real time, and provide a basis for power consumption in online task allocation.

9. A two-stage collaborative scheduling method for forest fire emergency material delivery considering flexible battery swapping of UAVs, as described in claim 1, is characterized in that: The dynamic detection of the truck and drone status, and the re-planning of the route if an abnormality occurs, includes: The location, remaining tasks, battery status, accessibility type, and urgency of the demand points for trucks and drones are refreshed at fixed intervals. If an abnormal situation occurs, the attention-weighted ant colony-taboo search hybrid algorithm and the adaptive two-layer emergency allocation and rescue mechanism will be re-triggered to re-plan the path; The anomalies include at least drone malfunctions, truck route disruptions, and changes in demand point priorities.

10. A two-stage collaborative scheduling method for forest fire emergency material delivery considering flexible battery swapping of UAVs, as described in claim 9, is characterized in that: If an abnormal situation occurs, the attention-weighted ant colony-taboo search hybrid algorithm and the adaptive two-layer emergency allocation and rescue mechanism will be re-triggered to re-plan the path, including: When a drone malfunctions, the unfulfilled service requests of the malfunctioning drone are treated as new requests, and the unfulfilled service requests are reallocated to trucks or drones based on an adaptive two-layer emergency allocation and rescue mechanism. When a truck's path is blocked, the affected trucks are replanned locally based on the attention-weighted ant colony-taboo search hybrid algorithm. If the truck has already entered the restricted section, its unfinished demand points are reassigned to other unaffected drones or trucks based on the adaptive two-layer emergency allocation and rescue mechanism. When the priority of a demand point changes, the demand point whose priority is increased is regarded as a new demand point, and the truck or drone service is reallocated to the demand point based on the adaptive two-level emergency allocation and rescue mechanism.

Images