A transport scheduling method and system based on multi-vehicle cooperation, a terminal and a storage medium

Abstract

The present application relates to the technical field of wisdom logistics, and discloses a transport scheduling method and system based on multi-vehicle cooperation, a terminal and a storage medium, comprising: obtaining first-stage task data of a logistics distribution task, and constructing a cooperative transport network model based on the first-stage task data; using a self-adaptive large neighborhood search algorithm to solve the cooperative transport network model to obtain a first transport scheduling scheme corresponding to the first-stage task data; generating second-stage task data of the logistics distribution task based on the first transport scheduling scheme; constructing a target function corresponding to the second-stage task data, and solving the target function to obtain a second-stage transport scheduling scheme corresponding to the second-stage task data; and merging the first-stage transport scheduling scheme and the second-stage transport scheduling scheme to output a complete transport scheduling scheme corresponding to the logistics distribution task. The present application can solve the problem of time-space synchronization of multi-stage transfer of the transport scheduling scheme, and improve the flexibility of end distribution.

Description

Technical Field

[0001] This invention relates to the field of smart logistics technology, and in particular to a transportation scheduling method, system, terminal and storage medium based on multi-vehicle collaboration. Background Technology

[0002] With the explosive growth of global e-commerce and the continuous increase in urban population density, urban last-mile delivery systems have become a core link connecting large-scale freight networks with end customers.

[0003] Traditional urban logistics typically employs a two-stage "port-warehouse-customer" model, where container trucks transport goods from the port to inland transit warehouses, and then trucks depart from the transit warehouses to complete last-mile delivery. However, this traditional single-ground transportation model suffers from significant efficiency bottlenecks and cost pressures.

[0004] Using drones to assist ground vehicles in air-ground collaborative delivery is a popular research direction. However, existing technologies focus on the single last-mile delivery stage, neglecting the spatiotemporal coupling between container trunk transportation from ports to transit warehouses and last-mile delivery, and lacking optimization of resource allocation for the entire system. In urban scenarios, building height also affects delivery patterns, but most existing delivery models treat customers as planar point coordinates, ignoring service constraints in the vertical dimension. Furthermore, there is a strong structural dependency between drone and truck collaborative transportation. Drones can flexibly dock between different trucks. This strong coupling characteristic makes it easy for heuristic algorithms to get trapped in local optima when dealing with large-scale instances. Operator selection lacks adaptability, making it difficult to balance search efficiency and solution quality. Therefore, how to construct a system transportation scheduling scheme is a technical challenge that urgently needs to be solved in the field of smart logistics.

[0005] Existing technologies still have limitations, such as the inability of transportation scheduling schemes to achieve time and space synchronization in multi-stage transfers and their inability to adapt to the characteristics of drone transportation. Therefore, existing technologies need further improvement. Summary of the Invention

[0006] The technical problem to be solved by the present invention is that, in view of the defects of the prior art, the present invention provides a transportation scheduling method based on multi-vehicle collaboration, so as to solve the problems that the existing transportation scheduling schemes cannot solve the problem of multi-stage transfer time and space synchronization and cannot adapt to the characteristics of UAV transportation.

[0007] The technical solution adopted by this invention to solve the technical problem is as follows: In a first aspect, the present invention provides a transportation scheduling method based on multi-vehicle cooperation, comprising: Obtain the first-stage task data of the logistics and delivery task, and construct a collaborative transportation network model based on the first-stage task data. The adaptive large neighborhood search algorithm is used to solve the cooperative transportation network model to obtain the first transportation scheduling scheme corresponding to the first stage task data. Based on the first transportation scheduling scheme, the second phase of task data for the logistics and delivery task is generated; Construct an objective function corresponding to the second-stage task data, and solve the objective function to obtain the second-stage transportation scheduling scheme corresponding to the second-stage task data; The first-stage transportation scheduling scheme and the second-stage transportation scheduling scheme are combined to output the complete transportation scheduling scheme corresponding to the logistics and distribution task.

[0008] In one implementation, the first-stage task data includes any one or more combinations of: order request data, first ground vehicle resource data, second ground vehicle resource data, air vehicle resource data, and warehouse data.

[0009] In one implementation, constructing the cooperative transportation network model based on the task data from the first stage includes: Based on the order request data and the warehouse data, determine the objective function of the collaborative transportation network model; Based on the first ground vehicle resource data, the second ground vehicle resource data, the air vehicle resource data, and the warehouse data, the constraints of the collaborative transportation network model are determined. The cooperative transportation network model is constructed based on the objective function and the constraints.

[0010] In one implementation, the step of using an adaptive large neighborhood search algorithm to solve the cooperative transportation network model to obtain a first transportation scheduling scheme corresponding to the first stage task data includes: An initial solution for the cooperative transportation network model is generated based on a greedy algorithm; wherein, the initial solution is used to plan transportation routes for the first ground vehicle, the second ground vehicle, and the air vehicle; An adaptive large neighborhood search algorithm is used to iteratively optimize the initial solution to obtain the first transportation scheduling scheme corresponding to the first stage task data.

[0011] In one implementation, generating the initial solution of the cooperative transportation network model based on a greedy algorithm includes: Construct a first transportation route for the first ground vehicle; wherein the starting point of the first transportation route is the starting point of the logistics distribution task, and the ending point of the first transportation route is a transit warehouse; A second transportation route for the second ground vehicle is constructed based on a greedy algorithm; wherein the starting point of the second transportation route is the transit warehouse, and the ending point of the second transportation route is the end point of the logistics delivery task; A third transportation route for the air vehicle is constructed based on the second transportation route; wherein the starting point of the third transportation route is the location of the second ground vehicle on the second transportation route, and the ending point of the third transportation route is the end point of the logistics delivery task.

[0012] In one implementation, the step of using an adaptive large neighborhood search algorithm to iteratively optimize the initial solution to obtain a first transportation scheduling scheme corresponding to the first stage task data includes: Based on the adaptive large neighborhood search algorithm, the state space, action space and reward function are defined, and a Markov decision process model is constructed. A destruction operator and a repair operator are generated, and an iterative solution process is performed a preset number of times based on the destruction operator, the repair operator, and the simulated annealing algorithm; Output the target solution of the cooperative transportation network model, and use the transportation scheduling scheme corresponding to the target solution as the first transportation scheduling scheme corresponding to the task data of the first stage.

[0013] In one implementation, the iterative solution process based on the destruction operator, the repair operator, and the simulated annealing algorithm, performing a preset number of iterations, includes: The initial solution is used as the target solution of the cooperative transportation network model. In each iteration, a pair of destruction operators and repair operators are selected. The solution corresponding to the destruction operator is destroyed; wherein, the destruction operator includes random client deletion operator, greedy deletion operator, correlation deletion operator and cascade destruction operator; Repair the solution corresponding to the repair operator and output a new solution; wherein, the repair operator includes: global greedy repair operator, local greedy repair operator, regret value repair operator and multi-node cascade repair operator; The simulated annealing algorithm is used to determine whether to accept the new solution. If the new solution is accepted, it is taken as the target solution; otherwise, the next iteration process is executed. When the number of iterations meets the condition, the target solution of the cooperative transportation network model is output.

[0014] Secondly, the present invention provides a transportation scheduling system based on multi-vehicle cooperation, comprising: The data acquisition module is used to acquire the first-stage task data of the logistics and distribution task, and to build a collaborative transportation network model based on the first-stage task data. The first model solving module is used to solve the cooperative transportation network model using an adaptive large neighborhood search algorithm to obtain the first transportation scheduling scheme corresponding to the first stage task data. The second task generation module is used to generate the second-stage task data of the logistics and distribution task based on the first transportation scheduling scheme. The second model solving module is used to construct the objective function corresponding to the second stage task data, and solve the objective function to obtain the second stage transportation scheduling scheme corresponding to the second stage task data. The result output module is used to merge the first-stage transportation scheduling scheme and the second-stage transportation scheduling scheme to output the complete transportation scheduling scheme corresponding to the logistics and distribution task.

[0015] Thirdly, the present invention provides a terminal, comprising: a processor and a memory, wherein the memory stores a multi-vehicle cooperative transportation scheduling program, which, when executed by the processor, is used to implement the operation of the multi-vehicle cooperative transportation scheduling method as described in the first aspect.

[0016] Fourthly, the present invention also provides a computer-readable storage medium storing a multi-vehicle cooperative transportation scheduling program, which, when executed by a processor, is used to implement the operation of the multi-vehicle cooperative transportation scheduling method as described in the first aspect.

[0017] The present invention, by employing the above technical solution, has the following effects: This invention discloses a transportation scheduling method, system, terminal, and storage medium based on multi-vehicle collaboration, comprising: acquiring first-stage task data of a logistics delivery task; constructing a collaborative transportation network model based on the first-stage task data; solving the collaborative transportation network model using an adaptive large neighborhood search algorithm to obtain a first transportation scheduling scheme corresponding to the first-stage task data; generating second-stage task data of the logistics delivery task based on the first transportation scheduling scheme; constructing an objective function corresponding to the second-stage task data; solving the objective function to obtain a second-stage transportation scheduling scheme corresponding to the second-stage task data; and merging the first-stage and second-stage transportation scheduling schemes to output a complete transportation scheduling scheme corresponding to the logistics delivery task. This invention can solve the spatiotemporal synchronization problem of multi-stage transfers in transportation scheduling schemes and improve the flexibility of last-mile delivery. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0019] Figure 1 This is a flowchart of the transportation scheduling method based on multi-vehicle collaboration in this invention.

[0020] Figure 2 This is a schematic diagram of a logistics transportation scheduling scheme corresponding to a logistics distribution task in one implementation of the present invention.

[0021] Figure 3 This is a schematic diagram of the algorithm flow for iterative optimization of the initial solution in one implementation of the present invention.

[0022] Figure 4 This is a schematic diagram of a transportation scheduling system based on multi-vehicle collaboration in one implementation of the present invention.

[0023] Figure 5 This is a functional schematic diagram of the terminal in one implementation of the present invention.

[0024] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0026] Exemplary methods Traditional urban logistics typically employs a two-stage "port-warehouse-customer" model, where container trucks transport goods from the port to inland transit warehouses, and then trucks depart from the transit warehouses to complete last-mile delivery. However, this traditional single-ground transportation model suffers from significant efficiency bottlenecks and cost pressures.

[0027] Using drones to assist ground vehicles in air-ground collaborative delivery is a popular research direction. However, existing technologies focus on the single last-mile delivery stage, neglecting the spatiotemporal coupling between container trunk transportation from ports to transit warehouses and last-mile delivery, and lacking optimization of resource allocation for the entire system. In urban scenarios, building height also affects delivery patterns, but most existing delivery models treat customers as planar point coordinates, ignoring service constraints in the vertical dimension. Furthermore, there is a strong structural dependency between drone and truck collaborative transportation. Drones can flexibly dock between different trucks. This strong coupling characteristic makes it easy for heuristic algorithms to get trapped in local optima when dealing with large-scale instances. Operator selection lacks adaptability, making it difficult to balance search efficiency and solution quality. Therefore, how to construct a system transportation scheduling scheme is a technical challenge that urgently needs to be solved in the field of smart logistics.

[0028] Existing technologies still have limitations, such as the inability of transportation scheduling schemes to achieve time and space synchronization in multi-stage transfers and their inability to adapt to the characteristics of drone transportation. Therefore, existing technologies need further improvement.

[0029] To address the above technical problems, this invention provides a transportation scheduling method based on multi-vehicle collaboration, comprising: acquiring first-stage task data for a logistics delivery task; constructing a collaborative transportation network model based on the first-stage task data; solving the collaborative transportation network model using an adaptive large neighborhood search algorithm to obtain a first transportation scheduling scheme corresponding to the first-stage task data; generating second-stage task data for the logistics delivery task based on the first transportation scheduling scheme; constructing an objective function corresponding to the second-stage task data; solving the objective function to obtain a second-stage transportation scheduling scheme corresponding to the second-stage task data; and merging the first-stage and second-stage transportation scheduling schemes to output a complete transportation scheduling scheme corresponding to the logistics delivery task. This application can solve the spatiotemporal synchronization problem of multi-stage transfer in transportation scheduling schemes and improve the flexibility of last-mile delivery.

[0030] like Figure 1 As shown, this embodiment of the invention provides a transportation scheduling method based on multi-vehicle cooperation, including the following steps: Step S100: Obtain the first-stage task data of the logistics and distribution task, and construct a collaborative transportation network model based on the first-stage task data.

[0031] Specifically, in one implementation of this embodiment, step S100 includes the following steps: Step S101: Obtain the first phase of task data for the logistics and delivery task.

[0032] In this embodiment, the first-stage task data obtained includes any one or more combinations of: order request data, first ground vehicle resource data, second ground vehicle resource data, air vehicle resource data, and warehouse data.

[0033] Specifically, the order request data includes order floor attributes, delivery order requirements, pickup order requirements, and time windows. Different order floor attributes correspond to different vehicles; for example, customers on higher floors use air vehicles for service, while customers on lower floors can use either air vehicles or secondary ground vehicles.

[0034] The first ground vehicle resource data includes the number, location, carrying capacity, and mission time of the first ground vehicle; in this embodiment, the first ground vehicle is a truck, used to transport containers containing goods from the port to the transit warehouse. The starting point of the first ground vehicle is the port, and the destination is the transit warehouse.

[0035] The second ground vehicle resource data includes the number, location, carrying capacity, and mission time of the second ground vehicles; in this embodiment, the second ground vehicle is a truck, which is used to transport goods from the transit warehouse to the customer's location.

[0036] The aerial vehicle resource data includes the number, location, carrying capacity, and mission time of aerial vehicles; in this embodiment, the aerial vehicle used is a drone, which is used to transport goods from the location of the truck to the location of the customer.

[0037] Warehouse data includes port data and transit warehouse data, corresponding to the starting point of each stage of the transportation task.

[0038] In this embodiment, the logistics transportation scheduling scheme corresponding to the logistics delivery task is as follows: Figure 2 As shown, in this scheme, nodes are divided into three types: ports, transit warehouses, and customers, corresponding to warehouse data and order request data. The transportation route of the truck is the first ground vehicle, the transportation route of the truck 1 is the second ground vehicle, the transportation route of the drone is the air vehicle, and the transportation route of the truck 2 is the transportation route for the second ground vehicle and the air vehicle to jointly perform the transportation task.

[0039] Step S102: Determine the objective function of the collaborative transportation network model based on the order request data and the warehouse data.

[0040] In this embodiment, the objective function of the collaborative transportation network model is determined based on the order request data and the warehouse data. Specifically, the order request data and warehouse data are initialized using Python, and the geographical location, time window, delivery volume, pickup volume, and floor attributes of each transportation request in the order request data are parsed to minimize the total transportation cost as the objective function.

[0041] In this embodiment, the total transportation cost includes the transportation cost of the first ground vehicle. Second ground vehicle transportation costs Air transport costs Vehicle fixed commissioning cost and transit waiting time costs The objective function can be expressed as: ; = ; = ; = ; = ; = ; in, This represents the total transportation cost of the cooperative transportation network model; This represents the first set of ground vehicles, which in this embodiment corresponds to the set of trucks / tractors; This represents the second set of ground vehicles, which in this embodiment corresponds to the set of trucks; This represents a collection of air vehicles, which in this embodiment corresponds to a collection of drones; This represents the set of nodes that the first ground vehicle can pass through, including port nodes, transit warehouse nodes, tractor warehouse nodes, and trailer warehouse nodes; This represents the set of nodes that can be served by the second ground vehicle and air vehicle, including transit warehouse nodes and customer nodes; Represents the set of transit warehouse nodes; Represents a node With nodes The transportation distance between them.

[0042] This represents the unit distance transportation cost of the first ground vehicle; This represents the unit distance transportation cost of the second ground vehicle; This represents the unit distance transportation cost of air vehicles. Indicates vehicle The fixed activation cost incurred when the device is activated; Indicates the cost per unit of waiting time; Indicates the first ground vehicle From node Drive to the node ; Indicates the second ground vehicle From node Drive to the node ; Indicates air vehicles From node Fly to the node ; Indicates vehicle Is it enabled? and They represent the first ground vehicle. Arrival and departure from transit warehouse nodes The time.

[0043] In the above objective function, the transportation cost of the first ground vehicle This represents the transportation costs incurred by the first ground vehicle in performing container transport tasks between ports, transit warehouses, and vehicle depots; and the transportation costs of the second ground vehicle. This represents the transportation costs incurred by a second ground vehicle in performing delivery or pickup tasks between a transit warehouse and a customer node; air vehicle transportation costs. This represents the flight costs incurred by air vehicles performing delivery or pickup missions between transit warehouses and customer nodes; vehicle fixed activation costs. This represents the fixed activation cost of the activated vehicle; transit waiting time cost. This represents the waiting time cost incurred by the first ground vehicle at the transit warehouse node.

[0044] Step S103: Determine the constraints of the collaborative transportation network model based on the first ground vehicle resource data, the second ground vehicle resource data, the air vehicle resource data, and the warehouse data.

[0045] The constraints include mission service constraints, path continuity constraints, vehicle capacity constraints, time window constraints, floor service constraints, air vehicle endurance constraints, ground-air coordination constraints, and phase transition constraints.

[0046] The task service constraints are used to ensure that the delivery and pickup needs of each customer node are met, and that each customer node completes its service in the same transportation task using only one feasible service method, avoiding duplicate or missed services. For example, for customer nodes... Its delivery needs can be served directly by the second ground vehicle, or by an air vehicle launched from the node where the second ground vehicle is located, but it cannot be delivered repeatedly in the same mission.

[0047] The path continuity constraint ensures that the transport routes of the first ground vehicle, the second ground vehicle, and the air vehicle all meet the requirements of origin, destination, and flow conservation. For example, the first ground vehicle departs from a port, tractor warehouse, or trailer warehouse and arrives at the corresponding transit warehouse; the second ground vehicle departs from the transit warehouse, returns to or arrives at the designated transit warehouse after completing customer node service; the air vehicle is launched from the node where the second ground vehicle is located and is recovered to the node where the second ground vehicle is located or another feasible recovery node after completing customer service. This type of constraint can be achieved through the flow conservation relationship that "the number of paths entering a node equals the number of paths leaving that node."

[0048] The container allocation and transshipment warehouse selection constraints are used to ensure that each container is assigned to a corresponding trailer and to determine its destination transshipment warehouse. For example, each container may be assigned to only one trailer for a transport task, and each container may select only one transshipment warehouse as its unloading or transshipment destination. These constraints ensure consistency in cargo flow between port-to-transshipment warehouse transport tasks and subsequent last-mile delivery tasks.

[0049] The vehicle capacity constraints are used to ensure that the load of various vehicles during transportation does not exceed their maximum carrying capacity. For example, the sum of the remaining delivery amount and the cumulative pickup amount of a second ground vehicle on any road segment does not exceed its maximum load; the load of an air vehicle in any flight mission does not exceed its maximum load. The model in the paper also imposes constraints on the load updates and capacity limits for trucks and UAVs.

[0050] The time window constraint is used to ensure that the service start time of the client node meets the corresponding earliest and latest service time requirements.

[0051] The floor service constraint is used to determine the available service mode based on the floor attribute of the customer node. For example, customers on higher floors are served by air vehicles due to longer manual delivery times or lower ground delivery efficiency; customers on lower floors can be served by a second ground vehicle, or by air vehicles provided that the load capacity, range, and time window requirements are met. In other words, customers on higher floors are required to use air vehicles, while customers on lower floors can choose between a second ground vehicle or an air vehicle.

[0052] The air vehicle endurance constraint is used to ensure that the total flight distance, total flight time, or cumulative energy consumption of the air vehicle from the launch node, serving the customer node, and arriving at the recovery node does not exceed its maximum endurance.

[0053] The phase connection constraints are used to ensure that the container transportation from the port to the transit warehouse in the first phase can be connected with the last-mile delivery from the transit warehouse to the customer node, and to ensure that the goods picked up after the customer's pickup task is completed can be collected in the corresponding transit warehouse to generate the second phase container transportation task.

[0054] Step S104: Construct the cooperative transportation network model based on the objective function and the constraints.

[0055] In this embodiment, a cooperative transportation network model is constructed based on the objective function and constraints. The selection process is modeled as a Markov decision process and defined as a quadruple. The state, action, reward, and state transition correspond to the "solution features - operator decision - search feedback - solution evolution" in ALNS (Adaptive Large Neighborhood Search).

[0056] like Figure 1 As shown, this embodiment of the invention provides a transportation scheduling method based on multi-vehicle cooperation, including the following steps: Step S200: Solve the cooperative transportation network model using an adaptive large neighborhood search algorithm to obtain the first transportation scheduling scheme corresponding to the first stage task data.

[0057] Specifically, in one implementation of this embodiment, step S200 includes the following steps: Step S201: Generate an initial solution for the cooperative transportation network model based on a greedy algorithm; wherein the initial solution is used to plan transportation routes for the first ground vehicle, the second ground vehicle, and the air vehicle.

[0058] In this embodiment, the generation of the initial feasible solution using the reinforcement learning-heuristic transportation scheduling method for multi-vehicle cooperation of trucks, vans, and drones fully considers the multi-level structure of the container-truck-drone system and maintains consistency with the subsequent cascaded search mechanism. The construction process of the initial solution is divided into three stages: the first stage corresponding to the first ground vehicle, the second stage corresponding to the second ground vehicle, and the third stage corresponding to the air vehicle.

[0059] Step S2011: Construct a first transportation route for the first ground vehicle; wherein the starting point of the first transportation route is the starting point of the logistics distribution task, and the ending point of the first transportation route is a transit warehouse.

[0060] In this embodiment, the first stage corresponding to the first ground vehicle is the first transportation route. The starting point of the first transportation route is the starting point of the logistics distribution task, and the ending point of the first transportation route is the ending point of the transit warehouse. The first transportation route in this embodiment is a truck transportation route, which is used to assign a transportation path executed by a tractor-trailer combination for each container, that is, to transport the trailer carrying the container from the port to the designated transit warehouse by a tractor-trailer.

[0061] Step S2012: Construct a second transportation route for the second ground vehicle based on a greedy algorithm; wherein the starting point of the second transportation route is the transit warehouse, and the ending point of the second transportation route is the end point of the logistics delivery task.

[0062] The second stage, corresponding to the second ground vehicle, is the second transportation route. The starting point of the second transportation route is the transit warehouse, and the ending point is the midpoint of the logistics delivery task. Constructing the second transportation route using a greedy algorithm involves employing a greedy insertion strategy to build the truck route, sequentially inserting customer nodes and checking capacity and time constraints.

[0063] Step S2013: Construct a third transportation route for the air vehicle based on the second transportation route; wherein the starting point of the third transportation route is the location of the second ground vehicle on the second transportation route, and the ending point of the third transportation route is the end point of the logistics delivery task.

[0064] The third stage corresponding to the airborne vehicle is the third transportation route. The starting point of the third transportation route is the second ground vehicle of the second transportation route, and the ending point is the end point of the logistics delivery task. Therefore, after the second transportation route is established, the third transportation route needs to transfer goods that meet the constraints of the airborne vehicle from the second ground vehicle to the airborne vehicle to construct a launch-service-recovery third transportation route. Specifically, the goods are transferred from the truck to the drone, constructing the drone sub-route.

[0065] Step S202: Use an adaptive large neighborhood search algorithm to iteratively optimize the initial solution to obtain the first transportation scheduling scheme corresponding to the first stage task data.

[0066] In this embodiment, the adaptive large neighborhood search algorithm used is a reinforcement learning-enhanced cascaded adaptive large neighborhood search algorithm, which iteratively optimizes the initial solution, such as... Figure 3 The diagram shown illustrates the algorithm flow for iterative optimization of the initial solution in this embodiment. In this embodiment, a reinforcement learning-enhanced cascaded adaptive large neighborhood search algorithm is constructed, and the initial solution is input into this algorithm for iterative optimization. By introducing a reinforcement learning agent's operator selection strategy into the "destruction-reconstruction" iterative framework of the adaptive large neighborhood search algorithm, a replacement is found for the traditional adaptive large neighborhood search algorithm's operator selection method based on fixed scores or heuristic weight updates.

[0067] Step S202 includes the following steps: Step S2021: Based on the adaptive large neighborhood search algorithm, define the state space, action space and reward function, and construct a Markov decision process model.

[0068] In this embodiment, a cascaded adaptive large neighborhood search algorithm based on reinforcement learning is used to define a state space, action space, and reward function, and to establish a Markov decision process model. In each iteration, the reinforcement learning agent trained using proximal policy optimization dynamically selects a pair of destruction operators and repair operators from the operator set according to the current search state.

[0069] Specifically, the operator selection process is modeled as a Markov Decision Process (MDP), defined as a quadruple. The state, action, reward, and state transition correspond to "solution features - operator decision - search feedback - solution evolution" in ALNS. state It is designed as a composite vector, composed of three concatenated parts: ; in, Used to characterize the update status of the solution in the current search phase; Used to constrain the degree of slack; Temporal context information used to constrain the operation.

[0070] action The operator selection decision in one ALNS iteration is defined as a combination of the destruction operator and the reconstruction operator: ; in, This indicates a destructive operator used to remove parts of an order or node. This represents the reconstruction operator used to re-insert the removed order into the solution.

[0071] The reward function is used to evaluate the contribution of operator selection to the search process. Taking into account both the solution update and the improvement of the objective function, it is defined as: ; in, This indicates the contribution of a candidate solution to the update of the current solution or the global optimal solution; Represents the relative improvement of the objective function; coefficients Used to balance the weights between stable search and objective optimization.

[0072] In this embodiment, the reward design ensures sensitivity to improvements in the objective function while avoiding the high variance problem caused by relying solely on minute numerical changes by using discrete encoding of "whether to update the current solution / global optimal solution", thus facilitating stable training of the policy.

[0073] Furthermore, within the aforementioned MDP modeling framework, this embodiment employs Proximal Policy Optimization (PPO) to learn the operator selection policy, replacing the traditional ALNS-based operator selection mechanism that updates based on fixed scores or empirical weights. PPO belongs to the policy gradient method, directly optimizing the parameterized policy. It can stably learn the mapping relationship between "state-operator" in discrete action space, and is suitable for iterative and non-stationary search environments in RL-ALNS (Reinforcement Learning-based Adaptive Large Neighborhood Search).

[0074] In CTD-RL-CALNS, each ALNS iteration is considered a reinforcement learning decision step: the agent is in a state Select Action (i.e., the combination of destruction-reconstruction operators), the environment performs the corresponding neighborhood search and updates the current solution according to the simulated annealing criterion, and then returns the reward. and the next state Through continuous interaction with the search process, PPO learns more appropriate operator selection strategies at different search stages.

[0075] The core idea of ​​PPO is to limit the magnitude of policy changes between adjacent updates to ensure the stability of the training process. In this embodiment, the new and old policies are defined in state... Next action The probability ratio is: ; in, This represents the strategy parameters before the update.

[0076] Based on the advantage function PPO employs a shearing strategy objective function: ; in This is the shearing threshold, used to constrain the policy update magnitude and prevent policy oscillations caused by excessive improvement in a single search. For expectations; This is the shearing function. The dominant function. The reward, estimated by the reward signal and the value function, comprehensively reflects the update of the solution and the improvement of the objective. Therefore, the optimization objective can directly guide the strategy to favor the selection of operators that generate higher search returns.

[0077] After training is complete, fix the learned strategy. Furthermore, during the CTD-RL-CALNS search process, it directly outputs the operator selection decision, thereby achieving intelligent takeover of the operator selection phase. In this embodiment, PPO is only used to learn the operator selection strategy and does not change the "destruction-reconstruction-acceptance" search framework of ALNS; its role is to realize the transformation of operator selection from static heuristic rules to an adaptive decision-making mechanism based on experience and state feedback by perceiving the search state.

[0078] Step S2022: Generate a destruction operator and a repair operator, and perform an iterative solution process for a preset number of times based on the destruction operator, the repair operator, and the simulated annealing algorithm.

[0079] In this embodiment, by executing the selected destruction and repair operators, a cascading mechanism is used to handle the strong coupling relationship between the truck route and the UAV sub-route, synchronously updating the main truck route and the UAV sub-route dependent on the truck. A simulated return criterion is used to determine whether to accept the newly generated solution and to update the policy parameters of the reinforcement learning agent until the termination condition is met, outputting the first transportation scheduling scheme corresponding to the first stage task data.

[0080] It should be noted that the cascading mechanism includes both cascading destruction and multi-node cascading repair mechanisms.

[0081] The cascading destruction mechanism is defined as follows: when a client node is selected for removal, if the node participates in the drone sub-route service as a drone service node, launch node, or recovery node, a recursive deletion is triggered, which simultaneously removes all drone sub-routes associated with the node structure and the client nodes contained therein, until the deletion set no longer expands.

[0082] The node cascading repair mechanism is defined as treating the customer nodes to be inserted as an associated set, selecting seed customer nodes and constructing a candidate sequence containing multiple customers, and simultaneously constructing or merging a feasible truck-drone path that satisfies the drone's payload, range, and collaborative constraints for multiple customers in a single repair operation.

[0083] In this embodiment, the destruction operators include: random customer deletion operator, greedy deletion operator, correlation deletion operator, and cascading destruction operator.

[0084] The random customer deletion operator randomly selects a certain proportion of customer nodes from the truck path and deletes them to enhance the randomness and diversity of the search process. The greedy deletion operator evaluates the marginal impact of customer nodes on the objective function, prioritizing the deletion of customer nodes that significantly increase system costs, thereby enhancing local search capabilities; for any customer node Its marginal contribution is defined as ,according to Sort by largest to smallest, then select the first... The deletion set consists of customer nodes. ; The relevance deletion operator selects a group of spatially close customers for simultaneous deletion based on the spatial distance relationship between customer nodes, thereby achieving concentrated perturbation of the local path structure. A customer node is randomly selected. Define customers as seed nodes. and The correlation measure is: ;in, Represents a node and The distance between them. According to... Sort by size from smallest to largest and select the first few. Each customer node is used as the deletion set. ; When a deleted customer node participates in UAV subpath service (i.e., as a UAV service node, launch node, or recovery node), the cascading destruction operator triggers a cascading destruction mechanism. All UAV paths associated with that node are simultaneously removed, thus avoiding incomplete or infeasible collaborative structures. This mechanism enables consistent destruction across levels, effectively expanding the search neighborhood.

[0085] In this embodiment, the repair operators include: a global greedy repair operator, a local greedy repair operator, a regret value repair operator, and a multi-node cascade repair operator; The global greedy repair operator applies to each client to be inserted. Enumerate all feasible truck insertion positions and drone service solutions, and select the solution with the minimum objective function value while satisfying various constraints. Let... For customers Given a set of feasible insertion positions, the optimal insertion position is: ,in Indicates customers Insertion position The new solution obtained later; ; To reduce computational complexity, the locally greedy repair operator only inserts a truck along a randomly selected, insertable truck path. Perform the insertion operation above; The regret-repair operator prioritizes inserting customer nodes that "least should be delayed" by weighing the potential loss of delayed insertion; let the customer... The objective function values ​​corresponding to the first two optimal insertion schemes are respectively Then its regret value is defined as In each round, the customer node with the highest regret value is selected for insertion. ; ; ; The multi-node cascaded repair operator treats deleted customer nodes as a potential set of associations, rather than independent insertion objects, and attempts to restore the broken cooperative structure by constructing / incorporating a feasible truck-drone path for multiple customers simultaneously through a single repair operation. A seed customer node is selected. Construct a candidate customer sequence: ; in, The maximum number of customers that a drone is allowed to serve in a single flight; Constructing the drone path: ; in, and Let represent the launch and recovery nodes of the drone, respectively. This path must satisfy the following constraints. Among all feasible candidates, select the solution with the minimum incremental cost. And complete the repair of multiple nodes at once. Among them, the solution with the lowest incremental cost is: .

[0086] Based on the above, this embodiment, based on the destruction operator, the repair operator, and the simulated annealing algorithm, performs an iterative solution process a preset number of times, including: Step a: Take the initial solution as the target solution of the cooperative transportation network model, and select a pair of destruction operators and repair operators in each iteration. Step b: Destroy the solution corresponding to the destruction operator; Step c: Repair the solution corresponding to the repair operator and output the new solution; Step d: Based on the simulated annealing algorithm, determine whether to accept the new solution. If the new solution is accepted, take it as the target solution; otherwise, execute the next iteration. Step e: When the number of iterations meets the condition, output the target solution of the cooperative transportation network model.

[0087] Step S2023: Output the target solution of the collaborative transportation network model, and use the transportation scheduling scheme corresponding to the target solution as the first transportation scheduling scheme corresponding to the first stage task data.

[0088] In this embodiment, when the number of iterations meets the condition, the target solution of the cooperative transportation network model is output, and the transportation scheduling scheme corresponding to the target solution is used as the first transportation scheduling scheme corresponding to the first stage task data.

[0089] like Figure 1 As shown, this embodiment of the invention provides a transportation scheduling method based on multi-vehicle cooperation, including the following steps: Step S300: Based on the first transportation scheduling scheme, generate the second stage task data for the logistics and distribution task.

[0090] Specifically, after the container transportation of the first-stage delivery order is completed, the tractor and trailer return to the tractor warehouse and trailer warehouse respectively; after the second ground vehicle and air vehicle complete the customer pickup task and transport the picked-up goods to the corresponding transit warehouse for collection and packing, based on the first-stage delivery plan output in step S200, the container export demand of each transit warehouse is obtained, and the second-stage task data of the logistics delivery task is generated.

[0091] The second-stage task data is generated based on the pickup results, transit warehouse aggregation results, and container loading results from the first transportation scheduling plan. The second-stage task data includes output container task sets, transit warehouse sets, port node sets, tractor resource data, trailer resource data, container resource data, node distance data, node transportation time data, and transportation cost parameters.

[0092] The set of outgoing container tasks represents container transport tasks that need to be transported from a transit warehouse back to the port. Each outgoing container task includes a task number, originating transit warehouse, destination port, container number or quantity, earliest available departure time, and corresponding cargo origin information.

[0093] The tractor resource data includes the number of tractors, their initial positions, available time, maximum travel distance, and unit distance transportation cost; the trailer resource data includes the number of trailers, their initial positions, available time, and the number of containers each trailer can carry; the container resource data includes the number of containers to be returned, the location of the transit warehouse where the containers are located, the container loading completion time, and the target port of the containers.

[0094] The distance data and transportation time data between nodes are used to calculate the travel cost and time cost when a tractor-trailer performs a container transportation task. The transportation cost parameters include the tractor-trailer unit distance cost, the tractor-trailer fixed usage cost, and the unit waiting time cost.

[0095] The second phase of task data corresponds to the first transportation scheduling scheme: after the second ground vehicle and air vehicle complete the customer pickup task and transport the picked-up goods back to the transit warehouse in the first phase, the second phase of container transportation task from the transit warehouse to the port is generated based on the quantity of picked-up goods, the time of container loading completion, and the target port information gathered in each transit warehouse.

[0096] like Figure 1 As shown, this embodiment of the invention provides a transportation scheduling method based on multi-vehicle cooperation, including the following steps: Step S400: Construct the objective function corresponding to the second stage task data, and solve the objective function to obtain the second stage transportation scheduling scheme corresponding to the second stage task data.

[0097] In this embodiment, the second-stage transportation scheduling scheme is used to schedule the return transport of containers for goods that have been picked up and gathered at the transit warehouse in the first stage. Since the transportation object in the second stage is the output container loaded with the picked-up goods, the transportation method is consistent with the container transportation mode in the first stage, that is, the container transportation task is completed by the tractor-trailer in the first ground vehicle towing the trailer. Therefore, the objective function of the second stage can be constructed with reference to the container transportation cost item, the vehicle fixed activation cost item, and the waiting time cost item in the objective function of the first stage. The transportation solution unit is called to calculate the optimal scheduling and path of the tractor-trailer combination, with minimizing the total transportation cost as the objective function.

[0098] In this embodiment, the objective function of the second stage is used to minimize the total transportation cost incurred in transporting the output container from the transit warehouse to the port. The total transportation cost includes the cost of the first ground vehicle transportation. The first ground vehicle fixed commissioning cost and transit waiting time costs The objective function can be expressed as: ; = ; = ; = ; in, This represents the total cost of the second phase of container transport scheduling.

[0099] In the above objective function, the transportation cost of the first ground vehicle This represents the transportation costs incurred by the first ground vehicle in performing outbound container transport tasks between transit warehouses, ports, tractor-trailer warehouses, and trailer warehouses; transit waiting time costs. This represents the waiting time cost incurred by the first ground vehicle at the transit warehouse node due to waiting for container loading, mission coordination, or node services; the fixed activation cost of the first ground vehicle. This represents the fixed activation cost of the first ground vehicle to be activated.

[0100] The parameters involved in the objective function of the second stage include: the output container task set, the transit warehouse node set, the port node set, the tractor set, the trailer set, the container set, the distance between nodes, the transportation time between nodes, the unit distance transportation cost of the first ground vehicle, the fixed activation cost of the first ground vehicle, the unit waiting time cost, the container loading completion time, and the container loading and unloading service time.

[0101] By solving the objective function of the second stage, the matching relationship between the tractor and the trailer, the transportation sequence of the output containers, the starting transit warehouse, the destination port and the corresponding transportation route of each output container task are determined, thereby obtaining the second stage transportation scheduling scheme.

[0102] Furthermore, the second-stage transportation scheduling process satisfies the following constraints: task execution, tractor route, trailer route, tractor-trailer coordination, container allocation, travel distance, and stage transition time.

[0103] The task execution constraints are used to ensure that each outgoing container task is executed once. For example, any outgoing container task from a transit warehouse to a port should be assigned to a trailer and towed by a tractor to complete the transport, avoiding situations where outgoing containers are not transported or are transported repeatedly.

[0104] The tractor route constraints are used to ensure that the tractor transport route meets the requirements of origin, destination, and flow conservation. For example, each activated tractor departs from the tractor warehouse and returns to the tractor warehouse after completing one or more output container transport tasks; for intermediate nodes visited by the tractor, the number of paths entering that node should be equal to the number of paths leaving that node.

[0105] The trailer route constraints are used to ensure that trailer transport routes meet the requirements of origin, destination, and flow conservation. For example, each activated trailer departs from the trailer warehouse and returns to the trailer warehouse after completing the outgoing container transport task; for transit warehouses or port nodes visited by trailers, the number of paths entering the node should be equal to the number of paths leaving the node.

[0106] The tractor-trailer coordination constraint is used to ensure that the trailer can only move after being towed by the tractor, and that each tractor can tow at most one trailer at any given time.

[0107] The container allocation constraint is used to ensure that each outgoing container is assigned to a trailer for transport. For example, after an outgoing container loaded with pickup goods is packed at a transit warehouse, it should be uniquely assigned to a trailer, which should then be towed by a tractor unit from the corresponding transit warehouse to the target port.

[0108] The driving distance constraint is used to ensure that the total driving distance of the tractor and trailer does not exceed their maximum allowable driving distance. For example, when the tractor performs transportation tasks between the tractor warehouse, trailer warehouse, transit warehouse and port node, its cumulative driving distance shall not exceed the preset maximum driving distance; the cumulative transportation distance generated by the trailer during the movement with the tractor shall also not exceed its preset limit.

[0109] The aforementioned stage transition time constraint is used to ensure that the start time of the second stage container output task is not earlier than the time when the picked-up goods in the first stage are collected and packed in the transit warehouse. For example, the tractor can only pull the trailer to perform the container output transportation task from the transit warehouse to the port after the second ground vehicle and air vehicle have transported the customer's picked-up goods back to the transit warehouse and completed the corresponding container loading.

[0110] By using the same technical solution as in step S200, the objective function is solved to obtain the second-stage transportation scheduling scheme corresponding to the second-stage task data. The second-stage transportation scheduling scheme includes the loading completion time of each output container, the trailer number undertaking the transportation task, the matching relationship between the tractor and the trailer, the tractor's driving path, the trailer's following path, the arrival and departure times of each node, and the total cost of the second-stage transportation scheduling.

[0111] The tractor travel path represents the movement sequence of the tractor between the tractor warehouse, trailer warehouse, transit warehouse, and port node; the trailer following path represents the movement sequence of the trailer while being towed by the tractor to perform container transportation tasks; the tractor-trailer matching relationship represents whether the tractor tows the corresponding trailer in each transportation segment; the node arrival time and departure time represent the time scheduling results of the tractor at each node; and the total transportation scheduling cost of the second stage is used to evaluate the transportation cost, fixed activation cost, and waiting time cost of the output container during the process of returning from the transit warehouse to the port.

[0112] like Figure 1 As shown, this embodiment of the invention provides a transportation scheduling method based on multi-vehicle cooperation, including the following steps: Step S500: Merge the first-stage transportation scheduling scheme and the second-stage transportation scheduling scheme, and output the complete transportation scheduling scheme corresponding to the logistics and distribution task.

[0113] In this embodiment, the first-stage transportation scheduling scheme and the second-stage transportation scheduling scheme are merged, and the merged scheme is output as a complete transportation scheduling scheme for the logistics and distribution task, namely, the container transportation and last-mile delivery scheme.

[0114] This embodiment achieves the following technical effects through the above technical solution: By extracting floor attribute features, time window constraints, and spatiotemporal coupling features among multiple vehicles (trucks, freight trucks, and drones) in urban last-mile delivery tasks, and combining them with logistics transfer data from ports to transit warehouses, we can conduct an in-depth analysis of the operational efficiency and collaborative relationships of different vehicles in complex urban environments. To address the challenges of delivery to high-rise buildings and the high complexity of ground-air coordination, this study utilizes reinforcement learning agents to dynamically perceive search states and make decisions on operator combinations. It delves into the impact of cascading destruction and repair mechanisms on strongly coupled ground-air paths, constructing a delivery model that better suits real-world urban scenarios and providing decision support for intelligent urban logistics scheduling. To address the current problems in urban last-mile delivery, such as resource waste due to poor vehicle coordination, slow service response in high-rise buildings, and disconnect between trunk lines and last-mile delivery, a reinforcement learning-enhanced cascaded adaptive large neighborhood search algorithm is constructed. This algorithm fully considers high-rise / low-rise delivery constraints and combines cross-stage transportation connections to achieve joint optimization scheduling of container trucks, freight trucks, and drones. To address the problem of inefficient logistics resource scheduling, this project aims to achieve a more comprehensive multi-level transfer and air-ground collaborative transportation scheduling system. It utilizes a cascading mechanism to effectively handle the structural dependence between trucks and drones, providing a more robust solution for intelligent scheduling systems. This helps logistics companies to more rationally allocate transportation resources under complex constraints, significantly reducing the total transportation cost of the system.

[0115] Exemplary device Based on the above embodiments, the present invention also provides a transportation scheduling system based on multi-vehicle cooperation, such as... Figure 4 As shown, a transportation scheduling system based on multi-vehicle collaboration includes: Data acquisition module 21 is used to acquire the first-stage task data of the logistics and distribution task and to build a collaborative transportation network model based on the first-stage task data. The first model solving module 22 is used to solve the cooperative transportation network model using an adaptive large neighborhood search algorithm to obtain the first transportation scheduling scheme corresponding to the first stage task data. The second task generation module 23 is used to generate the second stage task data of the logistics and distribution task based on the first transportation scheduling scheme. The second model solving module 24 is used to construct the objective function corresponding to the second stage task data, and solve the objective function to obtain the second stage transportation scheduling scheme corresponding to the second stage task data. The result output module 25 is used to merge the first-stage transportation scheduling scheme and the second-stage transportation scheduling scheme to output the complete transportation scheduling scheme corresponding to the logistics and distribution task.

[0116] Based on the above embodiments, the present invention also provides a terminal, the principle block diagram of which can be as follows: Figure 5 As shown.

[0117] The terminal includes: a processor, a memory, an interface, a display screen, and a communication module connected via a system bus; wherein, the processor of the terminal provides computing and control capabilities; the memory of the terminal includes a computer-readable storage medium and internal memory; the computer-readable storage medium stores an operating system and computer programs; the internal memory provides an environment for the operation of the operating system and computer programs in the computer-readable storage medium; the interface is used to connect to external devices; the display screen is used to display relevant information; and the communication module is used to communicate with a cloud server or other devices.

[0118] When executed by a processor, this computer program is used to implement a transportation scheduling method based on multi-vehicle cooperation.

[0119] It will be understood by those skilled in the art that Figure 5 The schematic diagram shown is merely a partial structural diagram related to the present invention and does not constitute a limitation on the terminal to which the present invention is applied. A specific terminal may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0120] In one embodiment, a terminal is provided, comprising: a processor and a memory, the memory storing a multi-vehicle cooperative transportation scheduling program, which, when executed by the processor, is used to implement the operation of the multi-vehicle cooperative transportation scheduling method described above.

[0121] In one embodiment, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a multi-vehicle cooperative transportation scheduling program, which, when executed by a processor, is used to implement the operation of the multi-vehicle cooperative transportation scheduling method described above.

[0122] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, database, or other media used in the embodiments provided by this invention can include both non-volatile and volatile memory.

[0123] In summary, this invention provides a transportation scheduling method, system, terminal, and storage medium based on multi-vehicle collaboration, comprising: acquiring first-stage task data for a logistics delivery task; constructing a collaborative transportation network model based on the first-stage task data; solving the collaborative transportation network model using an adaptive large neighborhood search algorithm to obtain a first transportation scheduling scheme corresponding to the first-stage task data; generating second-stage task data for the logistics delivery task based on the first transportation scheduling scheme; constructing an objective function corresponding to the second-stage task data; solving the objective function to obtain a second-stage transportation scheduling scheme corresponding to the second-stage task data; and merging the first-stage and second-stage transportation scheduling schemes to output a complete transportation scheduling scheme corresponding to the logistics delivery task. This application can solve the spatiotemporal synchronization problem of multi-stage transfers in transportation scheduling schemes and improve the flexibility of last-mile delivery.

[0124] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A transportation scheduling method based on multi-vehicle cooperation, characterized in that, include: Obtain the first-stage task data of the logistics and delivery task, and construct a collaborative transportation network model based on the first-stage task data; wherein, the first-stage task data includes any one or more combinations of: order request data, first ground vehicle resource data, second ground vehicle resource data, air vehicle resource data, and warehouse data; The adaptive large neighborhood search algorithm is used to solve the cooperative transportation network model to obtain the first transportation scheduling scheme corresponding to the first stage task data. Based on the first transportation scheduling scheme, the second stage task data of the logistics and distribution task is generated; Construct an objective function corresponding to the second-stage task data, and solve the objective function to obtain the second-stage transportation scheduling scheme corresponding to the second-stage task data; Combine the first-stage transportation scheduling scheme and the second-stage transportation scheduling scheme to output the complete transportation scheduling scheme corresponding to the logistics and distribution task; The step of using an adaptive large neighborhood search algorithm to solve the cooperative transportation network model to obtain the first transportation scheduling scheme corresponding to the first stage task data includes: An initial solution for the cooperative transportation network model is generated based on a greedy algorithm; wherein, the initial solution is used to plan transportation routes for the first ground vehicle, the second ground vehicle, and the air vehicle; An adaptive large neighborhood search algorithm is used to iteratively optimize the initial solution to obtain the first transportation scheduling scheme corresponding to the first stage task data; wherein, the adaptive large neighborhood search algorithm is a cascaded adaptive large neighborhood search algorithm enhanced by reinforcement learning. The step of using an adaptive large neighborhood search algorithm to iteratively optimize the initial solution to obtain the first transportation scheduling scheme corresponding to the first stage task data includes: Based on the adaptive large neighborhood search algorithm, the state space, action space and reward function are defined, and a Markov decision process model is constructed. A destruction operator and a repair operator are generated, and an iterative solution process is performed a preset number of times based on the destruction operator, the repair operator, and the simulated annealing algorithm; wherein, in each iteration, a pair of destruction operators and repair operators are selected, and the destruction operator and the repair operator are optimized using a proximal policy optimization algorithm to learn an optimized operator selection strategy; Output the target solution of the cooperative transportation network model, and use the transportation scheduling scheme corresponding to the target solution as the first transportation scheduling scheme corresponding to the task data of the first stage.

2. The transportation scheduling method based on multi-vehicle collaboration according to claim 1, characterized in that, The construction of the collaborative transportation network model based on the task data of the first stage includes: Based on the order request data and the warehouse data, determine the objective function of the collaborative transportation network model; Based on the first ground vehicle resource data, the second ground vehicle resource data, the air vehicle resource data, and the warehouse data, the constraints of the collaborative transportation network model are determined. The cooperative transportation network model is constructed based on the objective function and the constraints.

3. The transportation scheduling method based on multi-vehicle collaboration according to claim 2, characterized in that, The initial solution for the cooperative transportation network model generated based on the greedy algorithm includes: Construct a first transportation route for the first ground vehicle; wherein the starting point of the first transportation route is the starting point of the logistics distribution task, and the ending point of the first transportation route is a transit warehouse; A second transportation route for the second ground vehicle is constructed based on a greedy algorithm; wherein the starting point of the second transportation route is the transit warehouse, and the ending point of the second transportation route is the end point of the logistics delivery task; A third transportation route for the air vehicle is constructed based on the second transportation route; wherein the starting point of the third transportation route is the location of the second ground vehicle on the second transportation route, and the ending point of the third transportation route is the end point of the logistics delivery task.

4. The transportation scheduling method based on multi-vehicle collaboration according to claim 1, characterized in that, The iterative solution process based on the destruction operator, the repair operator, and the simulated annealing algorithm, performing a preset number of iterations, includes: The initial solution is used as the target solution of the cooperative transportation network model. In each iteration, a pair of destruction operators and repair operators are selected. The solution corresponding to the destruction operator is destroyed; wherein, the destruction operator includes random client deletion operator, greedy deletion operator, correlation deletion operator and cascade destruction operator; Repair the solution corresponding to the repair operator and output a new solution; wherein, the repair operator includes: global greedy repair operator, local greedy repair operator, regret value repair operator and multi-node cascade repair operator; The simulated annealing algorithm is used to determine whether to accept the new solution. If the new solution is accepted, it is taken as the target solution; otherwise, the next iteration process is executed. When the number of iterations meets the condition, the target solution of the cooperative transportation network model is output.

5. A transportation scheduling system based on multi-vehicle collaboration, used to implement the transportation scheduling method based on multi-vehicle collaboration as described in any one of claims 1-4, characterized in that, include: The data acquisition module is used to acquire the first-stage task data of the logistics and distribution task, and to build a collaborative transportation network model based on the first-stage task data. The first model solving module is used to solve the cooperative transportation network model using an adaptive large neighborhood search algorithm to obtain the first transportation scheduling scheme corresponding to the first stage task data. The second task generation module is used to generate the second-stage task data of the logistics and distribution task based on the first transportation scheduling scheme. The second model solving module is used to construct the objective function corresponding to the second stage task data, and solve the objective function to obtain the second stage transportation scheduling scheme corresponding to the second stage task data. The result output module is used to merge the first-stage transportation scheduling scheme and the second-stage transportation scheduling scheme to output the complete transportation scheduling scheme corresponding to the logistics and distribution task.

6. A terminal, characterized in that, include: The processor and memory, wherein the memory stores a multi-vehicle cooperative transportation scheduling program, which, when executed by the processor, is used to implement the operation of the multi-vehicle cooperative transportation scheduling method as described in any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a multi-vehicle cooperative transportation scheduling program, which, when executed by a processor, is used to implement the operation of the multi-vehicle cooperative transportation scheduling method as described in any one of claims 1-4.

Images