A large-scale air-ground cooperative path planning method based on hybrid evolutionary algorithm

Abstract

The present application relates to a kind of large-scale air-ground cooperative path planning method based on hybrid evolutionary algorithm, belong to position and fairway control technical field, solve the prior art, without considering unmanned aerial vehicle can continuously serve multiple user points and can land on different transport car, there is the problem of high distribution cost, low distribution efficiency.The method obtains the description information of problem scene by obtaining actual distribution information, then constructs multi-transport car multi-unmanned aerial vehicle logistics distribution path planning model;Based on hybrid evolutionary algorithm, the model is solved, and the path planning scheme of the transport car and unmanned aerial vehicle is obtained.The method constructs model, designs algorithm, and unmanned aerial vehicle can continuously serve multiple user points and can land on different transport car, solve the multi-transport car multi-unmanned aerial vehicle path planning problem in logistics distribution, facilitate the flexibility of distribution scheme in actual logistics, reduce distribution cost, improve distribution efficiency.

Description

Technical Field

[0001] This invention relates to the field of location and airway control technology, and in particular to a large-scale air-ground cooperative path planning method based on a hybrid evolutionary algorithm. Background Technology

[0002] In recent years, with the rapid development of the social economy, the demand for drones in various fields has been increasing daily. Especially with the rapid development of the e-commerce industry, people have increasingly higher requirements and expectations for logistics services. The impact of the COVID-19 pandemic has led to the emergence of various shopping methods such as community group buying and online grocery shopping, further strengthening people's expectations for high-quality, efficient logistics and delivery.

[0003] Compared to traditional delivery vehicles, drones offer significant advantages in delivery speed and route selection, but their drawbacks are equally undeniable. For example, drones have limited payload capacity and insufficient range, making it difficult for them to independently complete the entire delivery process. However, coordinating drones and delivery vehicles can leverage their complementary strengths, greatly improving delivery efficiency and enhancing logistics service levels. Currently, existing solutions for dispatching delivery vehicles and drones primarily focus on single-vehicle / single-drone or single-vehicle / multi-drone models, with limited research on multi-vehicle / multi-drone models, which are undoubtedly more suitable for future practical applications. Furthermore, most existing algorithms do not consider scenarios where drones can continuously serve multiple user points and land on different delivery vehicles. Therefore, a path planning method that can coordinate multiple delivery vehicles and drones will play an invaluable role. However, the multi-vehicle / multi-drone path planning problem in logistics delivery presents challenges such as large solution scale, numerous constraints, and exponentially increasing difficulty. In real-world scenarios, the entire logistics delivery process needs to be completed in the shortest possible time. Simultaneously, multiple constraints such as drone range and drone retrieval by delivery vehicles must be considered, making this a combinatorial optimization problem.

[0004] In summary, current issues in dispatching transport vehicles and drones mainly target single-vehicle single-drone models and single-vehicle multi-drone models, which have limited delivery efficiency. Furthermore, they do not consider that drones can continuously serve multiple user points and land on different transport vehicles, resulting in high delivery costs and low delivery efficiency. Summary of the Invention

[0005] Based on the above analysis, the embodiments of the present invention aim to provide a large-scale air-ground cooperative path planning method based on a hybrid evolutionary algorithm to solve the problems of high delivery costs and low delivery efficiency in the prior art.

[0006] The objective of this invention is mainly achieved through the following technical solutions:

[0007] This invention provides a large-scale air-ground cooperative path planning method based on a hybrid evolutionary algorithm, comprising the following steps:

[0008] Obtain actual delivery information, including warehouse location information, location information of several users, number of transport vehicles, number of drones, and maximum flight time of a single drone;

[0009] Based on actual delivery information, obtain a description of the problem scenario;

[0010] Based on the aforementioned description information, a multi-transport vehicle and multi-UAV logistics delivery path planning model is constructed, including a directed graph, objective function, constraints, and path coding.

[0011] Based on the hybrid evolutionary algorithm, the multi-vehicle and multi-drone logistics delivery path planning model is solved to obtain the path planning schemes for the vehicle and drone.

[0012] Furthermore, the descriptive information includes functional descriptions of the transport vehicle and the drone, as well as a description of the drone's maximum flight time; obtaining the descriptive information of the problem scenario based on actual delivery information includes:

[0013] Based on the functions of the transport vehicle and the drone, functional description information of the transport vehicle and the drone is obtained respectively; wherein, the functions of the transport vehicle include being responsible for delivery, charging the drone, providing goods to the drone, and being allowed to leave and return to the warehouse only once; the functions of the drone include being responsible for delivery, taking off and landing on at least one of the different transport vehicles or the warehouse, and being able to continuously deliver to at least one of multiple users in different locations after taking off once.

[0014] Descriptive information about the drone's maximum flight time is obtained based on the drone's battery capacity.

[0015] Furthermore, the directed graph and the objective function include:

[0016] The directed graph G =<V,E> V represents a node, and E represents an arc;

[0017] Node V consists of one warehouse point and multiple user points; among which,

[0018] Multiple user points can be represented by a set of user points, denoted as C0 = {1, 2, 3, 4, ..., c};

[0019] The warehouse point is represented by the starting point {0} and the ending point {c+1};

[0020] Define the set of nodes containing the starting point as the first node set, denoted as C1 = C0 ∪ {0}; define the set of nodes containing the ending point as the second node set, denoted as C2 = C0 ∪ {c+1};

[0021] The objective function is to minimize the total delivery time, and the expression for the objective function f is:

[0022]

[0023] in, Let k be the time it takes for the kth transport vehicle to return to the warehouse. Let K be the time when the d-th drone returns to the warehouse, K be the set of transport vehicles, and D be the set of drones.

[0024] Furthermore, the constraints include: service path constraints, drone service constraints, transport vehicle service constraints, drone-transport vehicle interaction constraints, and sub-loop prevention constraints; wherein,

[0025] The service path restriction conditions are expressed as follows:

[0026]

[0027]

[0028]

[0029]

[0030] Where n represents the number of drones; x ijk Let y be a 0-1 variable. If its value is 1, it means that the k-th transport vehicle independently passes through arc ij; otherwise, it is 0. ijd Let K be a 0-1 variable. If its value is 1, it means that the d-th UAV has passed through arc ij; otherwise, it is 0. K is the set of transport vehicles, and D is the set of UAVs.

[0031] The restrictions on drone services are expressed as follows:

[0032]

[0033]

[0034] in, M represents the time required for the drone to travel from node i to node j, where M is a very large number without physical meaning, and L is the maximum flight time of the drone.

[0035] The service restrictions for the transport vehicles are expressed as follows:

[0036]

[0037] in, This represents the time required for the transport vehicle to travel from node j to node j;

[0038] The interaction constraints of the drone transport vehicle are expressed as follows:

[0039]

[0040]

[0041]

[0042]

[0043] The sub-loop prevention restriction condition is expressed as follows:

[0044]

[0045]

[0046] Among them, u i and p ij All are decision variables of the model; u i The order in which node i is visited is used to eliminate sub-loops, while p... ij It is a 0-1 variable. If its value is 1, it means that in the transport vehicle path, node i is visited before node j.

[0047] Furthermore, the path coding includes a total path code, a transport vehicle path code, and a drone path code; wherein,

[0048] There are a total of m transport vehicles and n drones;

[0049] In the overall path coding, the starting point is 0, the ending point is c+1, and the remaining points are user points;

[0050] In the vehicle route coding, there are a total of m groups of segments with 0 and c+1 as the two ends, representing the routes of m vehicles respectively. The value between 0 and c+1 in each group is the user point.

[0051] In UAV path coding, the UAV takeoff point, the user point served, and the landing point are all relative positions in the overall path code. -1 is the first-level distinguisher used to distinguish different UAVs; -2 is the second-level distinguisher used to distinguish different route codes of the same UAV.

[0052] Furthermore, based on a hybrid evolutionary algorithm, the multi-vehicle, multi-drone logistics delivery route planning model is solved to obtain a solution, including:

[0053] The multi-transport vehicle and multi-drone logistics delivery route planning model is initialized;

[0054] Calculate all current feasible solutions of the model, calculate the individual fitness of all current feasible solutions, and record the current optimal solution;

[0055] Select feasible solutions to obtain selected individuals; then perform sequential crossover on the selected individuals to obtain the crossed individuals, which are the current solutions;

[0056] Mutate the crossover individuals to obtain and determine whether to accept the new solution. If the new solution is accepted, first update the optimal solution and the current solution, and then determine whether the optimal solution at this temperature has reached stability. If the new solution is not accepted, directly determine whether the optimal solution at this temperature has reached stability. If it is stable, enter the outer loop; otherwise, continue the inner loop.

[0057] Enter the outer loop. If the current temperature reaches the termination temperature, the mutation ends, the solution space is updated, and the fitness of the new individuals is calculated again; otherwise, continue the inner loop.

[0058] The system obtains a new individual fitness and determines whether the maximum number of iterations has been reached. If it has, a solution is generated; otherwise, the process re-enters selection, sequential crossover, and mutation.

[0059] Furthermore, all feasible solutions of the current generation of the model are computed, and the individual fitness of all feasible solutions of the current generation is calculated. The current optimal solution is recorded, including:

[0060] Determine whether an individual satisfies the constraints; if it does, it is a feasible solution.

[0061] We obtain all feasible solutions in the current era, and then calculate the fitness F of all feasible solutions in the current era. The fitness calculation formula is as follows:

[0062] F = e -af

[0063] Where a is a constant representing the problem size; f is the objective function;

[0064] Record and save the solution with the highest fitness, i.e., the optimal solution s. best .

[0065] Further, feasible solutions are selected to obtain selected individuals, including:

[0066] Using a roulette wheel selection operator, feasible solutions are selected in the following way to obtain the selected individual:

[0067] Based on the fitness of all feasible solutions, the selection probability and cumulative probability of each feasible solution are calculated; the fitness of the i-th feasible solution is F. i Then the probability of choosing the i-th feasible solution is The cumulative probability of the i-th feasible solution is Where popsize is the population size;

[0068] Generate popsize random numbers r between 0 and 1; if r < q1, then select the first feasible solution; otherwise, select the i-th feasible solution such that q i-1 <r ≤ q i ; Take the selected feasible solution as the selected individual.

[0069] Furthermore, perform order crossover on the selected individual to obtain a crossover individual, i.e., the current solution, including:

[0070] Use the order crossover operator to perform crossover on the selected individual in the following way to obtain a crossover individual, i.e., the current solution s now :

[0071] According to the crossover probability coefficient P c Select parent individuals, and in the total path encodings of the two selected individuals, arbitrarily select a continuous segment for exchange, and the remaining genes are arranged in the original order of the individual to obtain a crossover individual, i.e., the current solution s now .

[0072] Furthermore, perform mutation on the crossover individual to obtain and determine whether to accept the new solution, including:

[0073] According to the mutation probability coefficient P m Judge the crossover individual, i.e., the current solution s now Whether to mutate; during the judgment process, generate a random number r' between 0 and 1. If r' < P m , then perform simulated annealing mutation on the current solution in the following way:

[0074] First, initialize the initial temperature t0 and the number of outer loop iterations k'. Let t0 = T, where T is the set initial temperature; k' = 0;

[0075] Secondly, use the Swap operator, DroneInsertion operator and TruckInsertion operator to perturb the current solution s now using local search, and record the feasible optimal solution s1 of the Swap operator, the feasible optimal solution s2 of the DroneInsertion operator and the feasible optimal solution s3 of the TruckInsertion operator respectively;

[0076] Then, select the solution with the largest fitness among s1, s2, and s3 as the new solution s new ; Among them,

[0077] The acceptance criterion for the new solution is based on the Metropolis criterion, and the acceptance probability P of the new solution is calculated in the following way tk’ :

[0078]

[0079] Among them, t k’ The current temperature is calculated using the following formula: k' represents the number of iterations in the outer loop;

[0080] Finally, generate a random number r” between 0 and 1. If r” is less than the acceptance probability, accept the new solution and update the current solution s. now and the optimal solution s best , that is, let s now =s new s best =s new Otherwise, new solutions will not be accepted.

[0081] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0082] 1. This invention solves the path planning problem of multiple transport vehicles and multiple drones in logistics distribution by constructing a model and designing an algorithm, thereby reducing distribution costs and improving distribution efficiency.

[0083] 2. This invention enables multiple transport vehicles and multiple drones to be deployed in logistics delivery. The drones can continuously serve multiple user points and can land on different transport vehicles, which is beneficial to the flexibility of delivery plans in actual logistics.

[0084] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description

[0085] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0086] Figure 1 This is a flowchart of the method of the present invention;

[0087] Figure 2 This is a flowchart of the algorithm of the present invention;

[0088] Figure 3 This is a flowchart of the mutation stage in this invention;

[0089] Figure 4 This is a schematic diagram of the encoding of the present invention;

[0090] Figure 5 This is a schematic diagram illustrating the local search using the Swap operator in this invention;

[0091] Figure 6 This is a schematic diagram illustrating the generation of a new unmanned path using the DroneInsertion operator in this invention.

[0092] Figure 7 This is a schematic diagram illustrating the insertion of an existing UAV path using the DroneInsertion operator in this invention.

[0093] Figure 8 This is a schematic diagram illustrating the local search performed using the TruckInsertion operator in this invention. Detailed Implementation

[0094] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0095] Example 1

[0096] A specific embodiment of the present invention discloses a large-scale air-ground cooperative path planning method based on a hybrid evolutionary algorithm, comprising the following steps:

[0097] S1. Obtain actual delivery information, including warehouse location information, location information of several users, number of transport vehicles, number of drones, and maximum flight time of a single drone.

[0098] S2. Obtain a description of the problem scenario based on the actual delivery information.

[0099] S3. Based on the described information, construct a multi-transport vehicle and multi-UAV logistics delivery path planning model, including a directed graph, objective function, constraints, and path coding.

[0100] S4. Based on the hybrid evolutionary algorithm, solve the multi-transport vehicle and multi-drone logistics delivery path planning model to obtain the path planning schemes for the transport vehicles and drones.

[0101] Compared with existing technologies, the large-scale air-ground cooperative path planning method based on hybrid evolutionary algorithm provided in this embodiment realizes path planning for multiple transport vehicles and multiple drones in logistics distribution, reducing delivery costs and improving delivery efficiency; and it takes into account that drones can continuously serve multiple user points and can land on different transport vehicles, which is beneficial to the flexibility of delivery plans in actual logistics.

[0102] Example 2

[0103] Based on Example 1, step S2 can be further refined as follows:

[0104] S21. Obtain the mission description information of the transport vehicle and the drone;

[0105] Specifically, in actual delivery, there are multiple transport vehicles and multiple drones to provide delivery services to users in different locations. Among them, the transport vehicles are responsible for delivery, charging the drones, and supplying goods to the drones, while the drones are only responsible for delivery. The drones can take off and land on the transport vehicles or in the warehouse, and the transport vehicles and drones can wait for each other.

[0106] S22. Obtain the interaction relationship and description information between the transport vehicle and the drone;

[0107] Specifically, drones can take off and land on different transport vehicles, and once a drone takes off, it can continuously deliver to multiple users in different locations.

[0108] Each transport vehicle is only allowed to leave and return to the warehouse once;

[0109] After drones and transport vehicles arrive at the user's location, the time spent serving the user and the take-off and landing time of the drones are ignored.

[0110] S23. Obtain energy limitation description information for the transport vehicle and the drone;

[0111] Specifically, the transport vehicle has no energy limitations; the drone has a limited battery capacity, meaning the drone must land within its maximum flight time.

[0112] Preferably, step S3 can be further refined into the following steps:

[0113] S31. Construct a directed graph for the multi-vehicle, multi-drone logistics delivery route planning model, wherein the directed graph G =<V,E> V represents a node, and E represents an arc;

[0114] Node V consists of one warehouse point and multiple user points; among which,

[0115] Multiple user points can be represented by a set of user points, denoted as C0 = [1, 2, 3, 4, ..., c};

[0116] The warehouse point is represented by the starting point {0} and the ending point {c+1};

[0117] Define the set of nodes containing the starting point as the first node set, denoted as C1 = C0 ∪ {0}; define the set of nodes containing the ending point as the second node set, denoted as C2 = C0 ∪ {c+1}.

[0118] S32. Construct the objective function for the multi-vehicle, multi-drone logistics delivery route planning model. The objective function is to minimize the total delivery time, that is, to ensure that the last vehicle or the last drone returns to the warehouse as early as possible. The expression for the objective function f is:

[0119]

[0120] in, Let k be the time it takes for the kth transport vehicle to return to the warehouse. Let K be the time when the d-th drone returns to the warehouse, K be the set of transport vehicles, and D be the set of drones.

[0121] S33. The constraints for the multi-vehicle, multi-drone logistics delivery route planning model are constructed; these constraints are mainly divided into the following categories: service path constraints, drone service constraints, vehicle service constraints, drone-vehicle interaction constraints, and sub-loop prevention constraints; among which...

[0122] The service path restriction conditions are expressed as follows:

[0123]

[0124]

[0125]

[0126]

[0127] Where n represents the number of drones; x ijk Let y be a 0-1 variable. If its value is 1, it means that the k-th transport vehicle independently passes through arc ij; otherwise, it is 0. ijd The variable is 0-1. If its value is 1, it means that the d-th drone has passed through arc ij; otherwise, it is 0. K is the set of transport vehicles, and D is the set of drones. The first four equations are used to constrain all transport vehicles and drones to leave the warehouse and return to the warehouse. The last two equations are used to constrain the in-degree and out-degree of the user point to be 1.

[0128] The restrictions on drone services are expressed as follows:

[0129]

[0130]

[0131] in, Let M represent the time required for the drone to travel from node i to node j, M be a large number without physical meaning, and L be the maximum flight time of the drone. The first inequality ensures that the drone's flight time does not exceed its maximum flight time. The second inequality constrains the time it takes for the drone to reach the user point.

[0132] The service restrictions for the transport vehicles are expressed as follows:

[0133]

[0134] in, This represents the time required for the transport vehicle to travel from node i to node j; this constraint also restricts the time it takes for the transport vehicle to reach the user point.

[0135] The interaction constraints of the drone transport vehicle are expressed as follows:

[0136]

[0137]

[0138]

[0139]

[0140] The first two inequalities constrain the transport vehicle and the drone to separate synchronously at node i, meaning their departure times must be the same. Note that the d-th drone takes off from the k-th transport vehicle. Similarly, the last two inequalities constrain the transport vehicle and the drone to arrive at node j at the same time. Note that the d-th drone does not necessarily take off from the k-th transport vehicle; that is, if the transport vehicle (or drone) arrives at node j first, it must wait for the drone (or transport vehicle) to arrive.

[0141] The sub-loop prevention restriction condition is expressed as follows:

[0142]

[0143]

[0144] Among them, u i and p ij All are decision variables of the model; u i The order in which node i is visited is used to eliminate sub-loops, while p... ij It is a 0-1 variable. If its value is 1, it means that in the transport vehicle path, node i is visited before node j.

[0145] S34. Construct the overall path code, transport vehicle path code, and drone path code for the multi-transport vehicle and multi-drone logistics delivery path planning model;

[0146] Specifically, suppose there are a total of m transport vehicles and n drones;

[0147] In the overall path coding, the starting point is 0, the ending point is c+1, and the remaining points are user points;

[0148] In the vehicle route coding, there are a total of m groups of segments with 0 and c+1 as the two ends, representing the routes of m vehicles respectively. The value between 0 and c+1 in each group represents the relative position of the user point on the total route.

[0149] The model proposed in this invention allows drone take-off and landing nodes to be on different transport vehicle routes, and a drone can serve multiple customers after a single take-off. To address these constraints, in the drone path coding, the drone take-off point, the customer point served, and the landing point are all relative positions within the overall path code, while -1 is a primary distinguisher used to differentiate between different drones; -2 is a secondary distinguisher used to differentiate between different trip codes for the same drone. For example... Figure 4 As shown, the drone takes off from user point 12 in the path of transport vehicle 2, serves user point 11, and lands at user point 9. The drone route is coded as: [12,10,7].

[0150] Preferably, step S4 can be further refined into the following steps:

[0151] S41. Initialize the basic parameters of the algorithm based on the hybrid genetic algorithm.

[0152] Specifically, the basic parameters of the algorithm include population size popsize, number of user points c, number of transport vehicles m, number of drones n, initial temperature t0, and cooling method. Termination temperature t F Maximum number of inner loops N, drone flight time limit L, crossover probability coefficient P c Coefficient of variation P m The maximum number of iterations in the termination condition.

[0153] S42. Initialize individual parameters based on a hybrid genetic algorithm.

[0154] Specifically, the initialization of individual parameters, i.e., the random generation of solutions to the mTSP problem, is achieved in the following way:

[0155] A non-repeating integer c is randomly generated between the start and end points in the total path encoding. i ∈[1,c], and c i Iterate through numbers 1 to c;

[0156] In the vehicle route coding, a unique integer k is randomly generated for each group of vehicle routes. i ∈[1,c] represents the relative position of the user point in the overall path encoding. At this point, it is necessary to ensure that each user point is assigned to a specific transport vehicle loop, so that each user point is visited exactly once;

[0157] The drone path encoding generates n+1 sets of primary identifiers -1 to distinguish different drones.

[0158] S43. Calculate all current feasible solutions to the model, and calculate the individual fitness of all current feasible solutions, recording the current optimal solution s.best .

[0159] Specifically, first determine whether an individual meets the constraint conditions. If it meets the constraint conditions, it is a feasible solution.

[0160] Obtain all the feasible solutions of the current generation, and then calculate the fitness F of all the feasible solutions of the current generation to obtain the fitness of all the feasible solutions of the current generation. The calculation formula for the fitness is:

[0161] F = e -af

[0162] where a is a problem scale constant, which is appropriately selected according to the problem scale; f is the objective function and satisfies:

[0163]

[0164] Record and save the solution with the maximum current fitness, that is, the optimal solution s best .

[0165] S44. Use the roulette wheel selection operator to select the feasible solutions to obtain the selected individuals.

[0166] Specifically, based on the fitness of all the feasible solutions of the current generation, calculate the selection probability and cumulative probability of each feasible solution; exemplarily, take the i-th feasible solution as an example. Its fitness is Fi, then its selection probability is Its cumulative probability is Then generate popsize random numbers r between 0 and 1. If r < q1, select the first feasible solution; otherwise, select the i-th feasible solution such that q i-1 < r ≤ q i ; Use the selected feasible solution as the selected individual.

[0167] S45. Use the order crossover operator to perform order crossover on the selected individuals to obtain the crossover individuals, that is, the current solution s now .

[0168] Specifically, according to the crossover probability coefficient P c Select the parent individuals. In the total path encoding of the two selected individuals, randomly select a continuous segment for exchange, and the remaining genes are arranged in the original order of the individual to obtain the crossover individuals, that is, the current solution s now ; Use the total path encoding to make the crossover operation faster and more convenient, and use the transporter path encoding to select the relative position of the user points in the total path encoding, so that after the total path encoding crossover, the transporter path encoding can represent a new transporter route.

[0169] S46. Use the mutation operator to mutate the crossover individual s now and calculate the individual fitness again.

[0170] Specifically, it needs to be based on the coefficient of variation P. m Determine the overlapping individuals, i.e., the current solution s. now Whether it has mutated. During the judgment process, a random number r' between 0 and 1 is generated. If r' <P m Then, simulated annealing mutation is applied to the current solution.

[0171] The specific steps for applying simulated annealing mutation to the current solution are as follows: Figure 3 As shown.

[0172] First, initialize the initial temperature t0 and the number of outer loops k', letting t0 = T, where T is the set initial temperature; k' = 0; and select the current solution s based on the crossover results. now and the optimal solution s best The optimal solution s best This is the current value.

[0173] Secondly, for the current solution s now Perturbation is performed using a local search;

[0174] Specifically, the local search employs three operators: Swap, DroneInsertion, and TruckInsertion; among them,

[0175] Using the Swap operator, two user points are randomly selected in the overall path encoding and their positions are swapped to generate a new neighborhood solution. The feasible optimal solution s1 is recorded. For example... Figure 5 The image shows the result of swapping the positions of nodes 2 and 3 in the encoding.

[0176] Using the DroneInsertion operator, user points previously served by transport vehicles are now served by drones. This is achieved by directly inserting a point from the transport vehicle's path code into the drone's path code. If both sides of the inserted point are distinguishers, a new drone path will be generated. The drone's takeoff and landing points are then selected to the left and right of this inserted point. A secondary distinguisher, -2, is then generated to the left of the takeoff point and to the right of the landing point. Figure 6 As shown; it should be noted that due to the positional constraints of the takeoff and landing points, to avoid infeasible solutions, the time when the drone arrives at the landing point must be later than the time when it takes off; if both sides after insertion are non-negative values ​​different from the value of the point, it means that the user point is inserted in the middle of an existing drone path; if one side of the value of the point after insertion is a node with the same value as the point, then the node needs to be changed to a takeoff point or a landing point, such as Figure 7 As shown, node 6 is removed from the transport vehicle path and inserted into the drone path; record the feasible optimal solution s2;

[0177] Using the TruckInsertion operator, a user point in one transport vehicle loop is inserted into another transport vehicle loop by inserting a point between 0 and c+1 into another set of 0 to c+1; the feasible optimal solution s3 is recorded. Figure 8 It describes how node 6 moves from path 1 to path 2.

[0178] Then, select the solution with the highest fitness among s1, s2, and s3 as the new solution s. new ;

[0179] Specifically, the acceptance criterion for the new solution is based on the Metropolis criterion, and the acceptance probability P of the new solution is calculated as follows: tk’ :

[0180]

[0181] Among them, t k’ The current temperature is calculated using the following formula: k' represents the number of iterations in the outer loop;

[0182] Generate a random number r between 0 and 1. If r is less than the acceptance probability P... tk’ If so, accept the new solution and update the current solution s. now and the optimal solution s best , that is, let s now =s new s best =s new Otherwise, no new solution will be accepted. It should be noted that since the inner loop does not occur more than once, there will generally be no solution update after one inner loop. If the optimal solution has not been improved after multiple inner loops, it means that the algorithm has converged. At this time, you can just wait for the algorithm to reach the termination condition and end the run.

[0183] Finally, by determining whether the temperature is stable, a decision is made on whether to enter the internal or external circulation.

[0184] Specifically, if the number of inner loop iterations, i.e. the number of local searches, reaches N, it is determined that the solution at that temperature has reached stability, and cooling measures are taken, setting k' = k' + 1, and entering the next outer loop; otherwise, the inner loop continues.

[0185] Entering the external circulation, if the current temperature t is satisfied... k Reaching the termination temperature t F If the mutation ends, the solution space is updated, and the process returns to step S31 to calculate the new individual fitness.

[0186] S47. After obtaining new individual fitness, determine whether the termination condition is met. If the maximum number of iterations has been reached, generate a solution; otherwise, re-enter the selection phase, sequential crossover phase, and mutation phase. Figure 2 As shown.

[0187] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0188] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A large-scale air-ground cooperative path planning method based on a hybrid evolutionary algorithm, characterized in that, Includes the following steps: Obtain actual delivery information, including warehouse location information, location information of several users, number of transport vehicles, number of drones, and maximum flight time of a single drone; Based on actual delivery information, obtain a description of the problem scenario; Based on the aforementioned description information, a multi-transport vehicle and multi-UAV logistics delivery path planning model is constructed, including a directed graph, objective function, constraints, and path coding. Based on the hybrid evolutionary algorithm, the multi-vehicle and multi-drone logistics delivery path planning model is solved to obtain the path planning schemes for the vehicle and drone. The constraints include: service path constraints, UAV service constraints, transport vehicle service constraints, UAV-transport vehicle interaction constraints, and sub-loop prevention constraints; among which... The service path restriction conditions are expressed as follows: in, n Indicates the number of drones; It is a 0-1 variable; if its value is 1, it represents the first... k The transport vehicles passed through the arc independently. Otherwise, it is 0; It is a 0-1 variable; if its value is 1, it represents the first... d A drone passed through the arc Otherwise, it is 0; K For transport vehicles, D For drones; The restrictions on drone services are expressed as follows: in, Indicates that the drone is from the node To the node The time required M It is a very large number that has no physical meaning. L This refers to the maximum flight time of the drone; The service restrictions for the transport vehicles are expressed as follows: in, Indicates that the transport vehicle departs from the node To the node The time required; The interaction constraints of the drone transport vehicle are expressed as follows: The sub-loop prevention restriction condition is expressed as follows: in, All of these are decision variables for the model; For nodes The order in which they are visited is used to eliminate sub-loops, and This is a 0-1 variable; if its value is 1, it indicates that the node in the transport vehicle's path... Comparison of nodes j Visit first; The path coding includes a total path code, a transport vehicle path code, and a drone path code; wherein, Total m A transport vehicle, n A drone; In the overall path coding, the starting point is 0, and the ending point is... The remaining points are user points; In the transport vehicle route coding, there are a total of m Groups with 0 and The segments at both ends represent, respectively m The route of the transport vehicles, where each group of 0 and The values ​​between these ranges represent user points; In UAV path coding, the UAV takeoff point, the user point served, and the landing point are all relative positions in the overall path code. -1 is the first-level distinguisher used to distinguish different UAVs; -2 is the second-level distinguisher used to distinguish different route codes of the same UAV.

2. The large-scale air-ground cooperative path based on the hybrid evolutionary algorithm as described in claim 1 The planning method is characterized by, The descriptive information includes functional descriptions of the transport vehicle and the drone, as well as a description of the drone's maximum flight time. The step of obtaining descriptive information about the problem scenario based on actual delivery information includes: Based on the functions of the transport vehicle and the drone, functional description information of the transport vehicle and the drone is obtained respectively; wherein, the functions of the transport vehicle include being responsible for delivery, charging the drone, providing goods to the drone, and being allowed to leave and return to the warehouse only once; the functions of the drone include being responsible for delivery, taking off and landing on at least one of the different transport vehicles or the warehouse, and being able to continuously deliver to at least one of multiple users in different locations after taking off once. Descriptive information about the drone's maximum flight time is obtained based on the drone's battery capacity.

3. The large-scale air-ground cooperative path planning method based on a hybrid evolutionary algorithm according to claim 2, characterized in that, The directed graph and objective function include: The directed graph , V Representative node, E Represents an arc; node V It includes one warehouse point and multiple user points; among which, Multiple user points can be represented by a set of user points, denoted as . ; From the starting point and the end point To represent warehouse locations; Define the set of nodes containing the starting point as the first node set, denoted as . Define the set of nodes containing the endpoint as the second node set, denoted as . ; The objective function is to minimize the total delivery time. The expression is: ， in, For the first k The time it takes for the transport trucks to return to the warehouse. For the first d The time it takes for the drone to return to the warehouse. K For transport vehicles, D For drones.

4. The large-scale air-ground cooperative path planning method based on a hybrid evolutionary algorithm according to any one of claims 1 to 3, characterized in that, Based on a hybrid evolutionary algorithm, a multi-vehicle, multi-drone logistics delivery route planning model is solved to obtain a solution, including: The multi-transport vehicle and multi-drone logistics delivery route planning model is initialized; Calculate all current feasible solutions of the model, calculate the individual fitness of all current feasible solutions, and record the current optimal solution; Select feasible solutions to obtain selected individuals; then perform sequential crossover on the selected individuals to obtain the crossed individuals, which are the current solutions; Mutate the crossover individuals to obtain and determine whether to accept the new solution. If the new solution is accepted, first update the optimal solution and the current solution, and then determine whether the optimal solution at this temperature has reached stability. If the new solution is not accepted, directly determine whether the optimal solution at this temperature has reached stability. If it is stable, enter the outer loop; otherwise, continue the inner loop. Enter the outer loop. If the current temperature reaches the termination temperature, the mutation ends, the solution space is updated, and the fitness of the new individuals is calculated again; otherwise, continue the inner loop. The system obtains a new individual fitness and determines whether the maximum number of iterations has been reached. If it has, a solution is generated; otherwise, the process re-enters selection, sequential crossover, and mutation.

5. The large-scale air-ground cooperative path planning method based on a hybrid evolutionary algorithm according to claim 4, characterized in that, Calculate all current feasible solutions to the model, and compute the individual fitness of all current feasible solutions. Record the current optimal solution, including: Determine whether an individual satisfies the constraints; if it does, it is a feasible solution. Obtain all feasible solutions in the current period, and then apply fitness to all feasible solutions in the current period. F The fitness of all feasible solutions in the present era is obtained through calculation. The formula for calculating fitness is: in, a The problem size is a constant; f The objective function is... Record and save the solution with the highest current fitness, i.e., the optimal solution. .

6. The large-scale air-ground cooperative path planning method based on a hybrid evolutionary algorithm according to claim 5, characterized in that, The feasible solutions are selected to obtain the selected individuals, including: Using a roulette wheel selection operator, feasible solutions are selected in the following way to obtain the selected individual: Based on the fitness of all feasible solutions in the present time, calculate the selection probability and cumulative probability of each feasible solution; i The fitness of each feasible solution is Then the first i The probability of choosing one feasible solution is: , No. i The cumulative probability of each feasible solution is ;in, popsize Population size; generate popsize indivual Random numbers between ;like If the first feasible solution is selected, then the second feasible solution is selected; otherwise, the third feasible solution is selected. i There are 1 feasible solution, such that Select the feasible solutions as the selected individuals.

7. The large-scale air-ground cooperative path planning method based on a hybrid evolutionary algorithm according to claim 6, characterized in that, Perform sequential crossover on the selected individuals to obtain the crossed individuals, i.e., the current solution, including: Using the sequential crossover operator, the selected individuals are crossed in the following way to obtain the crossed individuals, which is the current solution. : Based on the crossover probability coefficient Select a parent individual, and from the total path codes of the two selected individuals, arbitrarily choose a continuous segment to swap. The remaining genes are then arranged sequentially according to their original order within the selected individual, resulting in the crossover individual, which is the current solution. .

8. The large-scale air-ground cooperative path planning method based on a hybrid evolutionary algorithm according to claim 7, characterized in that, The process involves mutating the crossover individuals to obtain and determine whether to accept the new solution, including: Based on the coefficient of variation Determine the overlapping individuals, i.e., the current solution. Whether it has mutated; during the judgment process, generate a random number between 0 and 1. ,like < Then, simulated annealing mutation is applied to the current solution in the following way: First, regarding the initial temperature t 0 and the number of outer loops Perform initialization, let t 0 = T, where T is the initial temperature set. = 0; Secondly, the Swap, DroneInsertion, and TruckInsertion operators are used to analyze the current solution. Perturbations are performed using local search, and feasible optimal solutions for the Swap operator are recorded separately. Feasible optimal solution of the DroneInsertion operator Feasible optimal solution with TruckInsertion operator ; Then, select The solution with the highest fitness is used as the new solution. ;in, The acceptance criterion for the new solution is based on the Metropolis criterion, and the acceptance probability of the new solution is calculated as follows: : = in, The current temperature is calculated using the following formula: ; This represents the number of iterations in the outer loop. Finally, generate random numbers between 0 and 1. ,like If the probability is less than the acceptance probability, then accept the new solution and update the current solution. and optimal solution , that is to say = , = Otherwise, new solutions will not be accepted.

Images