Truck and unmanned aerial vehicle cooperative scheduling optimization method based on FTSPD

Abstract

The invention belongs to the field of logistics distribution scheduling, and particularly relates to a truck and unmanned aerial vehicle cooperative scheduling optimization method based on FTSPD, and the method comprises the steps: firstly obtaining warehouse point position information, customer point position information and customer demand, and taking the shortest total time for a truck and an unmanned aerial vehicle to cooperatively return to a warehouse to complete a task as a target; and constructing a truck and unmanned aerial vehicle cooperative scheduling model, and solving by adopting a hybrid artificial fish school genetic algorithm of a greedy strategy to obtain a scheduling scheme. According to the invention, through model innovation, strategy optimization and algorithm adaptation, the specific problems of inflexible cooperation mode, dispersed model research and insufficient attention to green logistics in the existing research are systematically solved, and the efficiency and practicability of cooperative distribution of the truck and the unmanned aerial vehicle are significantly improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of logistics distribution scheduling, and particularly relates to a truck and unmanned aerial vehicle collaborative scheduling optimization method based on FTSPD. BACKGROUND

[0002] With the rapid development of e-commerce and the increasing demand for timeliness of logistics distribution by consumers, the traditional logistics distribution mode is facing the challenge of high cost and low efficiency in the last mile delivery. Especially in remote areas, urban traffic congestion areas and emergency logistics, the traditional distribution method is difficult to effectively meet the actual logistics demand. Under this background, the unmanned aerial vehicle distribution technology is widely used in domestic and foreign distribution scenarios due to its high efficiency and flexibility, and has made effective progress.

[0003] However, the unmanned aerial vehicle still has certain limitations in endurance, load capacity and applicable scope, and is difficult to independently undertake large-scale and long-distance logistics distribution tasks. In order to overcome the defects of single distribution mode, a truck and unmanned aerial vehicle collaborative distribution mode is proposed, which combines the advantages of trucks and unmanned aerial vehicles, uses trucks for long-distance and large-volume goods transportation, and uses unmanned aerial vehicles for short-distance and rapid response end distribution, so as to realize efficient allocation and collaborative optimization of logistics resources. The truck as the mobile base station of the unmanned aerial vehicle can significantly expand the coverage of the unmanned aerial vehicle and improve the overall distribution efficiency. The schematic diagram of the truck carrying the unmanned aerial vehicle is shown in Figure 1 . The main feature is that the unmanned aerial vehicle takes off directly from the top of the truck to solve the problem of "last mile" delivery of large trucks in some remote areas.

[0004] The current truck and unmanned aerial vehicle collaborative distribution scheme mainly has the following problems: the collaborative mode is single, the flexibility is insufficient, which limits the full play of the flexibility of the unmanned aerial vehicle, the efficiency advantage of its straight flight is not fully utilized, and the overall collaboration degree is not high. Moreover, there is a lack of research on the distribution mode of new energy trucks. SUMMARY

[0005] The purpose of the present application is to provide a truck and unmanned aerial vehicle collaborative scheduling optimization method based on FTSPD, which is used to solve the problems of single collaborative mode, insufficient flexibility, scattered model research, lack of fusion and innovation, and insufficient attention to green logistics and new constraints in the prior art.

[0006] To achieve the above purpose, the present application adopts the following technical scheme:

[0007] A truck and unmanned aerial vehicle cooperative scheduling optimization method based on FTSPD, acquires warehouse point position information, customer point position information and its customer demand, takes the total time of the truck and unmanned aerial vehicle cooperative return to the warehouse to complete the task as the shortest as the target, constructs a truck and unmanned aerial vehicle cooperative scheduling model, adopts a hybrid artificial fish swarm genetic algorithm with a greedy strategy to solve, and obtains a scheduling scheme.

[0008] Further, the objective function is:

[0009] minimize T total

[0010] Wherein T total is the total time of the truck and unmanned aerial vehicle cooperative return to the warehouse to complete the task;

[0011] The mathematical expression of the time of the truck and unmanned aerial vehicle passing between nodes i and j is:

[0012]

[0013] The constraint part includes basic routing constraints, unmanned aerial vehicle action constraints, time continuity constraints and synchronization constraints. Further, when it is a new energy truck, the objective function is:

[0014] minimize T total ,E total

[0015] Wherein E total is the energy consumption;

[0016] Add energy constraints:

[0017] The new energy truck is equipped with a full battery at the beginning of the trip:

[0018] e0=E t

[0019] Track the battery power of the new energy truck after passing through the arc segment (i, j):

[0020]

[0021] The battery state of the new energy truck at the successor node after the charging node i:

[0022]

[0023] Set the maximum endurance distance of the unmanned aerial vehicle:

[0024]

[0025] The value of the polynomial is the maximum completion time of all trips:

[0026]

[0027] Total energy consumed by new energy truck and drone is calculated:

[0028]

[0029] Where n is the number of customer nodes that need to be served, R is the number of available charging stations, and r is the number of selected charging stations. Further, the basic routing constraints include:

[0030] Each customer can only be accessed by a truck or a drone once;

[0031]

[0032] The truck leaves the warehouse and drives to any one of the nodes in the customer and warehouse network;

[0033]

[0034] The truck returns to the warehouse from any one of the nodes in the customer and warehouse network;

[0035]

[0036] The truck must enter and leave a certain customer node to ensure the continuity of traffic in the truck route;

[0037] Sub-trip elimination constraint for truck trip connectivity;

[0038]

[0039] M is a large number, and only when all conditions are met, the constraint will be activated and strictly enforced;

[0040] Ensure the correct arrangement of the truck visiting customers;

[0041]

[0042] Further, the drone action constraints include:

[0043] Drone load weight constraint, the demand of the customers served by the drone is greater than the load weight constraint of the drone trip equal to 0:

[0044] Each drone can only be launched from each launch node at most once:

[0045]

[0046] Each drone can only be retrieved from each rendezvous node at most twice:

[0047]

[0048] When a drone is launched and retrieved at two different nodes, these nodes must be accessible by a truck:

[0049]

[0050] When the corresponding drone is launched from the depot and completes its task at node k, the truck must access the rendezvous node k e V from any node i e V: R

[0051]

[0052] Further, the time continuity constraints include:

[0053] The time at which the truck arrives at node k is greater than the time at which it arrives at the previous node i:

[0054]

[0055] The time at which the drone arrives at the rendezvous node k is greater than the time at which the drone arrives at the visited node j:

[0056]

[0057] The time at which the drone arrives at the rendezvous node k is greater than the time at which the drone arrives at the visited node j:

[0058]

[0059] Further, the synchronization constraints include:

[0060] The times at which the truck and the drone arrive at the launch node i are the same:

[0061]

[0062] The times at which the truck and the drone arrive at the rendezvous node k are the same:

[0063]

[0064] The truck cannot launch a drone if the drone is still serving a customer and has not yet been retrieved:

[0065]

[0066] The time at which all vehicles arrive at the garage node {0} is set to zero:

[0067]

[0068] According to the endurance constraint of the UAV, the maximum flight distance of the UAV is set:

[0069]

[0070] The value of the polynomial is set as the maximum completion time of all trips:

[0071]

[0072] Further, wherein n is the number of customer nodes requiring service, V is the set of all nodes in the network, V0 is the set of warehouse nodes and their virtual nodes, V D is the subset of customers that meet the UAV condition, V C is the set of customer nodes, V L is the set of start locations, V R is the set of end locations, V N is the set of all nodes except the warehouse nodes, A is the set of (i,j) arcs, where i,j∈V and i≠j, is the truck travel time between nodes i and j, is the UAV travel time between nodes i and j, is the truck travel distance from node i to j, is the UAV travel distance from node i to j, q i is the node demand, S t is the service time generated by the truck serving a customer node, S d is the service time generated by the UAV serving a customer node, Q d is the UAV weight limit, L is the UAV flight distance limit, v t is the truck travel speed, v d is the UAV travel speed, ρ L is the setup time required to launch a UAV, ρ R is the setup time required to retrieve a UAV, x ij is a binary variable: 1 if the truck moves from node i to node j; otherwise, 0, y ijk is a binary variable: 1 if the UAV performs visit (i,j,k); otherwise, 0, P ij is a binary variable: 1 if node i serves before j without having to be adjacent; otherwise, 0, is the time at which the truck arrives at node i, is the time at which the UAV arrives at node i, u i denotes the position of node i in the truck route.

[0073] Further, the steps of solving by using the hybrid artificial fish swarm genetic algorithm with a greedy strategy include:

[0074] (1) Initialization of parameters and population

[0075] When initializing the population:

[0076] The upper half of the array: the initial access path of the truck is generated by using a distance-based greedy strategy;

[0077] The lower half of the array: an array of the same length as the upper half is generated, and the beginning and end are fixed as 0, representing the warehouse.

[0078] (2) Fitness calculation and elite reservation selection

[0079] Each individual in the population is decoded to obtain the specific driving route and service scheme of the truck and the unmanned aerial vehicle, and the total time T of completing all distribution tasks is calculated total , the scheme is checked for violation of the endurance constraint and the load constraint of the unmanned aerial vehicle, if violated, a large penalty value is applied to the total time, and the fitness value Z after the penalty is obtained, the smaller the Z, the better the scheme;

[0080] (3) Update the population by following and foraging effect;

[0081] (4) Iteration and termination

[0082] The newly generated population is used as the initial population of the next generation. Steps (2) and (3) are repeatedly executed until the maximum number of iterations is reached, and after the algorithm terminates, the optimal solution found during the entire search process is output, that is, the truck and unmanned aerial vehicle cooperative scheduling scheme with the shortest total distribution time.

[0083] Further, when it is a new energy vehicle: add problem model parameters: electric vehicle energy consumption rate h t , unmanned aerial vehicle energy consumption rate h d , charging node electric vehicle charging rate R t , maximum energy capacity of electric vehicle, maximum energy capacity of unmanned aerial vehicle;

[0084] Total energy consumption Z2: calculate the total energy consumption of the new energy truck to complete the entire path, and add the penalty for the truck energy constraint. Add the "transform charging station" mutation operation: in the mutation process, traverse the upper half of the array, if the encountered node is a charging station, there is a certain probability to randomly replace the charging station with another optional charging station.

[0085] The present application has the following beneficial effects:

[0086] 1. A fusion collaborative model (FTSPD) and an optimized collaborative strategy are proposed: The present application innovatively proposes a fusion model (FTSPD) that combines the advantages of TSPD and STSPD models. The key lies in the design of a more flexible collaborative strategy, which clearly defines the order of service and launch / recovery of trucks and UAVs at the rendezvous node. This means that UAVs no longer have to return to the original launch truck, or launch and recovery can be carried out at different nodes, thereby significantly improving the flexibility of UAV use and the degree of collaboration of the entire system.

[0087] 2. Design of efficient hybrid heuristic algorithm: In order to effectively solve the proposed complex model, the present application designs a hybrid artificial fish swarm genetic algorithm combined with greedy strategy (for single-objective FTSPD) and a hybrid artificial fish swarm non-dominated sorting genetic algorithm (for multi-objective EC-FTSPD). These algorithms combine the advantages of various meta-heuristic algorithms, initialize the population through distance-based greedy strategy to ensure the quality of initial solution, accelerate convergence through the tail-chasing behavior of crossing with excellent individuals, and balance global and local search through mutation-based foraging behavior, so as to efficiently solve complex scheduling problems with high degree of collaboration.

[0088] 3. Extension to new energy truck and UAV collaborative scheduling: The present application further applies the proposed fusion model and collaborative strategy to the collaborative scheduling problem of new energy trucks and UAVs. For this purpose, a multi-objective optimization model (EC-FTSPD) considering truck energy consumption and charging is established, which optimizes the total distribution time and total energy consumption. Specific improvements are made in the algorithm, such as forcing charging stations not to be used as UAV service nodes in the coding, adding charging station transformation operations in the crossover and mutation, and using non-dominated sorting to handle multi-objective optimization, thereby providing a practical scheduling scheme and decision support for green logistics.

[0089] In summary, through model innovation (FTSPD), strategy optimization (flexible collaborative sequence) and algorithm adaptation (efficient hybrid heuristic algorithm), the present application systematically solves the specific problems of inflexible collaborative mode, scattered model research and lack of attention to green logistics in existing research, significantly improving the efficiency and practicality of truck and UAV collaborative distribution. BRIEF DESCRIPTION OF DRAWINGS

[0090] Figure 1 The figure is a schematic diagram of the STSPD model.

[0091] Figure 2 The figure is a case study of FTSPD and the corresponding coding scheme.

[0092] Figure 3 The figure is a sequential crossover process.

[0093] Figure 4Flow chart of hybrid artificial fish swarm genetic algorithm for greedy strategy.

[0094] Figure 5 Iteration chart for FTSPD model with 10 customer points.

[0095] Figure 6 Path chart for FTSPD model with 10 customer points.

[0096] Figure 7 Delivery schedule for FTSPD model.

[0097] Figure 8 Delivery schedule for three models.

[0098] Figure 9 Path chart for three models.

[0099] Figure 10 Comparison of optimal objective function values of four algorithms under different scale cases. DETAILED DESCRIPTION

[0100] The present application will be described in detail below with reference to the accompanying drawings and examples.

[0101] The single-truck and single-drone collaborative scheduling optimization method provided by the present application is implemented according to the following steps:

[0102] A truck and a drone collaboratively return to the warehouse to complete the task in the shortest total time, and a truck and drone collaborative scheduling model is constructed:

[0103] A truck carrying a drone needs to visit n sites, each site can only be visited once, and the truck serves as a mobile hub for the drone to provide goods replenishment, aiming to minimize the time to complete all delivery tasks, i.e., the time for all vehicles to return to the warehouse. The truck can only stop at the customer's location to launch the drone and retrieve it; the drone must return to the warehouse or meet the truck when launched from the truck; the drone can only serve a single customer request at a time; the drone can only serve customer points that do not exceed its weight limit; the drone replaces the battery at the truck after each task to ensure full power when launched; the needs of each customer can be met by a single visit by the truck or the drone; the drone's endurance, i.e., the drone's flight distance limit, cannot be exceeded in one trip; when the drone and the truck meet at the customer point, the drone's recovery task has higher priority than the service task; the time for the truck and the drone to serve a single customer point is fixed; the flight speed of the drone is greater than the driving speed of the truck; the same node is only allowed to launch the drone once.

[0104] First, the mathematical expression for the time taken by the truck and the drone to pass between nodes i and j is:

[0105]

[0106] Secondly, the objective function: the mathematical expression of minimizing the total time of truck and drone to return to the warehouse to complete the task is:

[0107] minimize T total

[0108] Finally, the constraint part includes basic routing constraints, drone action constraints, time continuity constraints, and synchronization constraints, and finally gives the value of all decision variables.

[0109] (1) Basic routing constraints:

[0110] It means that each customer can only be accessed by a truck or a drone once.

[0111]

[0112] Ensure that the truck leaves the warehouse and drives to any node in the customer and warehouse network.

[0113]

[0114] Ensure that the truck returns to the warehouse from any node in the customer and warehouse network.

[0115]

[0116] Guarantee the flow continuity in the truck route, that is, the truck must leave a customer node after entering it.

[0117]

[0118] Sub-trip elimination constraints to ensure truck trip connectivity.

[0119]

[0120] Ensure the correct arrangement of the truck visiting customers.

[0121]

[0122] (2) Drone action constraints:

[0123] Drone load weight constraint, the trip equal to 0 when the demand of the customer served by the drone is greater than the load weight constraint of the drone:

[0124]

[0125] Guarantee that each drone can only launch from each launch node at most once.

[0126]

[0127] Guarantee that each drone can retrieve at most twice from each rendezvous node, only possible if the first one adopts the 1 strategy and the second one adopts the 2 strategy.

[0128]

[0129] denotes the drone launching and retrieving at two different nodes. These nodes must be visited by the truck, although not necessarily in adjacent order.

[0130]

[0131] Guarantee that when its corresponding drone is launched from the depot and completes its task at node k, the truck must visit rendezvous node k∈V R from any node.

[0132]

[0133] (3) Temporal continuity constraints:

[0134] denotes that the truck arrives at node k later than it arrived at the previous node i.

[0135]

[0136] Guarantee that the drone arrives at a node j that meets the drone condition later than its corresponding truck (and drone itself) arrives at the launch node i.

[0137]

[0138] Guarantee that the drone arrives at a rendezvous node k later than the drone arrives at a visited node j.

[0139]

[0140] (4) Synchronization constraints:

[0141] Guarantee that the truck and the drone arrive at the launch node i at the same time.

[0142]

[0143] Guarantee that the truck and the drone arrive at the rendezvous node k at the same time.

[0144]

[0145]

[0146] Guarantee that the truck cannot launch the drone if the drone is still serving a customer and has not been recycled.

[0147]

[0148] Set the time of all vehicles to arrive at the depot node {0} to zero.

[0149]

[0150] denotes the endurance constraint of the drone, set the maximum flight distance of the drone.

[0151]

[0152] Set the value of the polynomial to the maximum completion time of all trips.

[0153]

[0154] where n is the number of customer nodes that need to be served, V is the set of all nodes in the network. V0 is the set of the depot node and its virtual nodes. V D is the subset of customers that meet the drone condition, V C is the set of customer nodes, V L is the set of start locations, V R is the set of end locations. V N is the set of all nodes except the depot node, A is the set of (i,j) arcs, where i,j∈V and i≠j, is the truck travel time between nodes i and j, is the drone travel time between nodes i and j. is the truck travel distance from node i to j, is the drone travel distance from node i to j, q i is the node demand, S t is the service time generated by the truck serving a customer node, S d is the service time generated by the drone serving a customer node, Q d is the drone weight limit, L is the drone flight distance limit, v t is the truck travel speed, v d is the drone travel speed, p L is the setup time required to launch a drone, p R is the setup time required to retrieve a drone, x ij is a binary variable: 1 if the new energy truck moves from node i to node j; otherwise, 0, y ijk is a binary variable: 1 if the drone performs visit (i,j,k); otherwise, 0, P ijBinary variable: 1 if node i is served before j (not necessarily adjacent), 0 otherwise, Time of truck arrival at node i, Time of drone arrival at node i, u i Denotes the position of node i in the truck route.

[0155] Simulate the objective function model:

[0156] To verify the effectiveness of the algorithm, 10 customer points are taken as an example. The location of 1 warehouse point, the location of 10 customer points and their demand are known. The example data is randomly generated. The coordinate information is randomly generated in a 100*100 map. The demand is randomly generated in 1 to 10. The Euclidean distance between customer points is taken as the distance matrix. The specific information is shown in Table 1. The specific steps are as follows:

[0157] Encoding: double-chromosome encoding scheme:

[0158] Generate the upper half array: generate the initial access path of the truck using the distance-based greedy strategy.

[0159] 1. Start from the warehouse point 0.

[0160] 2. In the unvisited customer points, select the nearest customer point to the current point to join the path.

[0161] 3. Repeat the above step until all customer points are visited, and finally return to the warehouse 0.

[0162] 4. This process is repeated G greedy times to get a better initial truck path.

[0163] Generate the lower half array: generate an array of the same length as the upper half array, with 0 at the beginning and end (representing the warehouse). The middle position is randomly filled with {0, 1, 2}, but the encoding must be feasible, for example, the left and right adjacent elements of the number 2 must be 0. This array defines the service mode of the drone. As Figure 2 shown.

[0164] Initialize the population: the population size is N, and each chromosome individual is initialized by the upper half array and the lower half array.

[0165] Fitness calculation: in order to handle the endurance constraint and payload constraint of the drone, the classic method of punishing infeasible solutions that violate constraints is used to separate the infeasible search space from the feasible search space.

[0166] Elitism strategy selection: The population is sorted by the fitness of different models, and the elite individuals selected from each model form an elite list (Elitism_list). These individuals are directly retained and participate in subsequent operations.

[0167] Based on the rear-end aggregation behavior of crossing with excellent individuals: The "sequential crossover" technique is adopted as shown in Figure 3 A rear-end aggregation behavior of crossing with excellent individuals is introduced. By retaining the excellent individuals in the parent population, the remaining individuals update their positions by crossing with these excellent individuals, forming a convergence trend in the population to gradually approach the global optimal solution.

[0168] Based on mutation-based foraging behavior: First, set the number of attempts to forage try_number, initialize the current fish's position as the best position, and calculate its fitness value. Loop to try foraging, perform small-range mutation operation with probability p, otherwise perform large-range mutation operation, get the new position of the current fish, and calculate its fitness value. If the fitness value of the new position is better than the current best fitness value, update the current fish's position and the current fish's best fitness value. Repeat the loop, and finally return the best position best_position found. The algorithm flow of the hybrid artificial fish swarm genetic algorithm combined with the greedy strategy to solve the FTSPD problem is shown in Figure 4 .

[0169] (1) Initialization of parameters and population

[0170] This step needs to set and generate all the basic data and initial solutions required for algorithm running. Specifically, it includes: a. Parameter setting:

[0171] Node set: warehouse node 0 and customer nodes {1, 2,..., n}.

[0172] Parameters of trucks and drones: truck speed v t , drone speed v d , truck service time S t , drone service time S d , launch drone setup time ρ L , retrieve drone setup time ρ R , maximum capacity of drone Q d , drone range L.b. Algorithm control parameters:

[0173] Population parameters: population size N.

[0174] Greedy initialization parameters: according to the number of customer points n customers Calculate the number of iterations of the variable greedy algorithm to balance the diversity and excellence of the initial population.

[0175] Specific operation parameters based on the foraging behavior of variation: try_number, the number of attempts to forage, p, the probability of performing small-scale variation (usually set as a function related to the current iteration number, for example, p = current iteration number / total iteration number). Termination condition: maximum iteration number.

[0176] c. Initialize the population:

[0177] Generate the upper half array: generate the initial access path of the truck using the distance-based greedy strategy.

[0178] 1. Start from warehouse point 0.

[0179] 2. In the unvisited customer points, select the customer point closest to the current point to join the path.

[0180] 3. Repeat the above step until all customer points are visited, and finally return to warehouse 0.

[0181] 4. This process is repeated G greedy times to get a better initial truck path.

[0182] Generate the lower half array: generate an array of the same length as the upper half array, with the first and last fixed as 0 (representing the warehouse). The middle positions are randomly filled with {0, 1, 2}, but need to ensure the feasibility of the code, for example, the left and right adjacent elements of number 2 must be 0. This array defines the service mode of the UAV.

[0183] (2) Fitness calculation and elite selection

[0184] a. Fitness calculation:

[0185] Decode each individual in the population (i.e., a complete upper and lower array code) to get the specific driving routes of the truck and UAV and the service scheme. Calculate the total time T total of the scheme to complete all delivery tasks. Check if the scheme violates the endurance constraint and the load constraint of the UAV. If it violates, apply a large penalty value to the total time to get the fitness value Z after the penalty. The smaller Z is, the better the scheme is.

[0186] b. Elite selection strategy:

[0187] Sort the population according to the fitness of different models, and select the elite individuals from each model to form the elite list (Elitism_list). These individuals are directly retained and participate in subsequent operations. This operation ensures that excellent individuals in the population will not be lost due to crossover and mutation, and also helps to accelerate convergence to the optimal solution.

[0188] (3) Artificial fish behavior operation: tail-chasing and foraging

[0189] This step simulates the aggregation and foraging behavior of fish population, and updates the non-elite individuals in the population through crossover and mutation operations.

[0190] a. Tail-chasing behavior:

[0191] For non-elite individuals, perform sequential crossover with the elite individuals retained in the previous step.

[0192] Randomly select two crossover points sp1 and sp2. Lower half: the offspring inherits the service mode gene in the [sp1, sp2] interval from the elite parent, and the rest comes from the current parent, with feasibility correction (ensuring that the left and right of 2 are 0). Upper half: the offspring inherits the customer node access order in the [sp1, sp2] interval from the elite parent. Then perform repair: first remove the customer nodes that already exist in the offspring from the current parent, and then fill the remaining customer nodes of the current parent into the empty positions of the offspring in turn. Finally, combine the newly generated upper and lower halves into a new offspring individual.

[0193] b. Foraging behavior:

[0194] Perform try_number attempts of mutation on the individual.

[0195] Each attempt, decide the mutation method with a probability p:

[0196] Small range mutation: randomly select two customer points close in distance, and exchange their access order and service mode in the sequence. This focuses on local fine search. Large range mutation: randomly select two arbitrary customer points for exchange. This focuses on global exploration and jumping out of local optimum. If the new position generated after mutation is better, update the current individual.

[0197] (4) Iteration and termination

[0198] Take the newly generated population as the initial population of the next generation. Repeat steps (2) and (3) until the maximum number of iterations is reached. After the algorithm terminates, output the optimal solution found during the entire search process, i.e. the shortest total delivery time of the truck and UAV cooperative scheduling scheme.

[0199] Model solution for new energy vehicles

[0200] (1) Initialize parameters and population

[0201] Increase problem model parameters: electric vehicle energy consumption rate h t , UAV energy consumption rate h d , electric vehicle charging rate R at charging nodes t , maximum energy capacity of electric vehicle, maximum energy capacity of UAV.

[0202] Initialize population:

[0203] Generate upper half:

[0204] 1. First, randomly select a certain number of charging stations from all available charging stations according to the set number of charging times.

[0205] 2. Then, generate an initial truck path containing only warehouse and customer points using the same distance-based greedy strategy as the truck solution.

[0206] 3. Finally, randomly insert the selected charging stations into the positions between the first and last warehouse points of the path to form the final upper half.

[0207] Generate lower half:

[0208] Add a mandatory rule: for the nodes corresponding to charging stations in the upper half, their values in the lower half must be set to "0" by force, indicating that charging stations can only be accessed by trucks and cannot be used as service nodes for drones.

[0209] (2) Multi-objective fitness calculation and non-dominated sorting selection

[0210] Transition from single-objective optimization to multi-objective optimization.

[0211] a. Multi-objective fitness calculation: decode each individual to obtain the collaborative scheduling scheme of trucks and drones.

[0212] Calculate the values of two objective functions simultaneously:

[0213] Total time Z1: the calculation method is the same as in Chapter 3, and a penalty is added for the endurance and load constraints of the drone.

[0214] Total energy consumption Z2: calculate the total energy consumption of the new energy truck to complete the entire path, and add a penalty for the energy constraints of the truck.

[0215] b. Elite preservation strategy (crowding-based non-dominated sorting):

[0216] Instead of using a single fitness ranking, the fast non-dominated sorting method of NSGA-II is adopted:

[0217] 1. According to the two objective values (Z1, Z2) of the individual, divide the entire population into multiple Pareto front layers. The first layer is the elite individual set that is not dominated by any other individual, the second layer follows, and so on.

[0218] 2. For individuals in the same Pareto layer, calculate their crowding distance to measure the distribution density of individuals in the target space, and prefer individuals with large crowding distance (individuals in sparse areas) to maintain the diversity of the population.

[0219] 3. According to the comprehensive comparison of the Pareto level (the higher the better) and the crowding degree (the larger the better), the elite individuals are selected and reserved to the next generation.

[0220] (3) Artificial fish behavior operation: tail chasing and foraging

[0221] Special processing for charging station nodes is added.

[0222] a. Tail chasing behavior:

[0223] The basic logic of the sequential crossover of non-elite individuals and elite individuals remains unchanged, and the key improvement is in the repair step after crossover: the upper half of the array: after completing the traditional sequential crossover and repair, it is necessary to check whether all customer points have been included in the offspring path. If it is found that some customer points are missing, these missing customer points will be inserted into the path at the additional charging station node position for service to ensure the completeness of the solution.

[0224] The lower half of the array: after completing the crossover of the service mode, in addition to ensuring that the left and right of the number "2" are "0", a second repair is also required, that is, to ensure that the value of all charging stations in the corresponding position of the lower half of the array in the upper half of the array is forcibly reset to "0".

[0225] b. Foraging behavior:

[0226] New "transform charging station" mutation operation: in the mutation process, the upper half of the array will be traversed, and if the node encountered is a charging station, there is a certain probability that the charging station will be randomly replaced with another optional charging station. This provides the algorithm with random search capability in charging station selection, which helps to find a better charging path planning.

[0227] Set the number of iterations to 100 times, and the algorithm iteration diagram is as follows: Figure 5As shown, the solving time is 4.74s, the optimal target value is 330.19, the optimal individual is [[0, 8, 9, 3, 2, 7, 5, 4, 1, 6, 10, 0], [0, 2, 0, 1, 0, 1, 0, 0, 1, 1, 0, 0]], and the path obtained by decoding is [('Drones', [(0, 8, 0), (9, 3, 2), (2, 7, 5), (4, 1, 10)]), ('Trucks', [0, 9, 2, 5, 4, 6, 10, 0])], the customer points served by the drones are [8, 3, 7, 1], the demand quantities are [1, 4, 2, 3], all of which meet the maximum load limit of the drones, the distances of the customer points served by the drones are [42.44, 43.61, 7.65, 46.48], all of which meet the endurance requirement of the drones, and the path diagram is as shown in Figure 6 The delivery schedule of the trucks and the drones in the FTSPD model is as shown in Figure 7

[0228] The GAFA-GS algorithm is used to solve the TSP model, the TSPD model and the STSPD model respectively under 10 customer points, and the delivery schedule is as shown in Figure 8 The optimal path diagrams obtained are as shown in Figure 9 It can be known from the comparison of the solving results of the four models in Table 3 that the GAFA-GS algorithm can solve the effective solutions of different models, and the superiority of the models is preliminarily proved.

[0229] Table 1 Example data

[0230]

[0231] Table 2 Parameters required in basic experiments

[0232]

[0233]

[0234] Table 3 Comparison of solving results of four models

[0235]

[0236] The above embodiments are only used to illustrate the technical solutions of the present application, but not limit the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those skilled in the art that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.​

Claims

1. A method for optimizing the coordination of a truck and a UAV based on FTSPD, characterized in that: The warehouse point position information, customer point position information and customer demand quantity are acquired, a truck and unmanned aerial vehicle (UAV) collaborative scheduling model is constructed with the shortest total time of the truck and UAV returning to the warehouse as the target, and a hybrid artificial fish swarm genetic algorithm with a greedy strategy is used to solve the model to obtain a scheduling scheme. 2.The FTSPD-based truck and drone collaborative scheduling optimization method of claim 1, wherein: The objective function is: minimize T total where T total is the total time for the truck and drone to return to the warehouse to complete the task; The mathematical expression of the time for the truck and UAV to pass between nodes i and j is: The constraint part includes basic routing constraints, UAV action constraints, time continuity constraints and synchronization constraints.

3. The FTSPD-based truck and UAV collaborative scheduling optimization method according to claim 2, characterized in that: When it is a new energy truck, the objective function is: minimize T total ,E total wherein E total is the energy consumption; An energy constraint is added: The new energy truck is equipped with a full battery at the beginning of the trip: e0 = E t The battery power of the new energy truck after passing through the arc segment (i, j) is tracked: The battery state of the new energy truck at the successor node after the charging node i: The maximum endurance distance of the UAV is set: The value of the polynomial is the maximum completion time of all trips: The total energy consumed by the new energy truck and the UAV is calculated. Wherein, n is the number of customer nodes that need to be served, R is the number of available charging stations, and r is the number of selected charging stations. 4.The FTSPD-based truck and drone collaborative scheduling optimization method of claim 2, wherein, The basic routing constraints include: Each customer can only be accessed by the truck or the UAV once; The truck leaves the warehouse and drives to any node in the customer and warehouse network; The truck returns to the warehouse from any node in the customer and warehouse network; The truck must enter and leave a certain customer node to ensure the flow continuity in the truck route; The truck trip connectivity sub-trip elimination constraint; M is a maximum number, and the constraint is activated and strictly implemented only when all conditions are met; The correct arrangement of the truck accessing customers is ensured; 5. The FTSPD-based truck and drone collaborative dispatch optimization method of claim 2, wherein, The UAV action constraints include: The UAV weight constraint, the demand of the customer served by the UAV is greater than the trip equal to 0: Each UAV can only be launched from each launch node at most once: Each UAV can only be retrieved from each rendezvous node at most twice: When the UAV is launched and retrieved at two different nodes, these nodes must be accessed by the truck: When the corresponding drone is launched from the warehouse and completes the task at node k, the truck must access the rendezvous node k∈V from any node R 6. The FTSPD-based truck and drone collaborative dispatch optimization method of claim 2, wherein, The time continuity constraints include: The time when the truck arrives at node k is greater than the time when it arrives at the previous node i: The time when the UAV arrives at the node j that meets the UAV condition is greater than the time when the corresponding truck and UAV itself arrive at the launch node i: The time when the UAV arrives at the rendezvous node k is greater than the time when the UAV arrives at the visited node j:

7. The FTSPD-based truck and drone collaborative dispatch optimization method of claim 2, wherein, The synchronization constraints include: The time when the truck and UAV arrive at the launch node i is the same: The time when the truck and UAV arrive at the rendezvous node k is the same: If the UAV is still serving customers and has not been recycled, the truck cannot launch the UAV: The time when all vehicles arrive at the depot node {0} is set to zero: The maximum flight distance of the UAV is set according to the endurance constraint of the UAV: The value of the polynomial is set to be the maximum completion time of all trips:

8. The FTSPD-based truck and drone collaborative dispatch optimization method according to any one of claims 3-5, wherein, n is the number of customer nodes requiring service, V is the set of all nodes in the network, V0 is the set of warehouse nodes and their virtual nodes, V D is the subset of customers eligible for UAV conditions, V C is the set of customer nodes, V L is the set of start locations, V R is the set of end locations, V N is the set of all nodes except the warehouse nodes, A is the set of (i,j) arcs, where i,j∈V and i≠j, is the truck travel time between nodes i and j, is the UAV travel time between nodes i and j, is the truck travel distance from node i to j, is the UAV travel distance from node i to j, q i is the node demand, S t is the service time generated by a truck serving a customer node, S d is the service time generated by a UAV serving a customer node, Q d is the UAV weight limit, L is the UAV flight distance limit, v t is the truck travel speed, v d is the UAV travel speed, p L is the setup time required to launch a UAV, p R is the setup time required to retrieve a UAV, x ij is a binary variable: 1 if truck moves from node i to node j; otherwise 0, y ijk is a binary variable: 1 if UAV performs visit (i,j,k); otherwise 0, P ij is a binary variable: 1 if node i serves before j without having to be adjacent; otherwise 0, is the time at which the truck arrives at node i, is the time at which the UAV arrives at node i, u i denotes the position of node i in the truck route.

9. The FTSPD-based truck and drone collaborative dispatch optimization method of claim 1, wherein, The steps of solving the hybrid artificial fish swarm genetic algorithm with a greedy strategy include: (1) Initialize parameters and population When initializing the population: The upper half array: generate the initial access path of the truck using the distance-based greedy strategy; The lower half array: generate an array with the same length as the upper half array, with 0 fixed at the beginning and end, representing the warehouse; (2) Fitness calculation and elitist selection Decode each individual in the population to get the specific driving route and service scheme of the truck and the drone, calculate the total time T for the scheme to complete all delivery tasks total , check whether the scheme violates the endurance constraint and the load constraint of the drone, if so, impose a very large penalty value on the total time , get the fitness value Z after the penalty, the smaller the Z, the better the scheme; (3) Update the population by the effects of pursuit and foraging; (4) Iteration and termination Repeat steps (2) and (3) using the new population as the initial population for the next generation until a maximum number of iterations is reached, and output the optimal solution found during the entire search process, i.e. the shortest total delivery time of the truck and UAV collaborative scheduling scheme.

10. The method of claim 8, wherein: When it is a new energy vehicle: add problem model parameters: electric vehicle energy consumption rate h t , unmanned aerial vehicle energy consumption rate h d , charging node electric vehicle charging rate R t , maximum energy capacity of electric vehicle, maximum energy capacity of unmanned aerial vehicle; Total energy consumption Z2: Calculate the total energy consumption of the new energy truck completing the entire path, and add a penalty for the truck energy constraint; Add a "transform charging station" mutation operation: During the mutation process, traverse the first half of the array, and if the node encountered is a charging station, there is a certain probability of randomly replacing the charging station with another available charging station.

Images