CN117291491A

CN117291491A - Urban logistics unmanned aerial vehicle path planning method considering dynamic wind speed and wind direction

Info

Publication number: CN117291491A
Application number: CN202311298865.6A
Authority: CN
Inventors: 杜鹏飞; 史悦强; 刘子悦; 卿朝进; 张学军
Original assignee: Xihua University
Current assignee: Xihua University
Priority date: 2023-10-09
Filing date: 2023-10-09
Publication date: 2023-12-26

Abstract

The invention provides a city logistics unmanned aerial vehicle path planning method considering dynamic wind speed and wind direction, which comprises the following steps: (1) system construction: (2) model construction: the wind speed and wind direction model construction is included; constructing a logistics unmanned aerial vehicle path planning model taking dynamic wind speed and direction into consideration; (3) The method comprises the steps of establishing a problem, and considering unmanned plane energy constraint, customer time window constraint and unmanned plane path planning problem representation under the influence of dynamic wind speed and wind direction; (4) problem solving: combining a large-scale neighborhood search algorithm LNS with GA to form a genetic algorithm GA-LNS combined with the large-scale neighborhood search algorithm to solve the problem of logistics unmanned aerial vehicle path planning based on static wind speed and wind direction. The invention can obtain better solution in reasonable time, and the distribution cost is reduced; compared with the condition without wind, static wind speed and wind direction, the path planning taking the dynamic wind speed and the wind direction into consideration can effectively avoid the problems of insufficient energy consumption constraint and customer satisfaction deviation caused by the former two conditions.

Description

Urban logistics unmanned aerial vehicle path planning method considering dynamic wind speed and wind direction

Technical Field

The invention provides a city logistics unmanned aerial vehicle path planning method considering dynamic wind speed and wind direction, and belongs to the technical field of unmanned aerial vehicle path planning.

Background

At present, the research of the path planning of the logistics unmanned aerial vehicle considers that the unmanned aerial vehicle lacks the influence on wind speed and wind direction on the basis of the energy constraint of the unmanned aerial vehicle, the time window of a client and other constraints, and particularly, the consideration of the unmanned aerial vehicle characteristics such as dynamic wind speed and wind direction is caused, so that the model is difficult to be applied to an actual transportation system.

At present, many researches on unmanned aerial vehicle route planning problems in logistics distribution scenes exist, and particularly, related researches on unmanned aerial vehicle logistics distribution route planning problems considering wind speed and wind direction are performed. The document [1] considers the influence of the actual load of the unmanned aerial vehicle on the flying speed and applies the influence to modeling of unmanned aerial vehicle logistics distribution path planning. On the basis of [ Funabashi Y, taniguchi I, tomiyama H.work-in-progress: routing of delivery Drones with load-dependent flight speed [ C ]//2019IEEE Real-Time Systems Symposium (RTSS) & IEEE,2019:520-523 ], the authors of [ Ito S, akaiwa K, funabashi Y, et al load and wind aware routing of delivery Drones [ J ]. Drones,2022,6 (2): 50 ]) further consider the effect of wind on time of flight, propose an extended VRP problem with the aim of minimizing total time of flight, and develop a dynamic planning algorithm to solve this problem. However, the authors do not fully consider unmanned energy consumption related effects and limitations, the customer point service time window, and dynamic changes in wind speed and direction. The authors of Sorbelli F B, cor co-fire F, palazzetti L, et al How the wind can be leveraged for saving energy in a truck-drone delivery system [ J ]. IEEE Transactions on Intelligent Transportation Systems,2023,24 (4): 4038-4049 ] have adjusted the flight path of the unmanned aerial vehicle according to the wind direction, and have transported the system in a truck-unmanned aerial vehicle series. The drone reacts positively to the wind by taking the "most downwind" available trajectory between the truck path and the cargo. To account for the risk of wind uncertainty resulting in time-of-flight variations and thus possible delivery delays [ Cheng C, aduyasak Y, rousseau L M, et al Robust drone delivery with weather information [ J ]. History,2020,1:1-37] authors have proposed a two-stage unmanned dispatch model to optimize delivery plans and use polar coordinates to represent wind vectors in the wind farm to determine speed variations of the unmanned in a single round trip warehouse point delivery. However, the authors only consider the delivery scenario for a single round trip warehouse point (warehouse-client-warehouse), failing to apply the wind farm further in the client-client delivery scenario. One unmanned aerial vehicle path planning problem that considers the influence of wind on energy consumption was studied in [ Radzki G, thiibotuwawa a, bocewicz g.uavs flight routes optimization in changing weather conditions-constraint programming approach [ J ]. Applied Computer Science,2019,15 (3): 5-12 ]. The authors assumed that each drone was flying at a fixed speed during the flight, that the wind speed and direction remained constant over time and that the total weight of the drone remained constant during the flight. However, because the parameters of the influencing factors are all fixed, the parameters can only represent a section of scene in the logistics distribution of the unmanned aerial vehicle and cannot be expanded into the whole logistics distribution process. [ Hamdi A, salim F D, kim D Y, et al Drone-as-a-service composition under uncertainty [ J ]. IEEE Transactions on Services Computing,2021,15 (5): 2685-2698.] proposes a sensing method for weather conditions in logistics distribution of unmanned aerial vehicles, which measures weather related parameters such as cloud coverage ratio, humidity, visibility and dew point temperature by considering sensor nodes on sites and airlines, and compares the measured parameters with set thresholds to judge whether the sites and the related airlines meet flight standards, and in addition, authors consider the influence of downwind and upwind on the actual flight speed of the unmanned aerial vehicle. The authors did not take into account the effect of crosswind on the actual flight speed of the drone. [ Peng L, murray C.Parallel Drone Scheduling Traveling Salesman Problem with Weather Impacts [ J ]. Available at SSRN 4254262,2022 ] quantitatively analyzes the influence of wind speed on the maximum distribution range of the unmanned aerial vehicle, and the areas and routes generated by rainfall and not meeting the flight standard, defines several bad weather scenes not meeting the flight standard, and researches how to reduce the influence of the bad weather scenes on the distribution time of the unmanned aerial vehicle under different weather conditions. However, the authors only consider the delivery scenario for a single round trip warehouse point (warehouse-client-warehouse), failing to apply the wind farm further in the client-client delivery scenario.

Disclosure of Invention

Based on the limitations, the invention solves the problem of logistics unmanned aerial vehicle path planning considering unmanned aerial vehicle energy consumption constraint, customer time window and wind speed and wind direction influence by utilizing an improved genetic algorithm. The method aims at adopting a genetic algorithm (GA-LNS) combined with a large-scale neighborhood search algorithm, and using a dynamic programming idea in a decoding process of the GA-LNS, and researching and considering unmanned aerial vehicle path programming problems under unmanned aerial vehicle energy constraint, client time window constraint and dynamic wind speed and wind direction influence so as to minimize the weight values of unmanned aerial vehicle fixed cost, energy consumption cost and client satisfaction punishment cost.

A city logistics unmanned aerial vehicle path planning method considering dynamic wind speed and wind direction comprises the following steps:

(1) And (3) system construction:

in the urban logistics distribution scene, the logistics distribution system consists of a plurality of charge type multi-rotor unmanned aerial vehicles with limited loads and capable of taking off and landing vertically, and the unmanned aerial vehicles provide logistics distribution services for delivering and taking goods for all client points in a distribution range according to a distribution path planned by a dispatching system after loading goods from a warehouse and replacing batteries.

(2) Model construction:

(2.1) wind speed and wind direction model construction

Establishing a vector speed triangle, wherein the direction vertically upwards along the y-axis in the plane is the positive direction of the system; airspeed represents the speed of flight of the unmanned aerial vehicle relative to air, abbreviated as TAS; wind speed represents the speed of movement of air relative to the ground, abbreviated WS; ground speed represents the flight speed of the unmanned aerial vehicle relative to the ground, abbreviated as GS; wind direction represents the angle defined clockwise from the positive direction to the direction of the wind, abbreviated as σ; the course angle represents the angle from the defined clockwise direction to the course line of the unmanned plane, and is abbreviated as alpha; the track angle represents the angle from the defined forward direction clockwise to the unmanned aerial vehicle track, abbreviated as beta; the drift angle represents the angle of the aircraft from the course under the influence of the crosswind, abbreviated as gamma, and the wind angle represents the included angle between the course line and the wind direction line, abbreviated as theta.

Due to the objective existence of air movement, the movement of the unmanned aerial vehicle in the air is decomposed into: operation of the unmanned aerial vehicle relative to air, movement of the air relative to the ground and movement of the unmanned aerial vehicle relative to the ground, wherein the three are in the relationship of a ground speed vector=an airspeed vector+a wind speed vector, namely:

by usingRepresenting n+1 nodes in the system consisting of 1 warehouse and n client points, where c ⁰ Represents a warehouse point, C' =c\ { C } ⁰ And represents a set of client points. Then for any starting point c in the system ⁱ And target point c ^j Let the coordinates of two nodes be (a) _i ,b _i ) And (a) _j ,b _j ) Then from c ⁱ Pointing to c ^j The calculation formula of the track angle beta is as follows:

and then the wind angle theta is obtained:

the drift angle gamma, the course angle alpha and the ground speed GS are respectively obtained by solving the vector speed triangle:

α＝β±θ (5)

equation (2) - (6) shows the solving process of unknown parameters in the vector velocity triangle, including track angle β, wind angle θ, drift angle γ, heading angle α and ground speed GS. The course angle alpha is used for eliminating the influence of drift on the flight track of the unmanned aerial vehicle, and ensuring the connecting line of the unmanned aerial vehicle to the target point along the starting point, namely the initial course when the course flies, and the ground speed GS is the speed which is changed for keeping flying along the course under the influence of wind speed and wind direction.

(2.2) Logistics unmanned aerial vehicle path planning model considering dynamic wind speed and direction

Describing the logistics distribution scene in a Cartesian plane and usingRepresenting n+1 nodes in the system consisting of 1 warehouse and n client points, where c ⁰ Represents a warehouse point, C' =c\ { C } ⁰ And represents a set of client points. Each node is connected withSet to {(s) ⁱ ,d ⁱ ,p ⁱ ,li,r ⁱ ) (s is therein ⁱ ＝(a _i ,b _i ) Representing the 2-dimensional position coordinates of node i, d ⁱ Representing delivery demand of node i, p ⁱ Representing the pick demand of node i (warehouse point c ⁰ Both delivery and pick-up demand of 0), l ⁱ And r ⁱ Then lambda is defined representing the earliest and latest acceptable service time windows at client point i, respectively ₁ And lambda (lambda) ₂ The system is a penalty coefficient when the actual arrival time of the unmanned aerial vehicle is earlier than the earliest service time window of the client point and later than the latest service time of the client point. For different combinations of customer point pickup delivery needs,set->Representation unmanned aerial vehicle _u The sum of the descent, ascent and hover service times at client point i. The relevant parameter symbols and their interpretation are shown in table 1.

TABLE 1 parameter symbols and interpretation in the model

By usingRepresenting unmanned aerial vehicle set in delivery system, beta _ij For track angle, alpha, between any two nodes _ij And GS _ij Heading and ground speed under the initial track keeping condition are respectively.

The dynamic change of the wind speed and the wind direction is scattered to a time window with a certain length, namely the wind speed and the wind direction in different time windows are different, meanwhile, the influence of the wind speed and the wind direction in the whole distribution range and the whole distribution process of the unmanned aerial vehicle is global, and the wind speed and the wind direction information of each time window is predicted in advance.

By usingRepresenting a set of time window sequences, there being different ground speeds and heading angles between different nodes and within different time window sequences, respectively defined as +.>And->Distribution of client point i to client point jThe service is provided with three time sequences, wherein each time sequence t has corresponding wind speed and wind direction, and the affected ground speed and course angle. The time of changing course angle during the flight is made negligible.

Setting variablesThe time of flight of unmanned plane u in time sequence t when performing the dispatch service of node i to node j is represented.

(1) Objective function

The objective function is constructed based on several factors:

(1.1) fixed cost

The fixed cost is related to the number of drones in the system, expressed as:

(1.2) energy consumption costs

The energy consumption power calculation formula under the hovering state of the multi-rotor unmanned aerial vehicle is as follows:

wherein W is the unmanned aerial vehicle's empty machine weight, contains the battery load, and m is unmanned aerial vehicle's actual load, and g is gravity acceleration, and ρ is air fluid density, and ζ and n are the area and the rotor number of rotor respectively, because in actual delivery process, the parameter under the root number in formula (8) is known, models unmanned aerial vehicle's energy consumption power as:

wherein the method comprises the steps ofm _ij Is a slave node of the unmanned aerial vehiclei actual payload of fly to node j, P _ij And (5) energy consumption power for the unmanned aerial vehicle to fly from the node i to the node j. Distribution distance dis between node i and node j _ij Equal to the sum of the cumulative flight distances at each time series, namely:

considering the time variability of wind speed and direction, the total energy consumption calculation formula in the logistics distribution system is as follows:

wherein the method comprises the steps ofIndicating whether the distribution service between node i and node j is performed by drone u.

(1.3) customer satisfaction penalty cost

Taking a soft time window [ l ] ⁱ ,r ⁱ ]Form definition of customer's acceptable earliest and latest service time windows, modeling customer satisfaction as penalty cost violating customer satisfaction, using λ ₁ And lambda (lambda) ₂ And penalty coefficients representing unit time when the actual arrival time of the unmanned aerial vehicle is earlier than the earliest service time window of the client point and later than the latest service time of the client point. By using The time that unmanned plane u arrives at node j from node i is represented, when the time that unmanned plane arrives at a client point accords with a time window of the client point, the client satisfaction punishment cost is 0, when the time that unmanned plane arrives at the client point is earlier or later than a time window acceptable by the client point, a certain punishment is received, and the satisfaction punishment of the client point is modeled as follows:

combining the fixed cost, the energy consumption cost and the customer satisfaction penalty cost, respectively increasing the fixed cost and the energy consumption cost from the operator perspective and the customer satisfaction penalty cost from the customer perspective by a weighted value k ₁ And kappa (kappa) ₂ ，κ ₁ +κ ₂ =1. The objective function of the logistics unmanned aerial vehicle path planning based on the dynamic static wind speed and direction is as follows:

(2) Constraint conditions

(2.1) unmanned aerial vehicle mission-related constraints

Introducing a 0-1 variable

To indicate whether the delivery service between node i and node j is performed by drone u, then set:

to ensure that each customer is at a point and only one drone performs the distribution service. By arranging

To ensure that the drone will leave from customer point j after performing the delivery task at customer point j.

(2.2) unmanned aerial vehicle reuse related constraints

Introducing a 0-1 variable

To indicate whether the unmanned plane u can execute the next delivery service again after returning to the warehouse, if the unmanned plane u returns to the warehouse after executing the delivery service of the client point i, and executes the delivery service of the client point j after reloading and battery replacement, then Otherwise->

In order to ensure that the unmanned aerial vehicle u returns to the warehouse to be reused to perform the delivery task of the next client point j after performing the delivery service of the client point i, the following constraint is set

(2.3) unmanned aerial vehicle number constraint

In order to ensure that the number of unmanned aerial vehicles performing tasks in the same period does not exceed the number of unmanned aerial vehicles arranged in the system, constraint is carried out:

(2.4) unmanned aerial vehicle heavy restraint

In order to ensure that the load of the unmanned aerial vehicle at any node is always smaller than the maximum weight limit Q of the unmanned aerial vehicle, the constraint is that:

0≤m _ij ≤Q,i∈C,j∈C.i≠j (21)

(2.5) unmanned aerial vehicle heavy restraint

Introducing variablesIndicating that unmanned plane u goes from node i to node jIn order to ensure that the unmanned aerial vehicle always keeps a full-power state when leaving the warehouse, and the residual electric quantity at each flying node can not be 0 all the time, constraint:

wherein E is _max Is the maximum electric quantity of the unmanned aerial vehicle. The electric quantity consumption of the unmanned aerial vehicle in the process of executing the distribution from the node i to the node j is as follows:

i∈C,j∈C',k∈C,i≠j,j≠k,u∈U,t∈T

(3) Problem establishment

Based on the system model, the objective is to minimize the weight of the unmanned plane fixed cost, the energy consumption cost and the customer satisfaction punishment cost by optimizing the decision variables in the unmanned plane path planning problem model under the consideration of unmanned plane energy consumption constraint, customer time window constraint and dynamic wind speed and wind direction influence.

Based on the model, unmanned plane path planning problems under the influence of unmanned plane energy constraint, client time window constraint and dynamic wind speed and wind direction are expressed as follows:

(4) Problem solving:

combining a large-scale neighborhood search algorithm LNS with GA to form a genetic algorithm GA-LNS combined with the large-scale neighborhood search algorithm to solve the problem of path planning of the logistics unmanned aerial vehicle based on static wind speed and direction;

(4.1) encoding

The code is in the form of an integer permutation. In the system, n client points are shared, a certain population number of chromosomes with the length of 2n+1 are generated during decoding, and the n client points at most need to complete distribution service through 2n+1 nodes, and the number of the client points is larger than that of the warehouse points.

Meanwhile, the coding form obtained after the earliest service time window acceptable by the client, the energy consumption constraint of the unmanned aerial vehicle and the weight constraint judgment of the unmanned aerial vehicle is the form of the initialized population. And introducing a penalty factor for violating the model constraint conditions in the encoding process, namely carrying out quantization processing on the unmanned aerial vehicle by the penalty factor under the condition of violating the energy consumption constraint of the unmanned aerial vehicle and the weight constraint of the unmanned aerial vehicle.

(4.2) population fitness calculation

The fitness of each chromosome in the population is calculated through a fitness function based on the coding form of the chromosome, and is used for evaluating the fitness of each chromosome, and the chromosomes are selected according to the fitness to carry out crossover and mutation operations so as to generate the next generation of chromosomes, and the chromosomes with high fitness are more likely to be selected and reserved, so that the fitness of the whole population is gradually improved.

(4.3) crossover and mutation

Crossover is used to generate next generation candidate solutions, and the mutation is to avoid sinking into the locally optimal solution and provide new search directions, through which GA can realize exploration and optimization in solution space.

(4.4) Large Scale neighborhood search

The LNS adjusts the solution in the neighborhood of the feasible solution mainly through the methods of destruction and repair, thereby obtaining the solution with high quality. The "destroy" and "repair" operations are implemented by removing operators and inserting operators, respectively.

For the "destroyed" part, the solution is adjusted by adopting a similarity removing operator, wherein the operator firstly randomly selects one client i, removes the client i from the existing distribution scheme, and then calculates the similarity measure R of the rest clients and the client i in the scheme _ij Taking the customer with the maximum similarity to performAnd removing, and repeating the above operation until the given number of clients are removed. R is R _ij The calculation formula is as follows:

wherein the distance dis from node i to node j _ij Embody the difference of distance, |l ⁱ -l ^j I represents the difference in the earliest service time window acceptable, |p ⁱ +d ⁱ -p ^j -d ^j The I represents the difference of the delivery demand and the picking demand of the customer point, and all three factors influence the measurement of the similarity, and eta is introduced in consideration of normalization ₁ 、η ₂ And eta ₃ The weight of each influencing factor is represented.

The dynamic wind speed and direction can enable the unmanned aerial vehicle to have dynamic change of ground speed when distributed among different nodes and different time window sequencesCourse angle changed with it>And (3) in the decoding process of the GA-LNS, solving the time for the unmanned aerial vehicle to reach each point in the distribution sequence under the dynamic wind speed and wind direction. Wherein i represents the node sequence label in the delivery sequence of the unmanned plane u, the unmanned plane sequentially accesses the nodes corresponding to the label, the loading and unloading of cargoes and the replacement of batteries are completed at a warehouse point, the picking and delivering service of cargoes is completed at a client point, t represents the current time window sequence, and different ground speeds and course angles corresponding to the unmanned plane when providing delivery service for different adjacent nodes in different time window sequences are represented by GS and alpha. For a time window sequence with time spans of ζ, it is divided into elapsed times t _used And unused time t _rest By t _start And t _end Representing a time window sequence in which delivery starts and a time window sequence in which delivery completes, and passing through a distance dis of neighboring nodes _i-1,i And +.>Determining a time window sequence t in which delivery is completed _end And further obtaining a time point when the unmanned aerial vehicle reaches the node i, and obtaining a set arr of the arrival time of each node through judging and accessing the nodes in the distribution sequence of the unmanned aerial vehicle u.

According to the method, based on the influence of the dynamic wind speed and the wind direction on the flight state of the unmanned aerial vehicle, the flight parameters of the unmanned aerial vehicle under the influence of the wind speed and the wind direction and the solving method thereof are determined, and modeling which considers the unmanned aerial vehicle energy constraint, the client time window and the logistics unmanned aerial vehicle path planning problem under the influence of the dynamic wind speed and the wind direction is completed. To solve this integer programming problem, the present invention combines a large-scale neighborhood search algorithm (LNS) with a Genetic Algorithm (GA) to form a genetic algorithm (GA-LNS) that combines the large-scale neighborhood search algorithm, and uses the dynamic programming concept in the decoding process of the GA-LNS to solve the dynamic problem. According to the invention, a large number of simulation experiments are carried out, and simulation results show that the method can obtain a better solution in a reasonable time, and compared with the traditional GA, the distribution cost is reduced by about 9%; compared with the condition without wind, static wind speed and wind direction, the path planning taking the dynamic wind speed and the wind direction into consideration can effectively avoid the problems of insufficient energy consumption constraint and customer satisfaction deviation caused by the former two conditions.

Drawings

FIG. 1 illustrates a municipal logistics unmanned aerial vehicle distribution system;

FIG. 2 vector speed triangle;

FIG. 3 is a graph of wind speed and direction dynamics;

FIG. 4 is a diagram of a single unmanned aerial vehicle logistics distribution process based on dynamic wind speed and direction;

FIG. 5 customer satisfaction and penalty cost schematic;

FIG. 6 is a schematic diagram of crossover and mutation operations;

FIG. 7 is a graph of maximum delivery distance versus different orientations;

FIG. 8 is a diagram of a different orientation Cheng Desu comparison;

FIG. 9 is a graph of energy consumption versus adjacent nodes for the same task;

FIG. 10 is a route diagram of an optimal delivery scheme;

FIG. 11 is an iterative convergence diagram;

FIG. 12 is a comparison of the algorithm;

FIG. 13 is a graph showing comparison of dispensing completion times under different conditions;

FIG. 14 is a genetic algorithm flow diagram incorporating a large-scale neighborhood search algorithm;

FIG. 15 is a flow chart of a method of similarity removal operator.

Detailed Description

The specific technical scheme of the invention is described by combining the embodiments.

(1) System description: in the urban logistics distribution scenario, as shown in fig. 1, the logistics distribution system is composed of a plurality of vertically-lifting rechargeable multi-rotor unmanned aerial vehicles with limited loads, and the unmanned aerial vehicles provide logistics distribution services for delivering and taking goods to each customer point in a distribution range according to a distribution path planned by a dispatching system after loading goods from a warehouse and replacing batteries.

The invention is based on the following assumptions: (1) the picking and delivering demands of each customer point are not detachable, and each customer point has and is only served by one unmanned aerial vehicle for picking and delivering goods. (2) The battery capacity of the unmanned aerial vehicle and the cargo load of the unmanned aerial vehicle are both limited. (3) The demand of a single client point cannot exceed the cargo weight limit of the unmanned aerial vehicle, and the position of the single client point cannot exceed the maximum distribution range of the unmanned aerial vehicle limited by the battery power. (4) The unmanned aerial vehicle carries out logistics distribution according to a path planned by the system, the air speed of the unmanned aerial vehicle is kept unchanged in the distribution process, and when flying across nodes or a time window sequence, the unmanned aerial vehicle can maintain flying on a preset track by changing the course, and meanwhile, the time spent by the action of changing the course and errors possibly caused are ignored. (5) The unmanned aerial vehicle has negligible time to replace the battery and load and unload the goods in the warehouse.

(2) Model construction:

(2.1) wind speed and wind direction model construction

When the unmanned aerial vehicle carries out the flight of logistics distribution, because the objectivity of air motion exists, the unmanned aerial vehicle can not always realize the point-to-point flight according to the route of expecting, and unmanned aerial vehicle's flight orbit and delivery time all can receive corresponding influence under no wind or quiet wind condition, down wind condition, upwind condition and crosswind condition. For ease of understanding, the present invention builds a vector velocity triangle as shown in FIG. 2, wherein the direction vertically upward in the plane along the y-axis is the positive direction of the system; airspeed represents the speed of flight of the unmanned aerial vehicle relative to air, abbreviated as TAS; wind speed represents the speed of movement of air relative to the ground, abbreviated WS; ground speed represents the flight speed of the unmanned aerial vehicle relative to the ground, abbreviated as GS; wind direction represents the angle defined clockwise from the positive direction to the direction of the wind, abbreviated as σ; the course angle represents the angle from the defined clockwise direction to the course line of the unmanned plane, and is abbreviated as alpha; the track angle represents the angle from the defined forward direction clockwise to the unmanned aerial vehicle track, abbreviated as beta; the drift angle represents the angle of the aircraft from the course under the influence of the crosswind, abbreviated as gamma, and the wind angle represents the included angle between the course line and the wind direction line, abbreviated as theta.

Due to the objective existence of air movement, the movement of the unmanned aerial vehicle in the air can be decomposed into: operation of the unmanned aerial vehicle relative to air, movement of air relative to the ground and movement of the unmanned aerial vehicle relative to the ground, the three being in relation of ground velocity vector = airspeed vector + wind velocity vector, i.e

Herein, use ofRepresenting n+1 nodes in the system consisting of 1 warehouse and n client points, where c ⁰ Represents a warehouse point, C' =c\ { C } ⁰ And represents a set of client points. Then for any starting point c in the system ⁱ And target point c ^j Let the coordinates of two nodes be (a) _i ,b _i ) And (a) _j ,b _j ) Then from c ⁱ Pointing to c ^j The calculation formula of the track angle beta is as follows:

i∈C,j∈C,i≠j

and then the wind angle theta can be obtained:

the drift angle gamma, the course angle alpha and the ground speed GS can be obtained by solving the vector speed triangle, and are respectively as follows:

α＝β±θ (5)

equation (2) - (6) shows the solution of unknown parameters in the vector velocity triangle, such as track angle β, wind angle θ, drift angle γ, heading angle α, and ground speed GS. The course angle alpha is used for eliminating the influence of drift on the flight track of the unmanned aerial vehicle, and ensuring the connecting line of the unmanned aerial vehicle to the target point along the starting point, namely the initial course when the course flies, and the ground speed GS is the speed which is changed for keeping flying along the course under the influence of wind speed and wind direction.

The present invention describes the logistics distribution scenario in a Cartesian plane and usesRepresenting n+1 nodes in the system consisting of 1 warehouse and n client points, where c ⁰ Represents a warehouse point, C' =c\ { C } ⁰ And represents a set of client points. Every node +.>Set to {(s) ⁱ ,d ⁱ ,p ⁱ ,l ⁱ ,r ⁱ ) (s is therein ⁱ ＝(a _i ,b _i ) Representing the 2-dimensional position coordinates of node i, d ⁱ Representing delivery demand of node i, p ⁱ Representing the pick demand of node i (warehouse point c ⁰ Both delivery and pick-up demand of 0), l ⁱ And r ⁱ Then lambda is defined representing the earliest and latest acceptable service time windows at client point i, respectively ₁ And lambda (lambda) ₂ The system is a penalty coefficient when the actual arrival time of the unmanned aerial vehicle is earlier than the earliest service time window of the client point and later than the latest service time of the client point. Setting +.>Representation unmanned aerial vehicle _u The sum of the descent, ascent and hover service times at client point i. For ease of understanding, the relevant parameter symbols and their interpretation are shown in table 1.

As shown in fig. 3, the present invention disperses the dynamic change of wind speed and direction into a time window with a certain length, that is, the wind speed and direction in different time windows are different, and meanwhile, it is assumed that the influence of wind speed and direction in the whole distribution range and the whole distribution process of the unmanned aerial vehicle is global, and the wind speed and direction information of each time window is predicted in advance.

By usingRepresenting a set of time window sequences, since wind speed and direction will dynamically change with time window sequences,therefore, as in fig. 4, the logistics distribution process of the single unmanned aerial vehicle has different ground speeds and course angles among different nodes and in different time window sequences, and the invention defines the ground speeds and the course angles as +.>And->For example, in fig. 4, the distribution service from client i to client j, the service has undergone five time sequences, each having a corresponding wind speed and direction, and the affected ground speed and heading angle. For ease of calculation, it is assumed herein that the time to change heading angle during flight is negligible.

To describe the dynamic changes described above, variables are setThe time of flight of unmanned plane u in time sequence t when performing the dispatch service of node i to node j is represented.

(1) Objective function

In unmanned aerial vehicle logistics distribution, the operator is the provider of the service and the customer is the recipient of the service. Operators desire to provide high quality services at lower cost with security ordered, and customers are more concerned about quality of service, i.e. time for picking/delivering goods. Considering the differences between operator and customer points of interest in combination, the present invention will construct an objective function based on several factors.

(1.1) fixed cost

When the unmanned aerial vehicle is dispatched to execute logistics distribution, certain fixed cost is generated, the cost is irrelevant to the flight distance and the flight time of the unmanned aerial vehicle, the unmanned aerial vehicle mainly comprises daily maintenance cost of the unmanned aerial vehicle, depreciation cost of the unmanned aerial vehicle and the like, and the fixed cost is related to the number of the unmanned aerial vehicles in the system and can be expressed as:

(1.2) energy consumption costs

wherein W is the air-machine weight of the unmanned aerial vehicle (including the battery load), m is the actual load of the unmanned aerial vehicle, g is the gravitational acceleration, ρ is the air fluid density, ζ and n are the area of the rotor and the number of rotors, respectively, and since the parameters under the root number in equation (8) are known in the actual delivery process, the energy consumption power of the unmanned aerial vehicle is further modeled herein as:

Wherein the method comprises the steps ofm _ij For the actual load of the unmanned aerial vehicle from node i to node j, P _ij And (5) energy consumption power for the unmanned aerial vehicle to fly from the node i to the node j. Distribution distance dis between node i and node j _ij Equal to the sum of the cumulative flight distances at each time series, namely:

wherein the method comprises the steps ofIndicating that the distribution service between node i and node j isWhether performed by the drone u.

(1.3) customer satisfaction penalty cost

The invention will take the form of a soft time window [ l ] ⁱ ,r ⁱ ]Form definition of acceptable earliest and latest service time windows for a customer, as shown in FIG. 5, customer satisfaction is modeled as a penalty cost violating customer satisfaction, expressed in lambda ₁ And lambda (lambda) ₂ And penalty coefficients representing unit time when the actual arrival time of the unmanned aerial vehicle is earlier than the earliest service time window of the client point and later than the latest service time of the client point. By usingRepresenting the time when the unmanned plane u starts from the node i to reach the node j, when the time when the unmanned plane reaches the client point accords with the time window of the client point, the client satisfaction punishment cost is 0, and when the time when the unmanned plane reaches the client point is earlier or later than the time window acceptable by the client point, a certain punishment is received, so the satisfaction punishment of the client point can be modeled as follows:

(2) Constraint conditions

(2.1) unmanned aerial vehicle mission-related constraints

Introducing a 0-1 variable

to ensure that each customer is at a point and only one drone performs the distribution service. By setting up:

(2.2) unmanned aerial vehicle reuse related constraints

The 0-1 variable was introduced:

to indicate whether the unmanned plane u can execute the next delivery service again after returning to the warehouse, if the unmanned plane u returns to the warehouse after executing the delivery service of the client point i, and executes the delivery service of the client point j after reloading and battery replacement, thenOtherwise->

In order to ensure that the unmanned aerial vehicle u returns to the warehouse after performing the delivery service of the client point i and can be reused to perform the delivery task of the next client point j, the following constraints are set:

(2.3) unmanned aerial vehicle number constraint

(2.4) unmanned aerial vehicle heavy restraint

In order to ensure that the load of the unmanned aerial vehicle at any node is always smaller than the maximum limit weight Q of the unmanned aerial vehicle, constraint is carried out

0≤m _ij ≤Q,i∈C,j∈C.i≠j (21)

(2.5) unmanned aerial vehicle heavy restraint

Introducing variablesThe method is characterized in that the residual electric quantity of the unmanned aerial vehicle u from the node i to the node j is represented, in order to ensure that the unmanned aerial vehicle always keeps a full-power state when leaving a warehouse, the residual electric quantity at each flying node can not be 0 all the time, and the constraint is that:

i∈C,j∈C',k∈C,i≠j,j≠k,u∈U,t∈T

(3) Problem establishment

Based on the system model, the invention aims to minimize the weight values of the unmanned plane fixed cost, the energy consumption cost and the customer satisfaction punishment cost by optimizing the decision variables in the unmanned plane path planning problem model under the consideration of unmanned plane energy consumption constraint, customer time window constraint and dynamic wind speed and wind direction influence.

Based on the foregoing model, unmanned plane path planning problems under the influence of unmanned plane energy constraints, customer time window constraints and dynamic wind speeds and wind directions are considered as follows:

(4) Problem solving:

since 1976, a Holland proposed Genetic Algorithm (GA) is commonly applied to solve various complex optimization problems, the GA is a highly parallel adaptive optimization algorithm, and has global searching capability, various solutions can be explored in a solution space through operations such as crossing and mutation, but meanwhile, the GA also has the problems of slower convergence speed and possibly trapping in local optimization. The flow is shown in fig. 14.

(4.1) encoding

The coding process in GA is the process of representing the solution of the model to be optimized as chromosome. Considering that the unmanned aerial vehicle logistics distribution path planning problem studied herein is that one warehouse point performs distribution service on a plurality of client points, continuity is provided among the client points, and generation of invalid solutions is reduced, so that the unmanned aerial vehicle logistics distribution path planning method takes the form of codes of permutation integers. As described in table 3, there are n customer points in the system, so in order to implement global searching in subsequent crossover and mutation operations as much as possible, a population number of chromosomes with lengths of 2n+1 are generated during decoding (the n customer points need to complete distribution service through 2n+1 nodes at most), and the numbers of the customer points are all numbers of warehouse points.

Meanwhile, the coding form obtained after the earliest service time window acceptable by the client, the energy consumption constraint of the unmanned aerial vehicle and the weight constraint judgment of the unmanned aerial vehicle is the form of the initialized population. For the violation of the model constraint conditions in the encoding process, the introduction of a penalty factor, namely, for the violation of the unmanned aerial vehicle energy consumption constraint and the unmanned aerial vehicle weight constraint, the quantization processing is carried out on the unmanned aerial vehicle energy consumption constraint and the unmanned aerial vehicle weight constraint through the penalty factor, is considered.

(4.2) population fitness calculation

(4.3) crossover and mutation

As shown in fig. 6, crossover and mutation are two key operations in the GA, crossover is used to generate next generation candidate solutions, mutation is used to avoid sinking into a locally optimal solution, and a new search direction is provided, through which the GA can realize exploration and optimization in the solution space.

(4.4) Large Scale neighborhood search

For the "destroyed" part, the invention adopts a method of removing the operator of the similarity to adjust the solution, as shown in fig. 15, the operator firstly randomly selects one client i, removes the client i from the existing distribution scheme, and then calculates the similarity measure R of the rest clients and the client i in the scheme _ij Removing the clients with the maximum similarity, and repeating the above operation until the clients with the set number are removed. R is R _ij The calculation formula is as follows:

The dynamic wind speed and direction can enable the unmanned aerial vehicle to have dynamic change of ground speed when distributed among different nodes and different time window sequences Course angle changed with it>Increasing the difficulty of solving. In order to solve the problem, the invention uses dynamic programming in the decoding process of the GA-LNS to solve the time when the unmanned aerial vehicle arrives at each point in the distribution sequence under the dynamic wind speed and direction. Wherein i represents the node sequence label in the delivery sequence of unmanned plane u, unmanned plane u accesses the nodes corresponding to the label in turn, finishes loading and unloading of goods and replacement of batteries at warehouse points, finishes the service of taking and delivering goods at client points, t represents the current time window sequence, and unmanned plane provides delivery service for different adjacent nodes in different time window sequencesThe various ground speeds and heading angles corresponding to the business hours are denoted by GS and alpha. For a time window sequence with time spans of ζ, the invention divides the time window sequence into the elapsed time t _used And unused time t _rest By t _start And t _end Representing a time window sequence in which delivery starts and a time window sequence in which delivery completes, and passing through a distance dis of neighboring nodes _i-1,i And +.>Determining a time window sequence t in which delivery is completed _end And further obtaining a time point when the unmanned aerial vehicle reaches the node i, and obtaining a set arr of the arrival time of each node through judging and accessing the nodes in the distribution sequence of the unmanned aerial vehicle u.

(4.4) details of the algorithm

/>

Input: task _u ,i＝1,t＝1,t _start ＝t _end ＝1,t _used ＝0,t _rest ＝ξ,GS,α

1. Judging the residual time t of the current time window sequence _rest Whether or not the distribution between the nodes i to i-1 can be completed

In a logistics distribution model, the invention adopts an ARK-150 six-rotor unmanned plane for logistics distribution, and the parameters of the model are that the aircraft weight is 100kg, the maximum load is 30kg, the battery power is 1000kwh, and the airspeed is 20m/s, and under the condition of being equipped with a high-energy battery, 50kg of cargoes can be loaded on the altitude of 4000m at the maximum to fly at the flying speed of 72km/h for about 20km.

Specific wind speed and direction and time window information are shown in table 2:

the simulation output result is as follows:

1. maximum delivery distance versus graph for different orientations:

maximum delivery distance pairs for different orientations centered on a warehouse point under four conditions no wind, 5m/s wind speed and 33 ° wind direction, 7m/s wind speed and 33 ° wind direction, 10m/s wind speed and 33 ° wind direction under full load are as shown in fig. 7:

the maximum delivery distance is reduced to a different extent when considering wind speed and wind direction conditions, compared to the maximum delivery distance of 10km when no wind, and the larger the wind speed, the larger the reduction amount of the maximum delivery distance is, the maximum delivery distance is reduced by 25% at the wind speed of 10m/s at most, which results in that the clients in the reduced distance section may not provide logistics distribution service by the unmanned aerial vehicle under the specific wind speed conditions.

2. Forward ground speed contrast map of different orientations:

as shown in fig. 8, the change of the forward ground speed at different azimuth angles in four cases under full load can be seen that there is a risk that the customer needs a longer delivery time due to the change of the forward ground speed, and thus a delivery delay is caused.

3. The energy consumption diagram of the adjacent nodes with the same task:

as shown in fig. 9, the present invention studied the influence of wind speed and wind direction on energy consumption under the same distribution task. The total energy consumption of the task under the condition of no wind is 83.92kwh, and the total energy consumption is increased along with the increase of the wind speed, specifically, the wind speed of 5m/s corresponds to the total energy consumption of 85.84kwh, the wind speed of 7m/s corresponds to the total energy consumption of 87.74kwh, and the wind speed of 10m/s corresponds to the total energy consumption of 92.15 kwh. Meanwhile, the energy consumption between adjacent nodes also has different changes, so that the risk that the customer point cannot finish logistics distribution according to a planned path due to the increase of the actual energy consumption exists.

4. An optimal delivery scheme roadmap, as shown in FIG. 10;

5. iterative convergence process diagram, as in fig. 11;

it can be seen that as the iteration number increases, the optimal function value, the energy consumption and the customer satisfaction penalty gradually decrease, and the convergence effect is achieved after a plurality of iterations.

6. Algorithm contrast graph, as in fig. 12;

as can be seen from fig. 12, under the same data set, parameter set and iteration number, the solution of GA is 2376.7, and the solution of GA-LNS is 2162.8, which improves about 9%, so that the GA-LNS can effectively improve the solution efficiency. Meanwhile, the GA has the problem of being trapped in local optimum, the convergence rate is not as fast as that of the GA-LNS, and the solution of the GA-LNS is better than that of the GA. It can be seen that the GA-LNS can get a better allocation scheme than the conventional GA.

7. A comparison of dispensing completion times under different conditions, as shown in fig. 13;

fig. 13 is a graph comparing the time for the unmanned aerial vehicle to perform the delivery task to reach each task node under the condition of no wind, static wind and dynamic wind, and it can be seen from the graph that the time for reaching each task node is different under different conditions, which may result in that the route planned without considering the conditions such as dynamic and the like may not meet the requirement of the customer time window, and the customer satisfaction is affected.

It can be seen that the method provided by the invention can obtain a better solution in a reasonable time, the distribution cost is reduced by about 9% compared with the traditional GA, and meanwhile, compared with the condition of no wind, static wind speed and wind direction, the path planning taking the dynamic wind speed and wind direction into consideration can also effectively avoid the problems of insufficient energy consumption constraint and customer satisfaction deviation caused by the former two conditions.

Claims

1. A city logistics unmanned aerial vehicle path planning method considering dynamic wind speed and wind direction is characterized by comprising the following steps:

(1) And (3) system construction:

in an urban logistics distribution scene, a logistics distribution system consists of a plurality of vertically-lifting rechargeable multi-rotor unmanned aerial vehicles with limited loads, and the unmanned aerial vehicles provide logistics distribution services for delivering and taking goods for all client points in a distribution range according to a distribution path planned by a dispatching system after loading goods from a warehouse and replacing batteries;

(2) Model construction:

the wind speed and wind direction model construction is included; constructing a logistics unmanned aerial vehicle path planning model taking dynamic wind speed and direction into consideration;

(3) Problem establishment

Based on the model, unmanned plane path planning problem representation under the influence of unmanned plane energy constraint, client time window constraint and dynamic wind speed and wind direction is considered;

(4) Problem solving:

combining a large-scale neighborhood search algorithm LNS with GA to form a genetic algorithm GA-LNS combined with the large-scale neighborhood search algorithm to solve the problem of logistics unmanned aerial vehicle path planning based on static wind speed and wind direction.

2. The urban logistics unmanned aerial vehicle path planning method considering dynamic wind speed and direction according to claim 1, wherein (2) the model construction specifically comprises:

(2.1) wind speed and wind direction model construction

Establishing a vector speed triangle, wherein the direction vertically upwards along the y-axis in the plane is the positive direction of the system; airspeed represents the speed of flight of the unmanned aerial vehicle relative to air, abbreviated as TAS; wind speed represents the speed of movement of air relative to the ground, abbreviated WS; ground speed represents the flight speed of the unmanned aerial vehicle relative to the ground, abbreviated as GS; wind direction represents the angle defined clockwise from the positive direction to the direction of the wind, abbreviated as σ; the course angle represents the angle from the defined clockwise direction to the course line of the unmanned plane, and is abbreviated as alpha; the track angle represents the angle from the defined forward direction clockwise to the unmanned aerial vehicle track, abbreviated as beta; the drift angle represents the angle of the aircraft from the course under the influence of the crosswind, and is abbreviated as gamma, and the wind angle represents the included angle between the course line and the wind direction line, and is abbreviated as theta;

by usingRepresenting n+1 nodes in the system consisting of 1 warehouse and n client points, where c ⁰ Represents a warehouse point, C' =c\ { C } ⁰ -representing a set of client points; then for any starting point c in the system ⁱ And target point c ^j Let the coordinates of two nodes be (a) _i ,b _i ) And (a) _j ,b _j ) Then from c ⁱ Pointing to c ^j The calculation formula of the track angle beta is as follows:

and then the wind angle theta is obtained:

α＝β±θ (5)

equations (2) - (6) show the solving process of unknown parameters in the vector velocity triangle, including the track angle beta, the wind angle theta, the drift angle gamma, the heading angle alpha and the ground speed GS; the course angle alpha is used for eliminating the influence of drift on the flight track of the unmanned aerial vehicle, ensuring the connection line of the unmanned aerial vehicle to the target point along the starting point, namely the initial course when the course flies, and the ground speed GS is the speed which is changed for keeping flying along the course under the influence of wind speed and wind direction;

Describing the logistics distribution scene in a Cartesian plane and usingRepresenting n+1 nodes in the system consisting of 1 warehouse and n client points, where c ⁰ Represents a warehouse point, C' =c\ { C } ⁰ -representing a set of client points; every node +.>Set to {(s) ⁱ ,d ⁱ ,p ⁱ ,l ⁱ ,r ⁱ ) (s is therein ⁱ ＝(a _i ,b _i ) Representing the 2-dimensional position coordinates of node i, d ⁱ Representing delivery demand of node i, p ⁱ Representing the pick demand of node i (warehouse point c ⁰ Both delivery and pick-up demand of 0), l ⁱ And r ⁱ Then lambda is defined representing the earliest and latest acceptable service time windows at client point i, respectively ₁ And lambda (lambda) ₂ The penalty coefficient is the penalty coefficient when the actual arrival time of the unmanned aerial vehicle is earlier than the earliest service time window of the client point and later than the latest service time of the client point; setting +.>Representing the sum of the descent, ascent and hover service times of the unmanned aerial vehicle u at the client point i; the relevant parameter symbols and their interpretation are shown in table 1;

by usingRepresenting unmanned aerial vehicle set in delivery system, beta _ij For track angle, alpha, between any two nodes _ij And GS _ij Respectively maintaining the course and the ground speed under the condition of an initial track;

dispersing the dynamic change of the wind speed and the wind direction into a time window with a certain length, namely, the wind speed and the wind direction in different time windows are different, and meanwhile, the influence of the wind speed and the wind direction in the whole distribution range and the whole distribution process of the unmanned aerial vehicle is global, and the wind speed and the wind direction information of each time window is predicted in advance;

By usingRepresenting a set of time window sequences, there being different ground speeds and heading angles between different nodes and within different time window sequences, respectively defined as +.>And->The distribution service from the client point i to the client point j is carried out by three time sequences, wherein each time sequence t has corresponding wind speed and wind direction, and the affected ground speed and course angle; the time of changing the course angle in the flying process is ignored;

setting variablesRepresenting the flight time of the unmanned plane u in the time sequence t when the unmanned plane u executes the distribution service from the node i to the node j;

(1) Objective function

The objective function is constructed based on several factors:

(1.1) fixed cost

The fixed cost is related to the number of drones in the system, expressed as:

(1.2) energy consumption costs

Wherein the method comprises the steps ofm _ij For the actual load of the unmanned aerial vehicle from node i to node j, P _ij The energy consumption power of the unmanned plane from the node i to the node j is obtained; distribution distance dis between node i and node j _ij Equal to the sum of the cumulative flight distances at each time series, namely:

wherein the method comprises the steps ofIndicating whether the delivery service between the node i and the node j is performed by the unmanned aerial vehicle u;

(1.3) customer satisfaction penalty cost

Taking a soft time window [ l ] ⁱ ,r ⁱ ]Form definition of customer's acceptable earliest and latest service time windows, modeling customer satisfaction as penalty cost violating customer satisfaction, using λ ₁ And lambda (lambda) ₂ The penalty coefficient of unit time when the actual arrival time of the unmanned aerial vehicle is earlier than the earliest service time window of the client point and later than the latest service time of the client point is represented; by usingThe time that unmanned plane u arrives at node j from node i is represented, when the time that unmanned plane arrives at a client point accords with a time window of the client point, the client satisfaction punishment cost is 0, when the time that unmanned plane arrives at the client point is earlier or later than a time window acceptable by the client point, a certain punishment is received, and the satisfaction punishment of the client point is modeled as follows:

Combining the fixed cost, the energy consumption cost and the customer satisfaction penalty cost, respectively increasing the fixed cost and the energy consumption cost from the operator perspective and the customer satisfaction penalty cost from the customer perspective by a weighted value k ₁ And kappa (kappa) ₂ ，κ ₁ +κ ₂ =1; based on dynamic static wind speed and directionThe objective function of the logistics unmanned aerial vehicle path planning is as follows:

(2) Constraint conditions

(2.1) unmanned aerial vehicle mission-related constraints

Introducing a 0-1 variable

to ensure that each customer has and only one drone to perform the delivery service; by setting up:

to ensure that the drone will leave from customer point j after performing the delivery task at customer point j;

(2.2) unmanned aerial vehicle reuse related constraints

Introducing a 0-1 variable

(2.3) unmanned aerial vehicle number constraint

(2.4) unmanned aerial vehicle heavy restraint

0≤m _ij ≤Q,i∈C,j∈C.i≠j (21)

(2.5) unmanned aerial vehicle heavy restraint

wherein E is _max The maximum electric quantity of the unmanned aerial vehicle; the electric quantity consumption of the unmanned aerial vehicle in the process of executing the distribution from the node i to the node j is as follows:

3. the urban logistics unmanned aerial vehicle path planning method considering dynamic wind speed and direction according to claim 2, wherein (3) problem establishment comprises:

based on the system model, the objective is to minimize the weight values of the unmanned plane fixed cost, the energy consumption cost and the customer satisfaction punishment cost by optimizing the decision variables in the unmanned plane path planning problem model under the consideration of unmanned plane energy consumption constraint, customer time window constraint and dynamic wind speed and wind direction influence;

4. the urban logistics unmanned aerial vehicle path planning method considering dynamic wind speed and direction according to claim 1, wherein the problem solving step (4) specifically comprises;

(4.1) encoding

Coding in the form of permutation integers; in the system, n client points are shared, a certain population number of chromosomes with the length of 2n+1 are generated during decoding, and the n client points at most need to complete distribution service through 2n+1 nodes, and the number of the client points is larger than that of the warehouse points;

meanwhile, the coding form obtained after the earliest service time window acceptable by the client, the energy consumption constraint of the unmanned aerial vehicle and the weight constraint judgment of the unmanned aerial vehicle is the form of an initialized population; introducing a penalty factor for violating the model constraint conditions in the encoding process, namely carrying out quantization processing on the unmanned aerial vehicle by the penalty factor under the condition of violating the energy consumption constraint and the unmanned aerial vehicle load constraint of the unmanned aerial vehicle;

(4.2) population fitness calculation

The fitness of each chromosome in the population is calculated by a fitness function based on the coding form of the chromosome, and is used for evaluating the fitness of each chromosome, and the chromosomes are selected to carry out crossover and mutation operations according to the fitness so as to generate the next generation of chromosomes, and the chromosomes with high fitness are more likely to be selected and reserved, so that the fitness of the whole population is gradually improved;

(4.3) crossover and mutation

The crossover is used for generating a next generation of candidate solutions, the mutation is to avoid sinking into a local optimal solution, and a new searching direction is provided, and through the two operations, the GA can realize exploration and optimization in a solution space;

(4.4) Large Scale neighborhood search

The LNS adjusts the solution in the neighborhood of the feasible solution mainly through the methods of destruction and repair, so as to obtain a high-quality solution; the "destroy" and "repair" operations are implemented by removing operators and inserting operators, respectively;

for the "destroyed" part, the solution is adjusted by adopting a similarity removing operator, wherein the operator firstly randomly selects one client i, removes the client i from the existing distribution scheme, and then calculates the similarity measure R of the rest clients and the client i in the scheme _ij Removing the clients with the maximum similarity, and repeating the above operations until the clients with the set number are removed; r is R _ij The calculation formula is as follows:

wherein the distance dis from node i to node j _ij Embody the difference of distance, |l ⁱ -l ^j I represents the difference in the earliest service time window acceptable, |p ⁱ +d ⁱ -p ^j -d ^j The I represents the difference of the delivery demand and the picking demand of the customer point, and all three factors influence the measurement of the similarity, and eta is introduced in consideration of normalization ₁ 、η ₂ And eta ₃ A weight representing each influencing factor;

the dynamic wind speed and direction can enable the unmanned aerial vehicle to have dynamic change of ground speed when distributed among different nodes and different time window sequencesCourse angle changed with it>The dynamic programming is used in the decoding process of the GA-LNS, and the time for the unmanned aerial vehicle to reach each point in the distribution sequence under the dynamic wind speed and wind direction is solved; wherein i represents the node sequence label in the delivery sequence of the unmanned plane u, the unmanned plane sequentially accesses the nodes corresponding to the label, the loading and unloading of cargoes and the replacement of batteries are completed at a warehouse point, the picking and delivering service of cargoes is completed at a client point, t represents the current time window sequence, and different ground speeds and course angles corresponding to the unmanned plane when providing delivery service for different adjacent nodes in different time window sequences are represented by GS and alpha; for a time window sequence with time spans of ζ, it is divided into elapsed times t _used And unused time t _rest By t _start And t _end Representing a time window sequence in which delivery starts and a time window sequence in which delivery completes, and passing through a distance dis of neighboring nodes _i-1,i AndDetermining a time window sequence t in which delivery is completed _end And further obtaining a time point when the unmanned aerial vehicle reaches the node i, and obtaining a set arr of the arrival time of each node through judging and accessing the nodes in the distribution sequence of the unmanned aerial vehicle u.