Method for optimizing collaborative delivery path of heterogeneous system under road network and energy consumption constraints

By establishing a collaborative delivery path model and optimization algorithm for heterogeneous systems, the problem of low delivery efficiency in unmanned vehicles and unmanned aerial vehicles is solved, and continuous delivery of multiple delivery addresses is achieved, reducing the total time and energy consumption .

CN119047673APending Publication Date: 2024-11-29DALIAN MARITIME UNIVERSITY
View PDF 0 Cites 7 Cited by

Patent Information

Application Number
CN202411184572.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-27
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

The existing heterogeneous collaborative delivery system composed of drones/drones cannot comprehensively consider the drone path during the delivery process and the movement of drones in the road network, resulting in the inability to ensure sufficient drones' power. The continuous delivery of multiple delivery addresses is performed, increasing the total time and energy consumption of heterogeneous system delivery.

Method used

A method of co-delivery path optimization of heterogeneous systems under road network and energy consumption constraints is proposed. By establishing a coordinated delivery path distribution area model of the heterogeneous system of drone/unmanned vehicle, a double-layer optimized hybrid integer planning model is constructed, and the use of Improve the sparrow search algorithm and hybrid genetic algorithm to optimize the delivery sequence of drones and the stop time and location of the unmanned vehicle to ensure the shortest total delivery time.

Benefits of technology

It realizes continuous delivery of multiple delivery addresses while ensuring sufficient power of the drone, significantly reducing the total time and energy consumption of heterogeneous system delivery, and improving the efficiency of collaborative delivery.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119047673A_ABST
    Figure CN119047673A_ABST
Patent Text Reader

Abstract

The invention discloses a heterogeneous system cooperative delivery path optimization method under road network and energy consumption constraints. The method comprises the following steps: S1, establishing a cooperative delivery path distribution area model about an unmanned aerial vehicle / unmanned vehicle heterogeneous system; s2, constructing a heterogeneous system path planning model based on the collaborative delivery path distribution area model; s3, constructing a double-layer optimization mixed integer programming model according to the heterogeneous system path programming model; s4, according to the total delivery time target function, obtaining a first optimization target function of the shortest time of the unmanned aerial vehicle without considering the energy consumption limitation, and obtaining an optimal unmanned aerial vehicle delivery sequence based on an improved sparrow search algorithm; and S5, according to the total delivery time target function, obtaining a second optimization target function considering the total time of the unmanned aerial vehicle under the energy consumption limitation, and optimizing the parking time and the parking position of the unmanned vehicle under the optimal unmanned aerial vehicle delivery sequence based on a hybrid genetic algorithm to obtain an optimal heterogeneous system cooperative delivery path. The problems that an existing heterogeneous cooperative delivery system composed of unmanned vehicles / unmanned aerial vehicles cannot comprehensively consider the paths of the unmanned aerial vehicles in the delivery process and movement of the unmanned aerial vehicles in a road network, and continuous delivery of multiple delivery addresses cannot be achieved on the premise that it is guaranteed that the unmanned aerial vehicles are sufficient in electric quantity; and the total delivery time and energy consumption of the heterogeneous system are greatly increased.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the fields of geographic information processing, logistics distribution and network analysis technology, and in particular to a method for optimizing a heterogeneous system collaborative delivery path under road network and energy consumption constraints. Background Art

[0002] At present, although drones have certain delivery capabilities and are safe, there are still the following challenges in applying logistics drones to actual cargo delivery. 1. The drone’s endurance is limited. That is, the energy consumption required for drones to pick up goods from distribution centers and then deliver them to the delivery destination often exceeds the drone’s maximum endurance. Therefore, it is necessary to find a way to maximize the use of drone endurance. [1] For the trajectory planning problem of using drones for wireless energy transmission, the optimal solution is obtained by relaxing the problem in an ideal situation, and then the optimal solution and constraints are combined and an effective trajectory design is obtained by applying continuous convex programming optimization technology. Coelho [2] Considering the energy consumption constraints of drones, assuming that drones can be charged from customer locations and charging stations, a new time-based drone formation path planning problem is introduced and solved. 2. Due to the limitation of their own load capacity, drones do not have strong continuous delivery capabilities. The following studies have considered the impact of drone load on delivery capacity: Dorling et al. [3] The linear relationship between UAV energy consumption and payload is derived and verified, and applied to obtain suboptimal solutions for realistic problem sizes using a simulated annealing (SA) inspired algorithm. [4] Aiming at the limited loading capacity of UAVs and the influence of loading weight on flight capability, a step-by-step progressive task assignment (RHTA) heuristic method was developed and tested on the delivery in island areas.

[0003] In addition, there are the following studies on drone single-unit delivery: Faust [5] A reinforcement learning approach for aerial delivery of cargo in scenarios with static obstacles is studied. [6] This paper explores whether the use of drones for logistics distribution can enhance the sustainability of the delivery process. It has been proven that the use of drones to optimize delivery routes can save energy and reduce carbon emissions, which is environmentally friendly. [7] The study analyzed several existing drone delivery solutions and tested different solutions using small and large datasets. However, most of the existing research methods are aimed at a specific challenge, and it is difficult to find a good method to uniformly solve the multiple challenges faced by drone delivery.

[0004] In mainstream research, vehicles are often used for logistics delivery. [8]The application of genetic algorithm (GA) in the basic vehicle routing problem (VRP) is considered, and the genetic algorithm is mixed with the neighborhood search algorithm to verify the advantages of this method in solution time and solution quality. [9] A heterogeneous unified heuristic algorithm is proposed to solve five different variants of the vehicle routing problem: the vehicle routing problem with time windows (VRPTW), the capacity vehicle routing problem (CVRP), the multi-depot vehicle routing problem (MDVRP), the station-dependent vehicle routing problem (SDVRP), and the open vehicle routing problem (OVRP).

[10] The fuel consumption rate (FCR) is added to the classic capacity vehicle routing problem (CVRP), and a simulated annealing algorithm with hybrid exchange rules is developed to solve the proposed optimization model FCVRP.

[11] An adaptive large neighborhood search heuristic algorithm is proposed for two-echelon vehicle routing problem (2E-VRP) and location routing problem (LRP). Previous studies have shown that vehicle delivery has better endurance and carrying capacity than drones, but the vehicle is slower than drones, and in actual situations, vehicles cannot reach any location in the delivery area due to restrictions on roads and obstacles.

[0005] In order to solve the above challenges, some studies have combined drones and unmanned vehicles to form a heterogeneous system for delivery. The advantage of this is that unmanned vehicles have strong endurance and carrying capacity. Compared with the use of single drones for delivery, the combination of the two can make up for the shortcomings of drone load and endurance by using unmanned vehicles to carry drones and goods. Compared with the use of single unmanned vehicles, drones can complete delivery tasks more quickly and avoid unmanned vehicles from traveling extra miles. Chen

[12] A classification of existing unmanned aerial vehicle and vehicle systems (UAGVS) is proposed, and a generalized optimization framework is established to uniformly describe the decision-making problems of different types of UAV systems.

[13] It is proposed to introduce a dedicated team of charging robots during long-term multi-robot missions, with which drones can dock for charging. The charging robots refer to the trucks that charge the drones.

[14] For the first time, the cooperation between drones and unmanned vehicles was introduced into the field of logistics and distribution. Two mathematical programming models, namely the flying partner traveling salesman problem (FSTSP) and the parallel drone scheduling TSP (PDSTSP), were proposed, and two heuristic solution methods were proposed to solve the problem of actual size. Sacramento, D

[15] The Vehicle Routing Problem with Drones is studied, and an adaptive large neighborhood search metaheuristic method is proposed with time limit and minimum cost as the objective function.

[16] A new mixed integer programming model (MILP) is proposed, and a branch-and-price algorithm is developed to solve the drone and unmanned vehicle delivery model.

[17] The method of package delivery using delivery trucks and multiple drones or unmanned vehicles is studied. A heuristic solution consisting of three sub-problems is proposed, and the potential time savings associated with the use of multiple drones are additionally quantified.

[18] A new variant of the UAV TSP (TSP-D) is studied. The first algorithm is based on the method proposed by Murray and Chu, which is implemented by local search, and the second algorithm is the greedy randomized adaptive search process (GRASP).

[19] The study used a hybrid genetic algorithm to allow an unmanned vehicle to carry multiple drones, which could deliver packages to customers in different locations at the same time.

[20] Using drones as repeaters in vehicle networks, a new data transmission scheme named UAV-Assisted Data Transmission (UADD) is proposed to assist data transmission.

[21] It demonstrates the extent to which a delivery system in which drones deliver packages to customers, unmanned vehicles carry goods and drones can improve the quality of logistics services.

[0006] This study starts from the actual scenario. In real cities, unmanned vehicles do not travel to arbitrary locations, but only travel on main roads. Therefore, the movement of unmanned vehicles is restricted to the urban road network. A heterogeneous collaborative delivery system consisting of unmanned vehicles and drones is designed to solve the classic "last mile problem" in logistics planning problems. The goal is to minimize the total delivery time. The unmanned vehicle carries the drone from the starting point to various areas. The movement of the unmanned vehicle is restricted to the road network, and the drone delivers the goods to the addresses of each customer. During the drone delivery, the unmanned vehicle also needs to find a suitable location to dock and meet the drone and charge it. After the delivery is completed, the drone is recovered, and the two go to the destination or to the next distribution area to continue delivery. This study was influenced by Chen

[22] Inspired by the research, the unmanned vehicle is also constrained in the road network. The drone unloads the package from the unmanned vehicle and delivers it to the customer. Chen considered the location of the charging point during the delivery process, but his research has some limitations. For example, the drone will only deliver to one customer each time it is released. When the customers are densely distributed, the drone will be repeatedly released and recovered in a short period of time, which greatly affects the delivery efficiency.

[0007] In summary, the existing heterogeneous collaborative delivery system composed of unmanned vehicles / drones cannot comprehensively consider the drone paths during the delivery process and the movement of the unmanned vehicles in the road network. It is impossible to achieve continuous delivery to multiple delivery addresses while ensuring that the drones have sufficient power, which greatly increases the total delivery time and energy consumption of the heterogeneous system. Summary of the invention

[0008] The present invention provides a method for optimizing the collaborative delivery path of heterogeneous systems under road network and energy consumption constraints to overcome the above technical problems.

[0009] In order to achieve the above object, the technical solution of the present invention is:

[0010] A method for optimizing a delivery path of a heterogeneous system under road network and energy consumption constraints specifically comprises the following steps:

[0011] S1: Establish a delivery area model for the coordinated delivery path of UAV / AV heterogeneous systems;

[0012] S2: Based on the collaborative delivery path distribution area model, a heterogeneous system path planning model is constructed;

[0013] And the heterogeneous system path planning model includes a UAV path planning model under energy consumption constraints and a UGV path planning model under road network constraints;

[0014] S3: Based on the heterogeneous system path planning model, a two-level optimization mixed integer programming model is constructed with the goal of minimizing the total collaborative delivery time of the UAV / AV heterogeneous system;

[0015] The two-level optimization mixed integer programming model includes a total delivery time objective function and corresponding constraints;

[0016] S4: Obtain the first optimization objective function of the shortest UAV delivery time without considering energy consumption constraints based on the total delivery time objective function, and obtain the optimal UAV delivery sequence based on the improved sparrow search algorithm;

[0017] S5: Based on the total delivery time objective function, the second optimization objective function of the total UAV time under energy consumption constraints is obtained, and the parking time and parking position of the unmanned vehicle are optimized under the optimal UAV delivery sequence based on the hybrid genetic algorithm to obtain the optimal heterogeneous system collaborative delivery path.

[0018] Furthermore, the distribution area model for the collaborative delivery path of the UAV / AV heterogeneous system established in S1 includes the road network, customer point locations, and stop points;

[0019] The road network is represented by an undirected network graph G=(V,E);

[0020] Wherein, V represents the numbered set of road network nodes, and V = {1, 2, ..., h}; E represents the set of edges between each road network node, that is, the driving path of the unmanned vehicle;

[0021] The customer point location is a sequence set C defined in an undirected graph, and C = {c1, c2, ..., c n};

[0022] The docking point is the location of the unmanned vehicle when releasing the drone, charging the drone, and recovering the drone during the delivery process;

[0023] And define the stop point set as P, and P = {p0, p i ,p m+1}, where p0 represents the position of the unmanned vehicle when the drone is released; p m+1 Indicates the position of the unmanned vehicle when recovering the drone; p i represents the position of the unmanned vehicle when charging the drone, and i = 1, 2, 3, …, m.

[0024] Furthermore, the UAV path planning model under the energy consumption constraint constructed in S2 is:

[0025]

[0026] Where: represents the total distance traveled by the drone during its two stops; d ij Represents the customer point location c i With the customer point position c j The distance between Indicates that the drone is moving from the customer point position c i Go to the k+1th stop p k+1 distance; Indicates that the drone is moving from the customer point position c j To the kth stop p k The distance; α ij ,β ij Indicates that the drone flies to the customer point location c i , about to go to customer point location c j , and

[0027]

[0028] The UGV path planning model under the road network constraints is:

[0029]

[0030] Where: represents the total distance travelled by the driverless car during two stops; D ij Represents the distance between network node i and network node j; Indicates that the unmanned vehicle stops at the kth stop p k The distance to the jth road network node; Indicates that the unmanned vehicle stops at the k+1th stop point p k+1 The distance to the i-th road network node; V k represents the sequence number set of the road network nodes that the unmanned vehicle passes through during the kth to k+1th stops, and V k ∈V;x ij ,γ ij represents the constraints that exist when the unmanned vehicle drives to the i-th road network node and is about to go to the j-th road network node, and

[0031]

[0032] Furthermore, the two-level optimization mixed integer programming model constructed in S3 includes the total delivery time objective function and the corresponding constraints;

[0033] And the total delivery time objective function f is

[0034]

[0035] Where: v a Indicates the speed of the drone; v g represents the speed of the driverless car; Indicates the distance traveled by the unmanned vehicle from the starting point to the first stop; Indicates the distance the unmanned vehicle travels from the last stop to the end point;

[0036] The constraints are

[0037]

[0038] Where: J represents the energy consumption of the drone per unit time; e k represents the total energy consumed by the UAV from the kth stop to the k+1th stop; e max Indicates the battery energy of the drone when it is fully charged; M,M - Represents a subset of the stop point set P.

[0039] Furthermore, the S4 specifically includes the following steps:

[0040] S41: Obtaining a first optimization objective function for the shortest delivery time of the UAV without considering energy consumption constraints according to the total delivery time objective function;

[0041] The expression of the first optimization objective function f′ is:

[0042]

[0043] S42: define the sparrow population and initialize it using the roulette wheel selection method;

[0044] The parameters of the sparrow search algorithm are determined as follows: the maximum number of iterations, the total number of individuals in the sparrow population, the ratio of producers, beggars and alarmists, the maximum number of iterations, and the positions and coding of all sparrow individuals;

[0045] The individual positions of the sparrows indicate the order of drone delivery;

[0046] S43: taking the first optimization objective function as the fitness function of the sparrow search algorithm, obtaining the individual fitness value of each individual position of the sparrow in the initial sparrow population, so as to obtain the initial global optimal solution according to the individual fitness value;

[0047] S44: updating the initial sparrow population through the position update formula of producers, beggars and alarms to obtain the current sparrow population;

[0048] The producer position update formula is:

[0049]

[0050] Where: represents the customer point number delivered by the i-th sparrow in the g-th generation at the j-th delivery, that is, the drone delivery order; α represents a random number and α∈(0,1]; iter max represents the maximum number of iterations; R2 represents the warning value and R2∈(0,1]; ST represents the safety threshold and ST∈[0.5,1.0]; Q represents a random number that follows a normal distribution; L represents an m-dimensional row vector whose elements are all 1;

[0051] The beggar's position update formula is:

[0052]

[0053] Where: represents the current global worst solution; n represents the number of sparrow individuals; Represents the optimal position occupied by the producer; A + represents an m-dimensional row vector in which each element is randomly assigned 1 or -1, and A + =A T (AA T )-1; A represents the set matrix;

[0054] The position update formula of the alarm is:

[0055]

[0056] Where: represents the current global optimal position; β represents a normal distribution of random numbers with a mean of 0 and a variance of 1; K represents a random value and K∈[-1,1]; f i Indicates the current sparrow's fitness value; f w represents the worst fitness value; ε represents a minimum constant to avoid division by zero error; f g Represents the current global optimal fitness value;

[0057] S45: performing a sparrow aggregation operation on the current sparrow population through a sparrow aggregation position update formula to optimize the current sparrow population;

[0058] And the sparrow gathering position update formula is:

[0059]

[0060] Where: S best represents the sequence corresponding to the optimal solution sparrow of the current sparrow population; Sa represents the sequence corresponding to sparrow a and its fitness is f(S a );S best represents the sequence corresponding to the sparrow with the best fitness; change(a,b) represents the position change sequence between the sequence corresponding to sparrow a and the sequence corresponding to sparrow b, and change(a,b)=S a -1 S b ; change(a,little) means swapping two random positions in the sparrow a sequence; r means a random number

[0061] S46: Obtain the individual fitness value of the individual position of the sparrow in the current sparrow population after optimization to obtain the current global optimal solution;

[0062] and comparing the current global optimal solution with the initial global optimal solution, confirming and obtaining the current optimal global optimal solution;

[0063] S47: Determine whether the maximum number of iterations has been reached;

[0064] If so, stop the iterative optimization of the individual sparrow positions, and the current optimal global optimal solution is the optimal drone delivery order;

[0065] Otherwise, the currently best global optimal solution is used as the initial global optimal solution, and S44 to S46 are repeatedly executed until the maximum number of iterations is reached.

[0066] Furthermore, the S5 specifically includes the following steps:

[0067] S51: Obtaining a second optimization objective function of the total time of the drone under energy consumption constraints according to the total delivery time objective function;

[0068] The expression of the second optimization objective function f″ is:

[0069] f″=1 / f

[0070] S52: Define the initial individual population of the hybrid genetic algorithm and initialize the initial individual population to I according to the optimal drone delivery order. GA ×(n-1) size matrix G;

[0071] Among them I GA represents the number of individuals in the hybrid genetic algorithm; n is the number of customer points;

[0072] And when G ij That is, when the i-th row and j-th column of matrix G are 1, it means that the drone will make a stop between customer point i and customer point j; ij That is, when the i-th row and j-th column of the matrix G are 0, no docking is performed;

[0073] Then, each individual in the initial individual population of the hybrid genetic algorithm is defined as a solution for the parking time and parking position of the UAV;

[0074] S53: taking the second optimization objective function as the fitness function of the hybrid genetic algorithm, obtaining the fitness values ​​of individuals of various populations in the initial individual population, so as to obtain the initial individual optimal solution and the initial global optimal solution according to the fitness values ​​of individuals of various populations;

[0075] S53: performing selection operation, crossover operation and mutation operation on the initial individual population in sequence according to the initial individual optimal solution to obtain the latest individual population;

[0076] And obtain the fitness values ​​of each population in the latest individual population to obtain the latest individual optimal solution and the latest global optimal solution;

[0077] S54: Compare the latest global optimal solution with the initial global optimal solution, confirm and obtain the current global optimal solution;

[0078] S55: Determine whether the maximum preset number of iterations has been reached;

[0079] If so, stop the iterative optimization of the population individuals, and the current global optimal solution is the optimal drone’s docking time and docking position to obtain the optimal heterogeneous system collaborative delivery path;

[0080] Otherwise, the current global optimal solution and the latest individual optimal solution are used as the initial individual optimal solution and the initial global optimal solution, and S53 to S54 are repeatedly executed until the maximum preset number of iterations is reached.

[0081] Beneficial effects: The present invention provides a method for optimizing the collaborative delivery path of heterogeneous systems under the constraints of road network and energy consumption. A collaborative delivery path distribution area model is established based on a heterogeneous system composed of unmanned vehicles and unmanned aerial vehicles to obtain a heterogeneous system path planning model for planning the delivery paths of unmanned vehicles and unmanned aerial vehicles by solving the delivery sequence and the stop time point to minimize the total delivery time. To this end, a heuristic solution method based on the layer-by-layer optimization method is constructed, that is, the optimal unmanned aerial vehicle delivery sequence is obtained based on the improved sparrow search algorithm, and the stop time and stop position of the unmanned vehicle under the optimal unmanned aerial vehicle delivery sequence are optimized based on the hybrid genetic algorithm to obtain the optimal heterogeneous system collaborative delivery path. By comprehensively considering the unmanned aerial vehicle path during the delivery process and the movement of the unmanned vehicle in the road network, continuous delivery to multiple delivery addresses is achieved while ensuring that the unmanned aerial vehicle has sufficient power, which greatly reduces the total delivery time and energy consumption of the heterogeneous system and improves the efficiency of collaborative delivery. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0083] Figure 1 It is a flow chart of the method for optimizing the collaborative delivery path of heterogeneous systems under the road network and energy consumption constraints of the present invention;

[0084] Figure 2 This is a schematic diagram of the distribution area model of the collaborative delivery path in this embodiment;

[0085] Figure 3 This is the overall flow chart of the CDPOA-E algorithm in this embodiment;

[0086] Figure 4 This is a schematic diagram of road network discretization in this embodiment;

[0087] Figure 5 This is a schematic diagram of stop point selection in this embodiment;

[0088] Figure 6 Schematic diagram of the main road network within the Third Ring Road of Beijing in this embodiment;

[0089] Figure 7 In this embodiment, according to Figure 5 The obtained road network is modeled and the customer points are marked on the network diagram;

[0090] Figure 8It is the fitness curve diagram of the ISSA algorithm and the traditional algorithm in this embodiment;

[0091] Fig. 9 It is a fitness curve diagram of the HGA algorithm and the traditional algorithm in this embodiment;

[0092] Fig.10 This is a schematic diagram of the shortest path of the single-release single-service algorithm in this embodiment;

[0093] Fig.11 This is a schematic diagram of the shortest path optimized by the CDPOA-E algorithm in this embodiment. DETAILED DESCRIPTION

[0094] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0095] This embodiment provides a method for optimizing the delivery path of heterogeneous systems under road network and energy consumption constraints. Figure 1 As shown, the specific steps include:

[0096] S1: Establish a delivery area model for the coordinated delivery path of UAV / AV heterogeneous systems;

[0097] Specifically, the collaborative delivery path distribution area model includes the road network, customer point locations, and stop points;

[0098] The road network is represented by an undirected network graph G=(V,E);

[0099] Wherein, V represents the numbered set of road network nodes, and V = {1, 2, ..., h}; E represents the set of edges between each road network node, that is, the driving path of the unmanned vehicle;

[0100] The customer point location is a sequence set C defined in an undirected graph, and C = {c1, c2, ..., c n};

[0101] The docking point is the location of the unmanned vehicle when releasing the drone, charging the drone, and recovering the drone during the delivery process;

[0102] And define the stop point set as P, and P = {p0, p i ,p m+1}, where p0 represents the position of the unmanned vehicle when the drone is released; p m+1Indicates the position of the unmanned vehicle when recovering the drone; p i represents the position of the unmanned vehicle when charging the drone, and i = 1, 2, 3, …, m.

[0103] Figure 2 It shows the process of delivering to customer points in a delivery area. There are four customer points c1, c2, c3, and c4 in the area that need to be delivered. Figure 2 The lower left corner is the starting point, and the upper right corner is the end point. The blue arrow indicates the driving path of the unmanned vehicle, the orange arrow indicates the flight path of the drone, the red dot indicates the release point of the drone, the green dot indicates the charging point where the two meet, and the yellow dot indicates the drone receiving point, where the unmanned vehicle receives the drone for charging and releases it;

[0104] S2: Based on the collaborative delivery path distribution area model, a heterogeneous system path planning model is constructed;

[0105] And the heterogeneous system path planning model includes a UAV path planning model under energy consumption constraints and a UGV path planning model under road network constraints;

[0106] Specifically, the sequence number set of the drone’s delivery to the customer point is defined as C k , C k ∈C; the total distance travelled by the drone between two stops It consists of the following three parts: the distance from the first customer point to the kth stop, C k The distance between each customer point in the delivery order and the distance from the last customer point to the k+1th stop point, the UAV path planning model under the energy consumption constraint is constructed as follows:

[0107]

[0108] Where: represents the total distance traveled by the drone during its two stops; d ij Represents the customer point location c i With the customer point position c j The distance between Indicates that the drone is moving from the customer point position c i Go to the k+1th stop p k+1 distance; Indicates that the drone is moving from the customer point position c j To the kth stop p k The distance; α ij ,β ij Indicates that the drone flies to the customer point location c i , about to go to customer point location c j , and

[0109]

[0110] where α ij Indicates whether the current drone is from c i Customer point flies to the next c in the planning sequence j If the decision variable is satisfied, it takes 1, and if it is not satisfied, it takes 0; β ij Indicates that the current drone is from c i The customer point No. flies to the corresponding docking point to charge, and then goes to the next customer point c in the planning sequence j If the value is 1, otherwise it is 0. i Customer No. Fly to C j There are only two situations in the process of customer point: 1) When the power is sufficient, complete the c j 2) When the battery is low, it needs to be docked for charging before completing the delivery to the c j Delivery to customer point.

[0111] In this embodiment, the sequence number set of the road network nodes that the unmanned vehicle passes through during the kth to k+1th stops is defined as V k , V k ∈V, the total distance travelled by the driverless car between two stops It consists of the following three parts: the distance from the kth stop on the shortest path to the first road network node, the distance from the kth stop to the last road network node on the shortest path, and the distance from the last road network node to the k+1th stop. The UGV path planning model under the road network constraint is:

[0112]

[0113] Where: represents the total distance travelled by the driverless car during two stops; D ij Represents the distance between network node i and network node j; Indicates that the unmanned vehicle stops at the kth stop p k The distance to the jth road network node; Indicates that the unmanned vehicle stops at the k+1th stop point p k+1 The distance to the i-th road network node; V k represents the sequence number set of the road network nodes that the unmanned vehicle passes through during the kth to k+1th stops, and V k ∈V; χ ij ,γ ij represents the constraints that exist when the unmanned vehicle drives to the i-th road network node and is about to go to the j-th road network node, and

[0114]

[0115] Among them, χij If the edge between the i-th road network node and the j-th road network node is included in the shortest path of the unmanned vehicle, it takes 1, otherwise it takes 0; γ ij It takes 1 when the unmanned vehicle stops from the i-th network node to the j-th network node, otherwise it takes 0. Constraint (8) means that between two stops, the unmanned vehicle visits each network node on the shortest path only once, and should decide whether to stop before going to the next network node;

[0116] S3: Based on the heterogeneous system path planning model, a two-level optimization mixed integer programming model is constructed with the goal of minimizing the total collaborative delivery time of the UAV / AV heterogeneous system;

[0117] The two-level optimization mixed integer programming model includes a total delivery time objective function and corresponding constraints;

[0118] Specifically, the total delivery time objective function f obtained based on equation (1) and equation (5) is:

[0119]

[0120] Where: v a Indicates the speed of the drone; v g represents the speed of the driverless car; Indicates the distance traveled by the unmanned vehicle from the starting point to the first stop; Indicates the distance the unmanned vehicle travels from the last stop to the end point;

[0121] The constraints are

[0122]

[0123] Where: J represents the energy consumption of the drone per unit time; e k represents the total energy consumed by the UAV from the kth stop to the k+1th stop; e max Indicates the battery energy of the drone when it is fully charged; M,M - represents a subset of the stop point set P; X li represents a 0-1 decision variable for the unmanned vehicle to go from road number l to road number i, indicating that the unmanned vehicle will not start from point i and return to point i from another point during the two stops; the objective function (9) aims to minimize the total delivery time; in constraint (10), e maxrepresents the battery energy of the drone when it is fully charged. This constraint means that the drone’s energy consumption between two stops should not exceed its maximum energy value. Constraint (11) means that the shortest path taken by the unmanned vehicle between two stops should not have a loop. Constraint (12) is an ideal equality constraint, which describes that the time taken by the drone and the unmanned vehicle to reach the next stop from the previous stop is the same. This constraint can avoid the time the two spend waiting for each other at the stop.

[0124] S4: Obtain the first optimization objective function of the shortest UAV delivery time without considering energy consumption constraints based on the total delivery time objective function, and obtain the optimal UAV delivery sequence based on the improved sparrow search algorithm;

[0125] like Figure 3 As shown in the figure, the Sparrow Search Algorithm (SSA) in this embodiment is a relatively new meta-heuristic algorithm. It has a strong global search capability due to its diverse search strategies when facing complex optimization problems. The algorithm is inspired by the foraging and anti-predation process of sparrow populations to optimize the position in the solution space. Therefore, this study selects the improved Sparrow Search Algorithm to solve the first-level problem of drone delivery order, and maps it into the drone delivery order by sorting the solutions.

[0126] The specific steps include:

[0127] S41: Obtaining a first optimization objective function for the shortest delivery time of the UAV without considering energy consumption constraints according to the total delivery time objective function;

[0128] The expression of the first optimization objective function f′ is:

[0129]

[0130] The difference between this embodiment and the original objective function formula (9) is that formula (9) solves the total time of the coordinated delivery of drones and unmanned vehicles, while formula (13) converts the objective function into the shortest time of drones without considering energy consumption constraints, that is, the delivery order is solved first;

[0131] S42: define the sparrow population and initialize it using the roulette wheel selection method;

[0132] The parameters of the sparrow search algorithm are determined as follows: the maximum number of iterations, the total number of individuals in the sparrow population, the ratio of producers, beggars and alarmists, the maximum number of iterations, and the positions and coding of all sparrow individuals;

[0133] The individual positions of the sparrows indicate the order of drone delivery;

[0134] Specifically, the sparrow search algorithm divides the sparrow population into producers and scroungers during the solution process. After the overall position of the population is updated, it is assumed that a proportion of sparrows in the population is aware of the danger, and these scouters are randomly generated in the population. In the problem of optimizing the delivery path of heterogeneous systems under the constraints of road network and energy consumption, the individuals in the sparrow population represent the delivery order of the drone, that is, the path of the drone visiting each customer point in a certain order:

[0135] Producer: Analogously, it is the drone delivery path that plays a leading role in the entire solution process. These paths can often approach the optimal solution earlier and represent a better delivery order.

[0136] Scrounger: It is analogous to the drone delivery path that improves its own path sequence by learning the producer's path information during the solution process. These paths may have poor results in the early stage, but through continuous iteration and learning, they may gradually approach the optimal solution;

[0137] Scouter: It is analogous to being responsible for exploring new path options during the search process. In particular, when the solution of the current path may fall into a local optimum, it introduces a new path sequence or adjusts the existing path to jump out of the local optimum and avoid the algorithm from converging prematurely.

[0138] In this embodiment, in order to make the sparrow search algorithm easier to solve this problem, an improved sparrow search algorithm (ISSA) is proposed, and the following two improvements are made to the original algorithm: 1) the population initialization uses the roulette selection method; 2) after a round of population update of the original sparrow algorithm, the sparrow aggregation behavior is added to improve the convergence accuracy of the solution;

[0139] Roulette wheel selection: Roulette wheel selection is an initialization method inspired by the random selection process in nature. Its steps are as follows:

[0140] Roulette initialization improves the initial solution of the sparrow search algorithm:

[0141] When initializing the first customer point in the delivery sequence, it should be as close to the starting point of the unmanned vehicle as possible. Therefore, the point on the road network closest to each customer point should be considered as a possible release point. After measuring the distance, the release point closest to the starting point of the unmanned vehicle should be selected to release the drone.

[0142] Calculating the selection probability: In this embodiment, the selection probability of the roulette selection method is calculated based on the distance between each customer point. For the current customer point cur and another customer point i, the selection probability P(i) is calculated based on the following formula:

[0143]

[0144] Where: d cur,i Indicates the distance between the current customer point cur and customer point i. If customer point i has been visited, it will be infinite.

[0145] Normalized probability: In order to make the sum of all undelivered customer point probabilities equal to 1, the selection probability of each customer point is normalized based on the following formula:

[0146]

[0147] Cumulative probability: Calculate the cumulative probability C(i) for each customer point before selecting the next customer point:

[0148]

[0149] Customer point selection: Generate a random number r between 0 and 1, and determine the next customer point to visit by selecting the first cumulative probability C(i) greater than r;

[0150] The advantage of using the roulette wheel selection method to initialize the population in this embodiment is that it can construct a solution with high effectiveness at the beginning of the algorithm, avoiding unnecessary waste of resources when solving the problem. At the same time, it also retains a certain degree of randomness, so that the population as a whole has a high diversity, so that the position where the drone is released and the first delivery are close to the starting point of the unmanned vehicle, making the solution more inclined to the optimal;

[0151] Sparrow gathering behavior: Sparrows in nature have spontaneous gathering behavior when facing danger. After the early warning person finds the natural enemy and moves, other sparrows will imitate its behavior and gather to the safer position in the center of the current group. This embodiment adds the behavior of sparrow gathering after the early warning person position is updated in the original algorithm. Similar to the early warning person position update strategy, sparrow gathering occurs on 10%-20% of the sparrow individuals in the entire population. The difference is that gathering in this study is represented by directly exchanging the customer points in the delivery sequence;

[0152] S43: taking the first optimization objective function as the fitness function of the sparrow search algorithm, obtaining the individual fitness value of each individual position of the sparrow in the initial sparrow population, so as to obtain the initial global optimal solution according to the individual fitness value;

[0153] S44: updating the initial sparrow population through the position update formula of producers, beggars and alarms to obtain the current sparrow population;

[0154] The producer position update formula is:

[0155]

[0156] Where: represents the customer point number delivered by the i-th sparrow in the g-th generation at the j-th delivery, that is, the drone delivery order; α represents a random number and α∈(0,1]; iter max represents the maximum number of iterations; R2 represents the warning value and R2∈(0,1]; ST represents the safety threshold and ST∈[0.5,1.0]; Q represents a random number that follows a normal distribution; L represents an m-dimensional row vector whose elements are all 1;

[0157] The beggar's position update formula is:

[0158]

[0159] Where: represents the current global worst solution; n represents the number of sparrow individuals; Represents the optimal position occupied by the producer; A + represents an m-dimensional row vector in which each element is randomly assigned 1 or -1, and A + =A T (AA T )-1; A represents the set matrix;

[0160] The position update formula of the alarm is:

[0161]

[0162] Where: represents the current global optimal position; β represents a normal distribution of random numbers with a mean of 0 and a variance of 1; K represents a random value and K∈[-1,1]; f i Indicates the current sparrow's fitness value; f w represents the worst fitness value; ε represents a minimum constant to avoid division by zero error; f g Represents the current global optimal fitness value;

[0163] S45: performing a sparrow aggregation operation on the current sparrow population through a sparrow aggregation position update formula to optimize the current sparrow population;

[0164] And the sparrow gathering position update formula is:

[0165]

[0166] Where: S best S represents the sequence corresponding to the optimal solution sparrow of the current sparrow population; a represents the current sequence corresponding to sparrow a and its fitness is f(S a );S bestrepresents the sequence corresponding to the sparrow with the best fitness; change(a,b) represents the position change sequence between the sequence corresponding to sparrow a and the sequence corresponding to sparrow b, and change(a,b)=S a -1 S b ; change(a,little) means exchanging two random positions in the sparrow a sequence; r represents a random number. In formula (20), a )>f(S best ) is performed by position update, which moves sequence a to the optimal solution by multiplying it by a random number r while preserving the original sequence order to a certain extent; in f(S a )=f(S best ) is performed by position fine-tuning, which randomly changes the two positions of the optimal solution sequence, and maintains the diversity of the population through small changes;

[0167] S46: Obtain the individual fitness value of the individual position of the sparrow in the current sparrow population after optimization to obtain the current global optimal solution;

[0168] and comparing the current global optimal solution with the initial global optimal solution, that is, confirming that the individual position of the sparrow corresponding to the individual fitness value with the larger individual fitness value is the current optimal global optimal solution;

[0169] S47: Determine whether the maximum number of iterations has been reached;

[0170] If so, stop the iterative optimization of the individual sparrow positions, and the current optimal global optimal solution is the optimal drone delivery sequence, which minimizes the total delivery time;

[0171] Otherwise, the current best global optimal solution is used as the initial global optimal solution, and S44 to S46 are repeated until the maximum number of iterations is reached. When the algorithm ends, the final optimal delivery order and the corresponding shortest delivery time are output. This solution is obtained through multiple iterations and searches, and is the optimal solution for the entire population.

[0172] S5: Based on the total delivery time objective function, the second optimization objective function of the total UAV time under energy consumption constraints is obtained, and the parking time and parking position of the unmanned vehicle are optimized under the optimal UAV delivery sequence based on the hybrid genetic algorithm to obtain the optimal heterogeneous system collaborative delivery path.

[0173] This embodiment is based on the hybrid genetic algorithm to solve the parking time and parking position of the drone:

[0174] The docking time of the unmanned vehicle is solved by a hybrid genetic algorithm. In this embodiment, whether the drone rushes to the corresponding docking point to charge after delivering to a customer point is regarded as the docking time of the delivery process, and the impact of different docking times on the total delivery time is evaluated. Through the selection, crossover and mutation operations of the genetic algorithm, the docking time is gradually optimized so that the unmanned vehicle can dock at the most appropriate time to provide charging and replenishment for the drone, thereby reducing waiting time and energy consumption. Solving the docking time is to solve β in formula (3) ij , by ij The selection of this 0-1 decision variable determines whether the drone stops during the delivery process between two customer points. In this embodiment, a hybrid genetic algorithm (HGA) is used to solve the stop time. In the traditional genetic algorithm, there is no information exchange between different populations, and each population evolves independently. The hybrid genetic algorithm adopts the idea of ​​sharing the optimal position of the particle swarm algorithm. The advantage is that it can guide the search by sharing the global optimum, so that the whole converges to the optimal solution more quickly. It is a solution method that utilizes the idea of ​​group collaboration.

[0175] The specific steps include:

[0176] S51: Obtaining a second optimization objective function of the total time of the drone under energy consumption constraints according to the total delivery time objective function;

[0177] The expression of the second optimization objective function f″ is:

[0178] f″=1 / f (21)

[0179] In this embodiment, the fitness function of each individual in the hybrid genetic algorithm is calculated. The fitness function takes the inverse of the objective function formula (10) in the problem description. A shorter total time has a higher fitness, that is, the shorter the total time, the higher the fitness value;

[0180] S52: Define the initial individual population of the hybrid genetic algorithm and initialize the initial individual population to I according to the optimal drone delivery order. GA ×(n-1)-sized matrix G, where the rows of the matrix G represent customer points and the columns represent the corresponding docking decisions (i.e., whether the driverless vehicle docks between two customer points, 0 means no docking, and 1 means docking);

[0181] Among them I GA represents the number of individuals in the hybrid genetic algorithm; n is the number of customer points, each element in the matrix G is randomly initialized to 0 or 1; and when G ij That is, when the i-th row and j-th column of matrix G are 1, it means that the drone will make a stop between customer point i and customer point j; ij That is, when the i-th row and j-th column of the matrix G are 0, no docking is performed;

[0182] Then, each individual in the initial individual population of the hybrid genetic algorithm is defined as a solution for the parking time and parking position of the UAV;

[0183] S53: taking the second optimization objective function as the fitness function of the hybrid genetic algorithm, obtaining the fitness values ​​of individuals of various populations in the initial individual population, so as to obtain the initial individual optimal solution and the initial global optimal solution according to the fitness values ​​of individuals of various populations;

[0184] S53: performing selection operation, crossover operation and mutation operation on the initial individual population in sequence according to the initial individual optimal solution to obtain the latest individual population; wherein the selection operation, crossover operation and mutation operation are existing well-known technical means and will not be described in detail here;

[0185] And obtain the fitness values ​​of each population in the latest individual population to obtain the latest individual optimal solution and the latest global optimal solution;

[0186] S54: Compare the latest global optimal solution with the initial global optimal solution, that is, confirm that the individual position in the hybrid genetic algorithm corresponding to the individual fitness value with a larger individual fitness value is the current global optimal solution, so as to find a shorter total time;

[0187] S55: Determine whether the maximum preset number of iterations has been reached;

[0188] If so, stop the iterative optimization of the population individuals, and the current global optimal solution is the optimal drone’s docking time and docking position to obtain the optimal heterogeneous system collaborative delivery path;

[0189] Otherwise, the current global optimal solution and the latest individual optimal solution are used as the initial individual optimal solution and the initial global optimal solution, and S53 to S54 are repeatedly executed until the maximum preset number of iterations is reached;

[0190] The improvement of the hybrid genetic algorithm in this embodiment over the traditional genetic algorithm lies mainly in that it combines the global search capability of the genetic algorithm and the efficiency of the local search technology of the particle swarm algorithm. This hybrid method can more effectively explore the solution space, avoid premature convergence, and maintain a good balance in the process of finding the optimal solution. By introducing crossover and mutation operations, the algorithm can effectively jump out of the local optimum and improve the diversity and quality of the solution. In the process of solving the stop time, this embodiment also requires solving the optimal stop position corresponding to different stop times. This embodiment will discretize the road network in the urban road network and select several candidate stop points, such as Figure 4 As shown in Figure 2, by calculating the impact of the geographical location of each stop on the delivery path, the stop position of the unmanned vehicle is optimized, as shown in Figure 2. Figure 5As shown. In order to effectively calculate the shortest path of the unmanned vehicle in the road network, this embodiment adopts the Floyd-Warshall algorithm to select the point with the shortest path according to the current stop time, so that the total time for the drone and the unmanned vehicle to reach this point is the shortest. The goal is to select those stop points that can minimize the flight distance and time of the drone, thereby further shortening the total delivery time. Finally, through the comprehensive optimization of the stop time and stop position, the hybrid genetic algorithm can find the best stop strategy for the unmanned vehicle under the determined delivery order. This strategy not only takes into account the energy consumption constraints of the unmanned vehicle and the drone, but also optimizes the overall delivery path to ensure the shortest total delivery time.

[0191] Experimental verification and comparison results of this embodiment:

[0192] This embodiment uses simulation experiments to evaluate the effectiveness of the proposed CDPOA-E algorithm (i.e., based on the improved sparrow search algorithm and the hybrid genetic algorithm). The experimental scenario is modeled by the main roads within the Fourth Ring Road of Beijing. The road network modeling process is as follows: Figure 6 to Figure 7 As shown, the data source is Google Maps, the model consists of 52 nodes and 80 paths, and there are 10 customer points distributed in the entire scene. In order to verify the effectiveness of the proposed algorithm, this embodiment adopts the following experimental settings:

[0193] 1. Delivery area: The delivery area is constructed through a randomly generated undirected graph, which contains 10 road network nodes, 20 customer points and 5 charging points.

[0194] 2. Parameters of UAV and unmanned vehicle: The maximum flight time of UAV is 0.5h and the speed is 54km / h; the maximum speed of unmanned vehicle is 18km / h.

[0195] 3. Algorithm parameters and experimental environment: Improve the number of iterations in the sparrow search algorithm iter1 max =300, the population size is m1=500, and the number of iterations in the hybrid genetic algorithm is iter2 max = 200, the population size is m2 = 200, and the probability of generating an early warning person in the population is 10%. The proposed algorithm was verified in MATLAB, and all experiments were run on a computer with an Intel(R) Core(TM) i5-10400 CPU and 16.0GB of running memory under Windows 11.

[0196] 4. Comparison of algorithms: This embodiment compares the proposed CDPOA-E algorithm with the following two algorithms: Comparison 1: Comparison of the effectiveness of each layer algorithm in the CDPOA-E algorithm before and after improvement;

[0197] Comparison 2: Compare the proposed CDPOA-E algorithm with the algorithm of Chen

[22] ;

[0198] like Figure 8 The figure shows the fitness curves of the ISSA algorithm and the traditional algorithm. It can be seen that the initial solution of the ISSA algorithm has a lower fitness. This is because the population initialization of the ISSA algorithm (Improved Sparrow Search Algorithm) uses the roulette selection method. It can also be seen that it can converge to a better solution within a few iterations. This is due to the introduction of sparrow aggregation behavior, which reflects its strong solution-solving ability. The SSA algorithm (Sparrow Search Algorithm) completes convergence in 70 generations, and the optimal solution is 1.5427h; the ISSA algorithm completes convergence in 35 generations, and the optimal solution is 1.3158h; the PSO algorithm completes convergence in 41 generations, and the optimal solution is 1.6274h; the ACO algorithm completes convergence in 121 generations, and the optimal solution is 1.6917h; the two optimization processes in the ISSA algorithm cooperate with each other to improve the overall efficiency of the algorithm. Table 1 shows the fitness value comparison table corresponding to the number of iterations of the ISSA and comparison algorithms;

[0199] Table 1. Fitness values ​​corresponding to the number of iterations of ISSA and comparison algorithms

[0200]

[0201] like Fig. 9 The figure shows the fitness curves of the HGA algorithm (Hybrid Genetic Algorithm) and the traditional algorithm. It can be seen that the HGA algorithm converges to the optimal solution faster than the traditional GA algorithm. This is because the introduction of the particle swarm algorithm shares the idea of ​​the global optimal to guide the search. The GA algorithm completed convergence in 185 generations, and the optimal solution was 1.3444h; the HGA algorithm completed convergence in 196 generations, and the optimal solution was 1.2683h; the SA algorithm completed convergence in 198 generations, and the optimal solution was 1.5330h; the quality of the optimal solution calculated by the HGA algorithm was also improved compared with the GA algorithm. It can be seen that the introduction of the global sharing process of the particle swarm effectively improves the convergence speed and solution quality of the algorithm; as shown in Table 2, there is a comparison table of fitness values ​​corresponding to the number of iterations of HGA and the comparison algorithm;

[0202] Table 2. HGA and comparison algorithm iteration times corresponding to fitness values

[0203]

[0204] like Figure 10 to Figure 11 As shown, the planning results of the single-release single-service algorithm of the CDPOA-E algorithm and the Chen

[22] algorithm in the same scenario are shown;

[0205] The shortest path for the unmanned vehicle obtained by the single-release single-service algorithm is as follows:

[0206] v1→v 13 →v 31 →v 30 →v 11 →v 10 →v9→v 26 →v 25 →v 44 →v 43 →v 42 →v 24 →v 23 →v 22 →v 21 →v 19 →v 18 →v 37

[0207] The shortest path for the unmanned vehicle obtained by the CDPOA-E algorithm is as follows:

[0208] v1→v 13 →v 12 →v 11 →v 10 →v 11 →v 12 →v1→v2→v3→v 17 →v 18 →v 37

[0209] The single-release single-service algorithm obtains the shortest total delivery time of 3.74735h, and the CDPOA-E algorithm obtains the shortest total delivery time of 1.42598h. It can be seen that the CDPOA-E algorithm is superior to the single-release single-service algorithm of Chen's

[22] algorithm in terms of total delivery time. This result is mainly due to the two key designs of the CDPOA-E algorithm:

[0210] First, the CDPOA-E algorithm allows the drone to continuously deliver to multiple customer points after a single release, while the single-release single-service algorithm can only deliver to one customer point at a time. This difference will be magnified when the customer points are densely distributed. The CDPOA-E algorithm can make full use of the drone's endurance and reduce the repeated release and recovery of the drone, thereby significantly shortening the total delivery time. In contrast, the single-release single-service algorithm requires frequent release and recovery of drones because the drone only serves one customer point at a time, resulting in a longer total delivery time;

[0211] Second, the CDPOA-E algorithm further improves the overall efficiency of the system by optimizing the delivery sequence and docking location. It uses the improved sparrow search algorithm and hybrid genetic algorithm to optimize these two aspects respectively. This sophisticated path planning enables the CDPOA-E algorithm to fully utilize the synergy between unmanned vehicles and drones. Although the single-release single-service algorithm also considers the impact of drone endurance, it is relatively simple in path planning and cannot fully utilize the potential of heterogeneous systems.

[0212] In summary, the CDPOA-E algorithm has shown excellent performance in terms of total delivery time, algorithm convergence, and energy consumption, proving the effectiveness of the proposed heterogeneous system collaborative delivery optimization method. This embodiment studies the "last mile" delivery problem in a heterogeneous collaborative logistics system under the dual constraints of road network and energy consumption; and establishes a heterogeneous system consisting of unmanned vehicles and drones, in which the unmanned vehicles are responsible for carrying drones and goods and charging drones, and the drones are responsible for delivery; in this embodiment, the selection of the delivery order to the customer address can be regarded as a traveling salesman problem (TSP), the selection of the charging point location can be regarded as a facility location problem (Facility Location Problem), and the process of waiting for each other's time window during the delivery of drones and unmanned vehicles can be regarded as a job-shop scheduling problem (Job-shop Scheduling Problem); and overall, the problem is described as a bilevel optimization mixed integer programming (Bilevel Optimization MILP) model and is NP-hard. Double-layer optimization means that the solution of the lower function is the calculation parameter of the upper function. In the context of the embodiment, the upper function is the total delivery time of the drone and the unmanned vehicle, the lower function is the time taken by the unmanned vehicle to go to each stop, and the solution of the lower layer is the timing of the drone to stop the unmanned vehicle. For this complex model, this embodiment adopts a heuristic solution method based on the layer-by-layer optimization idea. The algorithm is called CDPOA-E, where "CDP" stands for "Collaborative Delivery Path", "OA" stands for "Optimization Algorithm", and "E" stands for "Energy". As a constraint, the algorithm effectively solves the problem of actual scale within a reasonable running time; that is, the drone delivery sequence and stop time are optimized by improving the sparrow search algorithm and the hybrid genetic algorithm respectively. The simulation results show that the CDPOA-E algorithm is superior to the comparison algorithm in terms of total delivery time, algorithm convergence and energy consumption, proving the effectiveness of the heterogeneous collaborative logistics delivery optimization method.

[0213] The relevant literature involved in this embodiment is as follows:

[0214] [1] Xu J, Zeng Y, Zhang R. UAV-Enabled Wireless Power Transfer: Trajectory Design and Energy Optimization[J]. Ieee Transactions on Wireless Communications, 2018, 17(8): 5092-5106.

[0215] [2] Coelho B N, Coelho V N, Coelho I M, et al. A multi-objective green UAV routing problem[J]. Computers&Operations Research, 2017, 88: 306-315.

[0216] [3] Dorling K, Heinrichs J, Messier G G, et al. Vehicle Routing Problems for Drone Delivery[J]. Ieee Transactions on Systems Man Cybernetics-Systems, 2017, 47(1): 70-85.

[0217] [4] Song B D, Park K, Kim J. Persistent UAV delivery logistics: MILP formulation and efficient heuristic[J]. Computers&Industrial Engineering, 2018, 120: 418-428.

[0218] [5] Faust A, Palunko I, Cruz P, et al. Automated aerial suspended cargo delivery through reinforcement learning[J]. Artificial Intelligence, 2017, 247: 381-398.

[0219] [6]Chiang W C,Li Y Y,Shang J,et al.Impact of drone delivery onsustainability and cost:Realizing the UAV potential through vehicle routingoptimization[J].Applied Energy,2019,242:1164-1175.

[0220] [7]Kuru K,Ansell D,Khan W,et al.Analysis and Optimization of unmannedAerial Vechicle in Swarms in Logistics:An Intelligent Delivery Platform[J].Ieee Access,2019,7:15804-15831.

[0221] [8]Baker B M,Ayechew M A.A genetic algorithm for the vehicle routingproblem[J].Computers&Operations Research,2003,30(5):787-800.

[0222] [9]Pisinger D,Ropke S.A general heuristic for vehicle routingproblems[J].Computers&Operations Research,2007,34(8):2403-2435.

[0223]

[10] Xiao Y Y,Zhao Q H,Kaku I,et al.Development of a fuel consumptionoptimization model for the capacitated vehicle routing problem[J].Computers&Operations Research,2012,39(7):1419-1431.

[0224]

[11] Hemmelmayr V C,Cordeau J F,Crainic T G.An adaptive largeneighborhood search heuristic for Two-Echelon Vehicle Routing Problemsarising in city logistics[J].Computers&Operations Research,2012,39(12):3215-3228.

[0225]

[12] Chen J,Zhang X,Xin B,et al.Coordination Between Unmanned Aerialand Ground Vehicles:A Taxonomy and Optimization Perspective[J].IeeeTransactions on Cybernetics,2016,46(4):959-972.

[0226]

[13] Mathew N,Smith S L,Waslander S L.Multirobot Rendezvous Planningfor Recharging in Persistent Tasks[J].Ieee Transactions on Robotics,2015,31(1):128-142.

[0227]

[14] Murray C C,Chu A G.The flying sidekick traveling salesmanproblem:Optimization ofdrone-assisted parcel delivery[J].TransportationResearch Part C:Emerging Technologies,2015,54:86-109.

[0228]

[15] Sacramento D,Pisinger D,Ropke S.An adaptive large neighborhoodsearch metaheuristic for the vehicle routing problem with drones[J].Transportation Research Part C-Emerging Technologies,2019,102:289-315.

[0229]

[16] Wang Z,Sheu J B.Vehicle routing problem with drones[J].Transportation Research Part B-Methodological,2019,122:350-364.

[0230]

[17] Murray C C,Raj R.The multiple flying sidekicks traveling salesmanproblem:Parcel delivery with multipledrones[J].Transportation Research PartC-Emerging Technologies,2020,110:368-398.

[0231]

[18] Ha Q M,Deville Y,Pham Q D,et al.On the min-cost TravelingSalesman Problem with Drone[J].Transportation Research Part C-EmergingTechnologies,2018,86:597-621.

[0232]

[19] Peng K,Du J X,Lu F,et al.A Hybrid Genetic Algorithm on Routingand Scheduling for Vehicle-Assisted Multi-Drone Parcel Delivery[J].IeeeAccess,2019,7:49191-49200.

[0233]

[20] Li W, Liu XX, Ma XY, et al. UAV-Aided Data Delivery Scheme Based on Opportunistic Virtual Intersections for Smart Transportation Networks[J]. Journal of Advanced Transportation, 2019, 2019:

[0234]

[21] Carlsson JG,Song S Y.Coordinated Logistics with a Truck and aDrone[J].Management Science,2018,64(9):4052-4069.

[0235]

[22] Chen Y, Chen MQ, Chen ZH, et al. Delivery path planning of heterogeneous robot system under road network constraints [J]. Computers&Electrical Engineering, 2021,92:

[0236] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing the collaborative delivery path of heterogeneous systems under road network and energy consumption constraints, characterized in that: The specific steps include: S1: Establish a delivery area model for the coordinated delivery path of UAV / AV heterogeneous systems; S2: Based on the collaborative delivery path distribution area model, a heterogeneous system path planning model is constructed; And the heterogeneous system path planning model includes a UAV path planning model under energy consumption constraints and a UGV path planning model under road network constraints; S3: Based on the heterogeneous system path planning model, a two-level optimization mixed integer programming model is constructed with the goal of minimizing the total collaborative delivery time of the UAV / AV heterogeneous system; The two-level optimization mixed integer programming model includes a total delivery time objective function and corresponding constraints; S4: Obtain the first optimization objective function of the shortest UAV delivery time without considering energy consumption constraints based on the total delivery time objective function, and obtain the optimal UAV delivery sequence based on the improved sparrow search algorithm; S5: According to the total delivery time objective function, the second optimization objective function of the total UAV time under energy consumption constraints is obtained, and the parking time and parking position of the unmanned vehicle are optimized under the optimal UAV delivery sequence based on the hybrid genetic algorithm to obtain the optimal heterogeneous system collaborative delivery path.

2. According to the method for optimizing the collaborative delivery path of heterogeneous systems under road network and energy consumption constraints in claim 1, it is characterized in that: The distribution area model for the collaborative delivery path of the UAV / AV heterogeneous system established in S1 includes the road network, customer point locations, and stop points; The road network is represented by an undirected network graph G=(V,E); Wherein, V represents the numbered set of road network nodes, and V = {1, 2, ..., h}; E represents the set of edges between each road network node, that is, the driving path of the unmanned vehicle; The customer point location is a sequence set C defined in an undirected graph, and C = {c1, c2, ..., c n }; The docking point is the location of the unmanned vehicle when releasing the drone, charging the drone, and recovering the drone during the delivery process; And define the stop point set as P, and P = {p0, p i ,p m+1 }, where p0 represents the position of the unmanned vehicle when the drone is released; p m+1 Indicates the position of the unmanned vehicle when recovering the drone; p i represents the position of the unmanned vehicle when charging the drone, and i = 1, 2, 3, …, m.

3. According to the method of optimizing the delivery path of heterogeneous systems under the constraints of road network and energy consumption according to claim 1, it is characterized in that: The UAV path planning model under the energy consumption constraint constructed in S2 is: Where: represents the total distance traveled by the drone during its two stops; d ij Represents the customer point location c i With the customer point position c j The distance between Indicates that the drone is moving from the client point position c i Go to the k+1th stop p k+1 distance; Indicates that the drone is moving from the client point position c j To the kth stop p k The distance; α ij ,β ij Indicates that the drone flies to the customer point location c i , about to go to customer point location c j , and The UGV path planning model under the road network constraints is: Where: represents the total distance travelled by the driverless car during two stops; D ij Represents the distance between network node i and network node j; Indicates that the unmanned vehicle stops at the kth stop p k The distance to the jth road network node; Indicates that the unmanned vehicle stops at the k+1th stop point p k+1 The distance to the i-th road network node; V k represents the sequence number set of the road network nodes that the unmanned vehicle passes through during the kth to k+1th stops, and V k ∈V; χ ij ,γ ij represents the constraints that exist when the unmanned vehicle drives to the i-th road network node and is about to go to the j-th road network node, and 4. According to the method for optimizing the delivery path of heterogeneous systems under road network and energy consumption constraints in claim 1, it is characterized in that: The two-level optimization mixed integer programming model constructed in S3 includes the total delivery time objective function and the corresponding constraints; And the total delivery time objective function f is Where: v a Indicates the speed of the drone; v g represents the speed of the driverless car; Indicates the distance traveled by the unmanned vehicle from the starting point to the first stop; Indicates the distance the unmanned vehicle travels from the last stop to the end point; The constraints are Where: J represents the energy consumption of the drone per unit time; e k represents the total energy consumed by the UAV from the kth stop to the k+1th stop; e max Indicates the battery energy of the drone when it is fully charged; M,M - represents a subset of the stop point set P.

5. According to the method of optimizing the delivery path of heterogeneous systems under the constraints of road network and energy consumption according to claim 1, it is characterized in that: The S4 specifically comprises the following steps: S41: Obtaining a first optimization objective function for the shortest delivery time of the UAV without considering energy consumption constraints according to the total delivery time objective function; The expression of the first optimization objective function f′ is: S42: define the sparrow population and initialize it using the roulette wheel selection method; The parameters of the sparrow search algorithm are determined as follows: the maximum number of iterations, the total number of individuals in the sparrow population, the ratio of producers, beggars and alarmists, the maximum number of iterations, and the positions and coding of all sparrow individuals; The individual positions of the sparrows indicate the order of drone delivery; S43: taking the first optimization objective function as the fitness function of the sparrow search algorithm, obtaining the individual fitness value of each individual position of the sparrow in the initial sparrow population, so as to obtain the initial global optimal solution according to the individual fitness value; S44: updating the initial sparrow population through the position update formula of producers, beggars and alarms to obtain the current sparrow population; The producer position update formula is: Where: represents the customer point number delivered by the i-th sparrow in the g-th generation at the j-th delivery, that is, the drone delivery order; α represents a random number and α∈(0,1]; iter max represents the maximum number of iterations; R2 represents the warning value and R2∈(0,1]; ST represents the safety threshold and ST∈[0.5,1.0]; Q represents a random number that follows a normal distribution; L represents an m-dimensional row vector whose elements are all 1; The beggar's position update formula is: Where: represents the current global worst solution; n represents the number of sparrow individuals; Represents the optimal position occupied by the producer; A + represents an m-dimensional row vector in which each element is randomly assigned 1 or -1, and A + =A T (AA T )-1; A represents the set matrix; The position update formula of the alarm is: Where: represents the current global optimal position; β represents a normal distribution of random numbers with a mean of 0 and a variance of 1; K represents a random value and K∈[-1,1]; f i Indicates the current sparrow's fitness value; f w represents the worst fitness value; ε represents a minimum constant to avoid division by zero error; f g Represents the current global optimal fitness value; S45: performing a sparrow aggregation operation on the current sparrow population through a sparrow aggregation position update formula to optimize the current sparrow population; And the sparrow gathering position update formula is: Where: S best S represents the sequence corresponding to the optimal solution sparrow of the current sparrow population; a represents the current sequence corresponding to sparrow a and its fitness is f(S a );S best represents the sequence corresponding to the sparrow with the best fitness; change(a,b) represents the position change sequence between the sequence corresponding to sparrow a and the sequence corresponding to sparrow b, and change(a,little) means swapping two random positions in the sparrow a sequence; r represents a random number S46: Obtain the individual fitness value of the individual position of the sparrow in the current sparrow population after optimization to obtain the current global optimal solution; and comparing the current global optimal solution with the initial global optimal solution, confirming and obtaining the current optimal global optimal solution; S47: Determine whether the maximum number of iterations has been reached; If so, stop the iterative optimization of the individual sparrow positions, and the current optimal global optimal solution is the optimal drone delivery order; Otherwise, the currently best global optimal solution is used as the initial global optimal solution, and S44 to S46 are repeatedly executed until the maximum number of iterations is reached.

6. The method for optimizing the delivery path of heterogeneous systems under road network and energy consumption constraints according to claim 1, characterized in that: The S5 specifically includes the following steps: S51: Obtaining a second optimization objective function of the total time of the drone under energy consumption constraints according to the total delivery time objective function; The expression of the second optimization objective function f″ is: f″=1 / f S52: Define the initial individual population of the hybrid genetic algorithm and initialize the initial individual population to I according to the optimal drone delivery order. GA ×(n-1) size matrix G; Among them I GA represents the number of individuals in the hybrid genetic algorithm; n is the number of customer points; And when G ij That is, when the i-th row and j-th column of matrix G are 1, it means that the drone will make a stop between customer point i and customer point j; ij That is, when the i-th row and j-th column of the matrix G are 0, no docking is performed; Then, each individual in the initial individual population of the hybrid genetic algorithm is defined as a solution for the parking time and parking position of the UAV; S53: taking the second optimization objective function as the fitness function of the hybrid genetic algorithm, obtaining the fitness values ​​of individuals of various populations in the initial individual population, so as to obtain the initial individual optimal solution and the initial global optimal solution according to the fitness values ​​of individuals of various populations; S53: performing selection operation, crossover operation and mutation operation on the initial individual population in sequence according to the initial individual optimal solution to obtain the latest individual population; And obtain the fitness values ​​of each population in the latest individual population to obtain the latest individual optimal solution and the latest global optimal solution; S54: Compare the latest global optimal solution with the initial global optimal solution, confirm and obtain the current global optimal solution; S55: Determine whether the maximum preset number of iterations has been reached; If so, stop the iterative optimization of the population individuals, and the current global optimal solution is the optimal drone’s docking time and docking position to obtain the optimal heterogeneous system collaborative delivery path; Otherwise, the current global optimal solution and the latest individual optimal solution are used as the initial individual optimal solution and the initial global optimal solution, and S53 to S54 are repeatedly executed until the maximum preset number of iterations is reached.

Citation Information

Cited By

  • Multi-warehouse distribution task path planning method and system

    CN119783933A

  • Truck and unmanned aerial vehicle collaborative distribution method based on hybrid eagle-variable neighborhood search

    CN119919036A

  • Unmanned aerial vehicle Dubins path planning method based on piecewise linear mixed integer programming

    CN120085681A

  • Unmanned Aerial Vehicle Dubins Path Planning Method Based on Piecewise Linear Mixed Integer Programming

    CN120085681B

  • Aircarrier path online planning method for air-land combined transportation task

    CN120252728A