A Solving Algorithm for Optimizing the Drone Delivery Network for Urban Instant Delivery

Abstract

The present invention discloses a solution algorithm for optimizing the drone delivery network for urban instant delivery, including the following processes: 1) Establishing a mixed-integer programming model, defining the state parameters and variables in drone delivery, determining the objective function of the drone delivery network optimization model, and the model needs to satisfy certain constraint conditions; 2) Establishing a set partitioning model; 3) Using a construction heuristic algorithm based on large neighborhood search to iteratively optimize the initial feasible path of drone delivery. In this invention patent, the problem of the drone delivery network considering the battery swapping and load scenarios is first modeled as a mixed-integer programming problem, and then this problem is rewritten as a set partitioning model. And a construction algorithm and a large neighborhood search algorithm are proposed to obtain an effective initial solution to this problem, and continuous optimization is carried out based on this initial solution to obtain the optimal solution of the drone delivery network, improving the efficiency of urban instant delivery.

Description

Technical Field

[0001] The present invention relates to a solution algorithm for optimizing the drone delivery network for urban instant delivery Technical Background

[0002] In recent years, with the continuous development of Internet technology, domestic e-commerce and online business in China have flourished. In the development of business, the logistics industry has also begun to further develop. An efficient logistics model has attracted much attention, and how to improve the efficiency of the existing delivery network is the key development point of major logistics companies at present. With the continuous development of online business, platform orders have emerged explosively. Facing the explosive order volume, traditional logistics delivery obviously cannot meet the requirements of urban instant delivery. Efficient urban instant delivery requires delivering goods to customers within 30 minutes. Facing such a fast delivery mode, traditional manual delivery not only cannot meet the time requirements, but also has huge labor costs and extremely poor employee safety. Therefore, drone delivery has begun to enter the logistics industry. Drones have inherent advantages: high-speed flight, vertical takeoff and landing, safe and reliable, all-weather operation, etc. Drone delivery is mainly applied to the urban instant delivery industry, and tens of thousands of delivery orders are generated every day in the city. Drone delivery can be unrestricted by ground traffic and can perform straight-line delivery, so it plays an important role in urban instant delivery

[0003] When Amazon's PrimeAir drone made its first appearance and test flight at the Emerging Technology Conference in 2013, the logistics model of drone delivery began to gradually develop. By December 2016, when drones were first used to deliver goods, the logistics model of drone delivery began to gradually enter the path of modern urban development. Compared with foreign countries, the development of domestic drone delivery technology lags behind. The history of domestic drone research and development is relatively short, and it is mainly used in the military. However, with the continuous deepening of the integration of military and civilian, commercial drones have also developed rapidly in China. Many large domestic enterprises have begun to research the core technologies related to drone delivery, including well-known enterprises such as JD.com, SF Express, Cainiao Network, and DJI

[0004] Although using drone delivery can effectively solve the traffic constraint problem of delivery, when the order volume is large enough, if the entire delivery network is not modeled and optimized. Then even using drones for delivery cannot meet the requirements of urban instant delivery. In the construction of smart cities, the number of orders is increasing day by day, and it is very necessary to optimize the overall delivery network. The present invention is based on the situation of a sharp increase in the order volume, and how to model and design an optimization algorithm for the drone delivery network. So as to make the efficiency of drone delivery reach the optimum, reduce the overall delivery cost, realize urban instant delivery, and create a modern delivery mode Summary of the Invention

[0005] In view of the above technical problems existing in the prior art, the purpose of the present invention is to provide a solution algorithm for optimizing the drone delivery network for urban instant delivery.

[0006] The solution algorithm for optimizing the drone delivery network for urban instant delivery includes the following processes:

[0007] Step 1: Establishment of a mixed integer programming model

[0008] 3.1 Symbol definition:

[0009] K = {1, 2, 3, …, k}: Set of drone numbers;

[0010] P = {k + 1, k + 2, …, k + n}: Set of drone pick-up nodes;

[0011] D = {k + n + 1, k + n + 2, …, k + 2n}: Set of drone delivery nodes;

[0012] K' = {1, 2, 3, …, k}: Set of initial position nodes of drones;

[0013] S = {k + 2n + 1}: Set of final summary nodes of drones;

[0014] N = {K', P, D, S}: Set of all nodes;

[0015] A = {(i, j)|i ∈ N\{k + 2n + 1}, j ∈ N\K', i ≠ j}: Node connection arc;

[0016] G = (N, A): Node graph;

[0017] q i : Load requirement of drone network node i;

[0018] d i : Service time of drone network node i;

[0019] [a i ,b i : Service time window of node i;

[0020] Q: Maximum load of drone (kg);

[0021] W: Unloaded take-off weight of drone (kg);

[0022] v = 10m / s: Flight speed of drone;

[0023] c ij : Flight cost of drone from node i to node j;

[0024] t ij : Flight time (s) of the UAV from node i to node j;

[0025] Δt: Time (s) consumed by the UAV for automatic battery replacement;

[0026] σ: Full charge energy (kwh) of the UAV lithium battery;

[0027] α: Energy density (kw / kg) of the UAV lithium battery;

[0028] M: A very large positive integer (e.g., M > 100000);

[0029] 3.2 Variables:

[0030] If UAV k flies from node i to node j, then Otherwise

[0031] z i : If the UAV performs battery replacement operation at node i, then z i = 1, otherwise z i = 0;

[0032] Time point when UAV k arrives at node i;

[0033] Q i : Payload when the UAV leaves node i;

[0034] Accumulated power consumption of the UAV when it arrives at node i;

[0035] Accumulated power consumption of the UAV when it leaves node i;

[0036] 3.3 Model objective function:

[0037] In this modeling, the system will have a time to accept orders. During this time, the main work of the system is to accept randomly generated orders. Each order will have a service time window [a i , b i , where a i is the earliest time for the UAV to arrive at node i, and b i is the latest time for the UAV to arrive at node i. According to the actual operation situation, every customer hopes that there will be a UAV for delivery service as soon as they place an order. Therefore, the earliest arrival time window a i for each node is set to 0, that is, a i= 0. According to the problem characteristics, there is a set of pick-up nodes P and a set of delivery nodes D in the UAV delivery network. There is a flight time t from each pick-up node p ∈ P to the delivery node d ∈ D ij , and it is necessary to ensure that the corresponding pick-up node i and delivery node i + n are served by the same UAV, otherwise there will be a chaotic phenomenon. For the entire UAV delivery network, what needs to be optimized is to minimize the total time for all UAVs to complete all orders, so that the UAVs can evenly serve each customer during the delivery and minimize the overall service time.

[0038] Objective function expression:

[0039]

[0040] The objective function in this design and modeling is to minimize the sum of the times for all UAVs to complete orders. Designing the objective function in this way ensures that the algorithm will not focus only on a single order during the solution process. The algorithm will evenly treat each order to be delivered during the solution process, minimizing the overall delivery time.

[0041] 3.4 Constraints:

[0042] In the solution algorithm model for optimizing the UAV delivery network for urban instant delivery designed in this invention patent, the constraints in actual UAV delivery are introduced. Compared with traditional delivery network optimization solution algorithms and modeling methods, this model is more in line with the actual delivery network and delivery conditions.

[0043] First is the in-flow and out-flow constraints for each node. For the UAV initial position nodes (i ∈ K'), there are no in-flow arcs in these nodes, only out-flow arcs, and the out-flow nodes cannot belong to the delivery nodes, but can only be pick-up nodes P or the end point S. Thus, the following flow constraint formula (4) is obtained:

[0044]

[0045] In this constraint, we design each node i = k. Because in the network node graph, the generated UAV initial node positions and UAV numbers are in one-to-one correspondence, so it is necessary to make i = k. This is also an aspect where this model differs from traditional flow constraints. Doing so can reduce the difficulty of model solution.

[0046] In addition to the flow constraints for the initial node of the drone, the pick-up nodes, delivery nodes, and their endpoints all require flow constraints. For the endpoint, only inflow is allowed and no outflow, so in the node graph, there are only arcs flowing towards the endpoint and no arcs flowing out from the endpoint. For the pick-up nodes and delivery nodes, the flow balance constraint must be satisfied, that is, the inflowing arcs are equal to the outflowing arcs, and it is necessary to ensure that the pick-up nodes corresponding to pick-up must be visited by the same drone before the delivery nodes. Because in the actual drone delivery network, an order of a customer must be served by the same drone and must satisfy the requirement of picking up the goods first and then delivering them. Therefore, more constraints need to be satisfied for the nodes in the pick-up set P and the delivery set D. The specific constraints are shown in the following formula (2-6):

[0047]

[0048] Expression (2) indicates that each customer's service is provided by exactly one drone;

[0049] Expression (3) indicates that a group of pick-up nodes and delivery nodes must be served by the same drone;

[0050] Expression (4) indicates that each drone starts from only one initial position node;

[0051] Expression (5) indicates that the flow balance constraint must be satisfied for pick-up nodes and delivery nodes;

[0052] Expression (6) indicates that each drone must finally return to the endpoint (i.e., the final aggregation node of the drone).

[0053] In this invention patent, time windows are considered during drone delivery. The drone must complete the visit to the current node within the time window range of each node, and when the drone needs to perform a battery replacement operation at the current node, an additional automatic battery replacement time Δt needs to be added. In the model, the time Δt for replacing the battery of the drone is designed as a constant. And the completion time of all orders must be within the promised delivery time. In addition, the time logic between nodes must also be reflected through constraints. In addition to time constraints, the drone needs to continuously pick up and deliver goods at network nodes, so the drone has a corresponding load at each node. At the pick-up node, the load of the drone increases by q i , i ∈ P. When the drone arrives at the delivery node, the load of the drone decreases by q i , i ∈ D. The maximum take-off weight of the drone must also be restricted. According to the actual situation in drone delivery, for example, the empty weight of the drone is about 15 kilograms and the maximum effective load is 3 kilograms. Therefore, when each drone is flying, the total weight of the goods in the cabin cannot exceed the maximum effective load of the drone, which is 3 kilograms.

[0054] After considering the time and load of the UAV in the model, it is necessary to impose constraints on battery swapping. Since when the UAV arrives at a node, it is necessary to consider whether the remaining available power of the current UAV can meet the flight requirements for the next leg of the journey. If the power required for the next flight leg of the UAV is greater than the current available power, then the UAV needs to perform a battery swapping operation at that node. At this time, there will be a phenomenon of sudden change in node power. For example, when the UAV flies to node 2, its power consumption is 85 kWh, and the power required for the next leg of the journey is 25 kWh. And the full charge of the UAV is 100 kWh, so the UAV must perform a battery swap at node 2. In addition, the access time of each node is matched one by one with a certain UAV, through the constraint variable This variable represents the time when the UAV numbered k leaves node i, where the node can be the initial position node, the pick-up node, the delivery node or the final convergence node. The specific constraints are shown as follows (7-20).

[0055]

[0056] Expression (7) represents the time logic constraint before and after the delivery of UAV k at the network node;

[0057] Expression (8) represents the load logic constraint of the UAV on the network node;

[0058] Expression (9) means that UAV k must first pick up goods at node i before going to the corresponding node n+i for delivery;

[0059] Expression (10) represents the power consumption constraint when the UAV arrives at the network node;

[0060] Expression (11) represents the power consumption constraint of the network node after the UAV swaps batteries;

[0061] Expression (12) represents the power consumption constraint of the network node before and after the UAV swaps batteries;

[0062] Expression (13) represents the relationship between the power consumption when the UAV leaves the node and the power consumption when it arrives at the node;

[0063] Expression (14) represents the power consumption constraint on the next network node after the UAV swaps batteries;

[0064] Expression (15) represents the power consumption constraint for all starting nodes of the UAV;

[0065] Expression (16) represents the time window constraint for UAV k to access node i;

[0066] Expression (17) represents the load constraint of the UAV on the network node;

[0067] The expression (18) represents the battery replacement and load constraints for the starting and ending nodes of the UAV.

[0068] The expressions (19)-(20) represent the types of variables.

[0069] Step 2: Establishment of the set partitioning model

[0070] 3.5 Set covering model:

[0071] After mathematically modeling the UAV delivery network optimization algorithm, the UAV delivery model can be obtained. Based on this model, a model transformation is carried out, transforming the above-mentioned mixed-integer programming model into a set partitioning model. In the set partitioning model, each solution corresponds to a feasible delivery path r of the UAV. The set partitioning model is to transform the above-mentioned mixed-integer programming model, and the meaning of the objective function of the set partitioning model must be consistent with the meaning of the objective function of the above-mentioned mixed-integer programming model.

[0072] A modeling transformation of an optimization algorithm for a UAV delivery network for urban instant delivery, characterized in that the modeling process of the set partitioning model is as follows:

[0073] 3.5.1 Symbol definition:

[0074] C r : The corresponding delivery cost of the UAV delivery path r;

[0075] Ω: The set of all UAV delivery paths;

[0076] Ω'∈Ω: The subset of UAV delivery paths;

[0077] d ij : The flight distance between nodes i and j;

[0078] a ir : Whether node i is included in the UAV delivery path r;

[0079] θ r =1: The UAV delivery path r is selected;

[0080] b ijr =1: The arc (i,j)∈A is included in the UAV delivery path r;

[0081] 3.5.2 Objective function:

[0082]

[0083] Among them: C r =∑ (i,j)∈A d ij b ijr ;

[0084] In formula (21), the objective function is to minimize the cost of all UAV delivery routes, and the cost of this delivery route corresponds to the time for all UAVs to complete order delivery in the above-mentioned mixed-integer programming model. There is a one-to-one correspondence between the two, meeting the requirements of the subsequent solution algorithm.

[0085] 3.5.3 Constraints:

[0086] In the set partitioning model, the constraints are imposed on the UAV delivery routes. When a UAV delivers goods, the generated delivery route must meet the requirements of completing the order, that is, the corresponding nodes K'∪P of the order must be included in the feasible delivery route of the UAV. The specific constraints are as follows:

[0087]

[0088] Expression (22) means that each initial position node and pick-up node must be included in a feasible route;

[0089] Expression (23) represents the type of variable;

[0090] Set partitioning model:

[0091]

[0092] Where: C r = ∑ (i,j)∈A d ij b ijr ;

[0093] 3.6 Construction algorithm:

[0094] This invention patent is a solution algorithm for optimizing the UAV delivery network. First, a mathematical model is established for this problem, and through mathematical constraints, this problem is modeled as a mixed-integer programming problem. After obtaining this mathematical model, an effective algorithm that can quickly solve this mathematical model needs to be designed. In this invention patent, a construction algorithm is mainly proposed first according to the characteristics of this problem. The factors to be considered in this construction algorithm are the effective load upper limit value of the UAV, the battery power of the UAV, the time window in the network nodes, and the total time of UAV delivery. In the construction algorithm, first, the positions of the UAVs are initialized, and each UAV needs to initialize its take-off starting point. When the take-off starting point is initialized, order allocation needs to be carried out according to the designed construction rules, which corresponds to the step-by-step allocation of order nodes in the network nodes. During the order allocation process, it needs to be allocated according to the designed construction rules. Once the design rules are violated, the corresponding order cannot be inserted into the delivery sequence of this UAV. After a UAV completes the allocation, the remaining orders are allocated one by one until all orders are completed.

[0095] Symbol Definition:

[0096] V = {K, P, D, S}: Set of nodes;

[0097] L: Promised delivery time;

[0098] R: List of UAV delivery routes;

[0099] K = {1, 2, 3, …, k}: Set of UAV numbers;

[0100] P = {k + 1, k + 2, …, k + n}: Set of UAV pick-up points;

[0101] D = {k + n + 1, k + n + 2, …, k + 2n}: Set of UAV delivery nodes;

[0102] K' = {1, 2, 3, …, k}: Set of initial UAV position nodes;

[0103] S = {k + 2n + 1}: Set of UAV final summary nodes;

[0104] N = {K', P, D, S}: Set of all nodes;

[0105] A = {(i, j)|i ∈ N\{k + 2n + 1}, j ∈ N\K', i ≠ j}: Node connection arc;

[0106] G = (N, A): Node graph;

[0107] [a i , b i : Service time window of node i;

[0108] d ij : Flight distance of UAV from node i to node j;

[0109] t ij : Flight time (s) of UAV from node i to node j;

[0110] Δt: Time (s) consumed by UAV for automatic battery replacement;

[0111] σ: Full charge energy (kwh) of UAV lithium battery;

[0112] In the construction algorithm, the input is the node set V = {K, P, D, S}, and the output is the initial feasible path R. The construction algorithm mainly obtains the initial delivery path of the UAV under the conditions of meeting the UAV capacity constraint, time window constraint, and power constraint. It is set that the order of the UAV accessing network nodes is to pick up goods first and then go to the corresponding delivery node to complete the order delivery. After completing this order, the UAV can carry out the delivery of the next order. During the order insertion process, orders are first assigned to the UAV according to the distance, and the nearest order is assigned according to the following formula.

[0113]

[0114] The UAV will choose the order closest to it for insertion. During the order insertion, the order will be inserted successively according to the corresponding pick-up node and delivery node of the order. It is necessary to check whether the inserted order meets the time window of the UAV when inserting. When the inserted order cannot be delivered within the agreed time, the order cannot be inserted into the delivery list R[i] of the UAV. After the order is inserted, the time window is divided into the following 4 cases:

[0115] ①

[0116] ②

[0117] ③

[0118] ④

[0119] In case ①, the order node j meets the requirements of the time window, and after inserting this order, the delivery of the order can still be completed within the promised delivery time L. At this time, the order j is inserted into the delivery list of the UAV k.

[0120] In ②, since the remaining power of the UAV is insufficient, the UAV changes the battery at node i. At this time, a battery change time Δt needs to be added to the delivery time of the UAV. If the order j still meets the requirements of the time window after insertion, then the order j can be inserted into the UAV list.

[0121] In ③, since the remaining power of the UAV at node j is insufficient, the UAV changes the battery at node j. At this time, a battery change time Δt needs to be added to the delivery time of the UAV. If the order j still meets the requirements of the time window after insertion, then the order j can be inserted into the UAV list.

[0122] In ④, since the remaining battery power of the UAV at nodes i and j is insufficient, battery swapping is required at both nodes i and j. At this time, a battery swapping time of 2Δt needs to be added. However, if order j can meet the time window requirement after being inserted at this time, then order j can be inserted into the UAV's delivery list. When the UAV traverses all orders, if any order cannot meet the time window requirement after being inserted, then the UAV for this flight has completed the order allocation. At this time, the algorithm proceeds to allocate orders for the next UAV until all orders are allocated to the UAVs and the algorithm ends. At this time, the algorithm outputs the delivery path list of the UAVs, and this list contains the delivery information of each UAV. The pseudo-code for constructing the algorithm is as follows Figure 1 as shown

[0123] In the construction algorithm, the way to obtain the UAV delivery path is to select the newly inserted order that is closest to the initial position of the UAV or the position where the UAV completed the previous order under the conditions of meeting the time window, battery power, and other constraints. The newly inserted order needs to meet the requirements of the time window, mainly including that the delivery time of the new order must be less than or equal to the time window of this order after insertion. In addition, it is also necessary to ensure that inserting this order will not affect all the previous orders of this UAV, and it must be guaranteed that the UAV can complete the delivery of all orders within the promised time L after inserting the new order.

[0124] Step 3: Use the construction heuristic algorithm based on large neighborhood search to iteratively optimize the initial feasible path of UAV delivery

[0125] 3.7 Heuristic optimization algorithm:

[0126] In this invention patent, a construction heuristic algorithm based on large neighborhood search is proposed. This heuristic algorithm is mainly used to iteratively optimize the initial feasible path of the UAV obtained by the construction heuristic algorithm. In the above UAV initial feasible solution construction algorithm, an initial feasible solution of the UAV can be obtained through this construction algorithm. This feasible solution contains the initial delivery path and delivery order of each UAV. Although this feasible solution meets the constraints such as the time window and battery power, it is not the optimal feasible solution. At this time, the time for the UAV to complete the delivery is still relatively long. Although there is an optimization effect on the overall UAV logistics distribution network, the quality of the solution is still not very good, and a heuristic algorithm needs to be used for further optimization and improvement. Using the construction heuristic algorithm based on the initial feasible solution can well improve the quality of the solution, optimize the overall UAV delivery network, and maximize the delivery efficiency.

[0127] In the large neighborhood construction heuristic algorithm, it is necessary to design the removal factor and the insertion factor of the algorithm. The process of this heuristic algorithm is based on the obtained initial feasible solution, select several orders to remove, and then re-insert the orders into the delivery path list of the drones. At this time, after insertion, there may be a situation where the time window is not satisfied or the battery constraint is not satisfied, requiring a battery replacement operation, and thus it cannot be delivered within the promised time after adding the battery replacement time. For this situation, in the algorithm designed this time, the order needs to be removed again, and then the order is inserted into the delivery path list of a random drone using a random strategy. The selection of the insertion position also uses a random strategy until the insertion of the order satisfies the time window and battery constraints of the drone.

[0128] Symbol definition:

[0129] R: Initial drone delivery path list solution;

[0130] n: Number of orders;

[0131] L: Promised delivery time;

[0132] K: Number of drones;

[0133] R': Current drone delivery path solution;

[0134] R”: List storing drone delivery path solutions;

[0135] R best : Current optimal delivery path solution;

[0136] f(R'): Target value corresponding to the current delivery path;

[0137] f best : Current optimal objective function value;

[0138] P: Sub-optimal solution acceptance probability;

[0139] P1: Random probability;

[0140] q: Removal factor;

[0141] L1: Storage list;

[0142] M: Maximum number of iterations;

[0143] Among them, the probability of accepting the sub-optimal solution is calculated through the following formula:

[0144]

[0145] When a sub-optimal solution appears, the designed heuristic algorithm does not completely reject it, but accepts the sub-optimal solution with a certain probability P. Among them, the random probability P1 is generated randomly, and the randomly generated probability P1 is compared with the acceptance probability P. When the random probability P1 ≥ P, the algorithm automatically accepts the sub-optimal solution; when the random probability P1 < P, the sub-optimal solution will not be accepted as the current solution.

[0146] In this heuristic algorithm, the selection of the removal factor q is directly related to the removal order. In the heuristic algorithm of this invention patent, the removal factor q is designed to select orders in a random way. The selected order is removed from the original delivery route, and then the removed order is randomly inserted into the delivery route in a random way. During the insertion process, the requirements of the time window and the power constraint need to be considered. When the inserted order cannot be delivered within the promised time, then the order cannot be inserted at this position. At this time, the algorithm needs to re-select a new insertion point until all the removed orders are reasonably re-inserted into the UAV delivery list, and at this time the algorithm completes one iteration. After completing the iteration, the algorithm will output the optimal solution in the iteration process, and this optimal solution is the optimal UAV delivery route.

[0147] The removal factor q is obtained by using the following formula:

[0148]

[0149] The insertion rule is obtained by using the following formula:

[0150]

[0151] Design of the algorithm termination criterion:

[0152] In the solution algorithm, it is necessary to design the termination condition of the algorithm iteration. In the heuristic algorithm, the maximum number of iterations of the algorithm is often limited. In this design, the maximum number of iterations is M. The pseudo-code of the heuristic algorithm is as Figure 2 shown.

[0153] This invention patent mainly designs a solution algorithm for optimizing the drone delivery network for urban instant delivery. With the continuous development of science and technology, the intelligent lifestyle of the Internet of Everything has gradually entered people's lives. In logistics distribution, fast, efficient, and intelligent logistics technologies have become the focus of the current smart city construction. Drone delivery is precisely the best representative of fast, efficient, and intelligent logistics. With the continuous improvement of living standards, the food delivery service has developed rapidly, and people have higher and higher requirements for the speed of food delivery. Facing the explosive order volume, drone delivery increasingly highlights its unique advantages. However, when the order volume surges, simply using drone delivery is not enough, and the optimization of the drone delivery network is also required to improve the overall delivery efficiency. The solution algorithm for optimizing the drone delivery network for urban instant delivery proposed in this invention patent is used to solve the scheduling optimization problem in the drone delivery network. There are pick-up and delivery nodes for orders in the drone delivery network. In addition, during the delivery process of the drone, the actual weight of the goods and the self-weight of the drone need to be considered, because the load has a great impact on the power consumption of the drone. In this invention patent, the main focus is on optimizing drone delivery in the urban instant delivery scenario and considering the scenario where the drone can be recharged. Each drone can complete the replacement of its own lithium battery while picking up and delivering goods, enabling the recharged drone to ensure a fully charged state and improve the delivery efficiency. In addition, the power consumption of the drone is also closely related to the flight time of the drone. In this invention patent, first, the problem of the drone delivery network considering the scenarios of recharging and load is modeled as a mixed-integer programming problem, and then this problem is rewritten as a set partitioning model. A construction algorithm and a large neighborhood search algorithm are proposed to obtain an effective initial solution for this problem, and continuous optimization is carried out based on this initial solution to obtain the optimal solution of the drone delivery network and improve the efficiency of urban instant delivery. This invention patent is different from the traditional network optimization modeling problem. In the usual delivery network, only the delivery capacity limit and time window limit are considered. In this invention patent, not only the delivery capacity limit of the drone is considered, but also the impact of the weight of the goods on the power consumption of the drone is considered. The actual power consumption of the drone will change with the weight of the goods and the flight mileage.

[0154] The decision on battery swapping needs to be automatically identified through algorithms, rather than manually detecting the remaining battery power of the battery for decision-making. Therefore, a decision variable for whether to swap the battery must be introduced during modeling. The nodes in the drone delivery network are divided into four categories. The first category is the location nodes where the initial drones are located, and the number of these nodes is determined according to the actual number of drones set in the system. If there are k drones, then there are k initial position nodes. The second category is the nodes (pickup) where the drones need to pick up goods, and these nodes require the drones to complete the pickup. The third category is the nodes (delivery) where the drones need to complete the delivery, and these nodes require the drones to deliver the picked-up goods. In the optimization of the entire drone delivery network, the drones must first pick up goods at the corresponding nodes and then go to the corresponding delivery nodes to complete the delivery. However, the drones can pick up all the goods in sequence and then go for delivery, or they can pick up one good and then deliver one good, and then continue to pick up goods and complete the delivery. The delivery efficiency and power consumption of these two methods are completely different, and which one is more efficient needs to be solved through algorithms, and then the actual routes of each drone and the positions where battery swapping is required are determined. What the algorithm finally outputs is the route of each drone and the position of battery swapping. In this invention patent, after network modeling, a delivery network can be obtained, and at this time, the algorithm does not have an initial feasible solution. At this time, the construction algorithm proposed in this invention patent is used to obtain an effective initial feasible solution, and after a certain number of iterations, the algorithm will calculate a better initial feasible solution. Then, based on this initial feasible solution, a large neighborhood search is carried out, the removal factor and insertion factor in the algorithm are designed, and then continuous iterative optimization is carried out to make the algorithm obtain the optimal solution. Output the optimal delivery routes and sequences of each drone. Description of the Drawings

[0155] Figure 1 It is a process diagram of the construction algorithm pseudocode;

[0156] Figure 2 It is a process diagram of the heuristic algorithm pseudocode;

[0157] Figure 3 It is a network diagram of drone delivery;

[0158] Figure 4 It is a route diagram of the drone delivery network for the case solved by the construction algorithm;

[0159] Figure 5 It is a schematic diagram of the delivery path of the initial solution;

[0160] Figure 6 It is a schematic diagram of removing an order;

[0161] Figure 7 It is a route diagram of the drone delivery after the order is re-inserted. Detailed Implementation Manner

[0162] For the convenience of those skilled in the art, the present invention will be further described below in combination with examples and drawings. The content mentioned in the embodiments does not limit the present invention.

[0163] An optimization solution algorithm for an unmanned aerial vehicle (UAV) delivery network for urban instant delivery has the following modeling process:

[0164] 4.1 Symbol Definition:

[0165] K = {1, 2, 3, …, k}: Set of UAV numbers;

[0166] P = {k + 1, k + 2, …, k + n}: Set of UAV pick-up points;

[0167] D = {k + n + 1, k + n + 2, …, k + 2n}: Set of UAV delivery nodes;

[0168] K' = {1, 2, 3, …, k}: Set of initial position nodes of UAVs;

[0169] S = {k + 2n + 1}: Set of final summary points of UAVs;

[0170] N = {K', P, D, S}: Set of all nodes;

[0171] A = {(i, j)|i ∈ N\{k + 2n + 1}, j ∈ N\K', i ≠ j}: Node connection arcs;

[0172] G = (N, A): Node graph;

[0173] q i : Load requirement of node i;

[0174] d i : Service time of node i;

[0175] [a i ,b i : Service time window of node i;

[0176] Q: Maximum load of UAV (kg);

[0177] W: Unloaded take-off weight of UAV (kg);

[0178] v = 10m / s: Flight speed of UAV;

[0179] c ij : Flight cost from node i to node j;

[0180] t ij : Flight time from node i to node j (s);

[0181] Δt: Time (s) consumed by the automatic battery replacement of the UAV;

[0182] σ: Full - charge energy (kwh) of the UAV lithium - battery;

[0183] α: Energy density (kw / kg) of the UAV lithium - battery;

[0184] M: A very large positive integer;

[0185] 4.2 Variables:

[0186] If UAV k flies from node i to node j, then Otherwise

[0187] z i : If the UAV performs battery replacement operation at node i, then z i = 1, otherwise z i = 0;

[0188] The arrival time of UAV k at node i;

[0189] Q i : The load capacity of the UAV leaving node i;

[0190] The cumulative power consumption of the UAV when arriving at node i;

[0191] The cumulative power consumption of the UAV when leaving node i;

[0192] The modeling and solution algorithm formula for the optimization of the UAV delivery network is as follows:

[0193]

[0194]

[0195] Expression (2) indicates that each customer's service is provided by exactly one UAV; Expression (3) indicates that a set of pick - up and delivery nodes must be served by the same UAV; Expression (4) indicates that each UAV starts from only one initial position;

[0196] Expression (5) indicates that the flow - balance constraint must be satisfied for pick - up nodes and delivery nodes;

[0197] Expression (6) indicates that each UAV must finally return to the end point;

[0198] Expression (7) indicates the time - logic constraint before and after the delivery of UAV k at the network node;

[0199] The expression (8) represents the load logical constraint on the network node;

[0200] The expression (9) indicates that the drone k must pick up goods at node i before delivering goods to the corresponding node n+i; the expression (10) represents the power consumption constraint for arriving at the network node;

[0201] The expression (11) represents the power consumption constraint of the network node after the drone changes its battery;

[0202] The expression (12) represents the power consumption constraints of the network node before and after the drone changes its battery;

[0203] The expression (13) represents the relationship between the power consumption when leaving and the power consumption when arriving;

[0204] The expression (14) represents the power consumption constraint on the next network node after the drone changes its battery;

[0205] The expression (15) represents the power consumption constraint for all the starting nodes of the drones;

[0206] The expression (16) represents the time window constraint for the drone k to visit node i;

[0207] The expression (17) represents the node load constraint;

[0208] The expression (18) represents the battery change and load constraints for the starting and ending nodes of the drone;

[0209] The expressions (19)-(20) represent the types of variables;

[0210] Construction algorithm:

[0211] Symbol definition:

[0212] V = {K, P, D, S}: Set of nodes;

[0213] L: Promised delivery time;

[0214] R: List of drone delivery routes;

[0215] K = {1, 2, 3, …, k}: Set of drone numbers;

[0216] P = {k+1, k+2, …, k+n}: Set of drone pick-up points;

[0217] D = {k+n+1, k+n+2, …, k+2n}: Set of drone delivery nodes;

[0218] K' = {1, 2, 3, …, k}: Set of initial position nodes of the drones;

[0219] S = {k + 2n + 1}: Set of the final summary points of the UAVs;

[0220] N = {K', P, D, S}: Set of all nodes;

[0221] A = {(i, j)|i ∈ N\{k + 2n + 1}, j ∈ N\K', i ≠ j}: Node connection arcs;

[0222] G = (N, A): Node graph;

[0223] [a i , b i : Service time window of node i;

[0224] d ij : Flight distance from node i to node j;

[0225] t ij : Flight time from node i to node j (s);

[0226] Δt: Time consumed for the UAV to automatically change the battery (s);

[0227] σ: Full - charge energy of the UAV's lithium battery (kwh);

[0228]

[0229] Where: C r = ∑ (i,j)∈A d ij b ijr ;

[0230] Insertion criterion:

[0231] ①

[0232] ②

[0233] ③

[0234] ④

[0235] Heuristic algorithm:

[0236] Symbol definition:

[0237] R: Initial list solution of the UAV delivery path;

[0238] n: Number of orders;

[0239] L: Promised delivery time;

[0240] K: Number of UAVs;

[0241] R': Current UAV delivery path solution;

[0242] R'': List storing UAV delivery path solutions;

[0243] R best : Current optimal delivery path solution;

[0244] f(R'): Target value corresponding to the current delivery path;

[0245] f best : Current optimal objective function value;

[0246] P: Sub - optimal solution acceptance probability;

[0247] P1: Random probability;

[0248] q: Removal factor;

[0249] L1: Storage list;

[0250] M: Maximum number of iterations;

[0251] Among them, the probability of accepting the sub - optimal solution is calculated by the following formula:

[0252]

[0253] The removal factor q is obtained by the following formula:

[0254]

[0255] The insertion rule is obtained by the following formula:

[0256]

[0257] Implementation Case 1

[0258] This Implementation Case 1 is used as an example to verify the scientificity and effectiveness of the above - mentioned solution algorithm for optimizing the UAV delivery network for urban instant delivery:

[0259] (1) The system starts to accept orders. At this time, the system can set the length of the time window by itself. In this case, the order acceptance time is set to 5 minutes. The algorithm is designed with a maximum number of iterations M = 500, the number of orders is set to n = 6, and the number of UAVs is K = 3. Then the following Figure 3 shown UAV delivery network node diagram is generated. In addition, the UAV promised delivery time L = 1800s is set this time.

[0260] Get the node set V = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 0, 11, 12, 13, 14, 15}, where {0, 1, 2} represents the initial position nodes of the UAVs; {3, 4, 5, 6, 7, 8} represents the pick-up nodes of the UAVs; {9, 10, 11, 12, 13, 14} represents the delivery nodes of the UAVs; {15} represents the rendezvous node of the UAVs; obtain the distance matrix between each node as shown in Table 1 below:

[0261] Table 1 Distribution distance matrix table between nodes

[0262] 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 0 2 1 1.2 1.3 2 2.5 1.3 1 0.8 0.6 0.5 1.5 3 2.5 0 1 / 0 2 2.1 1.8 3 2.3 1.5 0.8 1.4 5 1.3 1.5 3 4.5 0 2 / / 0 2.3 1.3 4.5 2.4 5.6 3.1 2 3 1.6 4 5 1.3 0 3 / / / 0 2.4 2.5 5 3 2 1 4 3.2 3.2 3.5 2.7 0 4 / / / / 0 3.5 4.2 2.6 5.2 3.4 4 5 1 2 4.5 0 5 / / / / / 0 2.2 1.5 1.6 1.8 3.4 2.2 2.4 2.6 2.7 0 6 / / / / / / 0 3.4 4 5 2.6 5.3 5.8 2 4 0 7 / / / / / / / 0 2.4 5.6 2.7 3.7 4.5 3.8 5.2 0 8 / / / / / / / / 0 2.8 5 3 1 4 5 0 9 / / / / / / / / / 0 1.3 1.8 1.9 2.2 2.1 0 10 / / / / / / / / / / 0 3.1 3.5 3.6 3.8 0 11 / / / / / / / / / / / 0 5 6 4 0 12 / / / / / / / / / / / / 0 2 3 0 13 / / / / / / / / / / / / / 0 1.7 0 14 / / / / / / / / / / / / / / 0 0 15 / / / / / / / / / / / / / / / 0

[0263] (2) Obtain the order information, which mainly includes the pick-up and delivery node numbers, the delivery weight q i , and the delivery time window. This

[0264] The information of six orders this time is as shown in Table 2 below:

[0265]

[0266]

[0267] (3) Construct an algorithm to obtain the initial feasible delivery route. First, initialize the initial position nodes of three UAVs, which are located at nodes 0, 1, and 2 respectively. Then, according to the designed construction algorithm, assign orders to UAV 0. By traversing the node distance matrix, it can be known that the order closest to node 0 is the pick-up node 8 of order 6. At this time, the distance d ij = d 0,8 = 1 (km). Calculate the flight time The required power for flight is At this time, the time The time required to complete the delivery of order 6 from node 8 to node 14 The power consumption is f 8,14 = 43 (kwh), and the remaining power at this time is 57 kwh. Then insert the order closest to node 14 into the delivery list R[0] of UAV 0. At this time, it is found that the closest one is the pick-up node 4 of order 3. Calculate the flight time as t 14,4 = 270 (s), and the power f 14,4 = 21.6 (kwh). The flight time required to complete the delivery of order 3 is t 4,10 = 400 (s), and the power f 4,10=34.13 (kws), so far the drone's cumulative time T = 100 + 500 + 450 + 400 = 1450 (s) < 1800 (s), the remaining power is F = σ-8-43-21.6-34.13 = -6.73 (kwh), so the drone needs to perform a battery replacement operation at node 4, so the extra battery replacement time Δt = 60 (s) is added, at this time T = 1450 + 60 = 1510 (s) <L,F=100-34.14=65.87(kwh)。于是算法继续寻找离节点10最近的订单,发现订单4的取货节点6最近,于是无人机飞行时间t 10,6 =260(s), delivery time t 6,12 =580(s), at this time it is found that if drone 0 inserts order 4, the time window will conflict and the drone cannot deliver within the promised time, so drone 0’s delivery list R[0]=[0,8,14,4,10,15].

[0268] (4) Construct an algorithm to start inserting the order of drone 1, select the order closest to drone 1, find that the pickup node 7 of order 5 is closest, and calculate the flight time t 1,7 =150(s), electric charge f 1,7 =12(kwh), delivery completion time t 7,13 =380(s), electric charge f 7,13 =31.41 (kwh). Now continue to insert the pickup node 6 of order 4 which is closest to node 13. The flight time t 13,6 =200, power f 13,6 =16(kwh). Delivery completion time t 6,12 =580(s), electric charge f 6,12 =49.5(kwh), so T=150+380+200+580=1310(S) 12 <L,F=100-12-31.41-16-49.5=-8.91(kwh),于是需要换电,此时T=150+380+200+580+Δt=1370(S)。此时无人机的剩余电量虽然可以继续插入订单但是所有订单的时间窗会超出,于是无人机1的配送列表R[1]=[1,7,13,6,12,15]。

[0269] (5) Similarly, insert the order of drone 2. After similar steps, the order of drone 2’s delivery list is calculated to be R[2] = [2, 3, 9, 5, 11, 15]. Then, the drone delivery list is obtained by constructing the algorithm:

[0270] ​R = [[0,8,14,4,10,15],[1,7,13,6,12,15],[2,3,9,5,11,15]]

[0271] (6) Optimization of the large neighborhood heuristic algorithm, setting the removal factor Then randomly select two orders for removal. The schematic diagram is as follows Figure 4 as shown. Randomly select orders 4 and 6 for removal to obtain the following Figures 5-6 removal schematic diagram as shown

[0272] (7) Then randomly insert the removed orders. During the insertion process, time window and power constraints are required, and the insertion is mainly carried out according to the following 4 rules:

[0273] ①

[0274] ②

[0275] ③

[0276] ④

[0277] Then the following Figure 3 is obtained as the case of the drone delivery network route map solved by the construction algorithm. At this time, the objective function of the current initial solution can be obtained as f(R') = 1510 + 1370 + 730 = 3610. At this time, the optimal solution is defaulted to the initial solution:

[0278] R best = R' = R[[0,8,14,4,10,15],[1,7,13,6,12,15],[2,3,9,5,11,15]]

[0279] (8) At this time, the heuristic algorithm starts to continuously iterate and optimize. During the iteration, the optimal solution and the current solution are continuously updated, and the following

[0280] probability is used to accept the sub-optimal solution:

[0281]

[0282] Compare the relationship between the randomly generated probability P1 and P. When P1 > P, the sub-sub-optimal solution will be accepted, otherwise it will not be accepted

[0283] This solution is used as the current solution. The current solution obtained is:

[0284] R' = [[0,4,10,15],[1,7,13,8,14,15],[2,3,9,5,11,6,12,15]]

[0285] At this time, the objective function is: f(R') = 4120(s) > f best = 3610(s). Thus, with a certain random generation probability P1 = 0.4, thus accept this solution as the current solution, and then

[0286] R' = [[0, 4, 10, 15], [1, 7, 13, 8, 14, 15], [2, 3, 9, 5, 11, 6, 12, 15]]

[0287] obtain the following Figure 7 inserted UAV delivery route map as shown below.

[0288] (9) Update the global optimal solution and output the list of the best UAV delivery paths. Through continuous iterative calculations, the following optimal solution is obtained:

[0289] R best = [[0, 8, 14, 4, 10, 15], [1, 7, 13, 6, 12, 15], [2, 3, 9, 5, 11, 15]]

[0290] f best = 3160(s)

[0291] The content described in this specification is only a list of the implementation forms of the inventive concept. The protection scope of the present invention should not be regarded as limited to the specific forms stated in the embodiments.

Claims

1. A solution algorithm for optimizing the drone delivery network for instant urban delivery, characterized in that It includes the following processes: Step 1. Establishment of the mixed-integer programming model 1) Based on the UAV delivery network, define the state parameters and variables in UAV delivery. The parameter symbols are defined as follows: K = {1, 2, 3, …, k}: Set of UAV numbers; P = {k + 1, k + 2, …, k + n}: Set of UAV pick-up nodes; D = {k + n + 1, k + n + 2, …, k + 2n}: Set of UAV delivery nodes; K' = {1, 2, 3, …, k}: Set of initial position nodes of UAVs; S = {k + 2n + 1}: Set of final summary nodes of UAVs; N = {K', P, D, S}: Set of all nodes; A = {(i, j)|i ∈ N\{k + 2n + 1}, j ∈ N\K', i ≠ j}: Node connection arcs; G = (N, A): Node graph; q i : Payload requirement of the UAV network node i; d i : Service time of the UAV network node i; [a i ,b i : The service time window of the UAV network node i; Q: Maximum load of UAV, kg; W: Empty take-off weight of UAV, kg; v: Flying speed of UAV; c ij : The flight cost of the UAV from node i to node j; t ij : Flight time of the UAV from node i to node j, s; Δt: Time consumed for UAV automatic battery replacement, s; σ: Full battery energy of UAV lithium battery, kwh; α: Energy density of UAV lithium battery, kw / kg; If the drone k flies from node i to node j, then Otherwise z i : If the drone performs a battery replacement operation at node i, then z i = 1, otherwise z i = 0; The time point when the drone k leaves node i; Q i : Payload of the drone leaving point i; The cumulative power consumption of the UAV when it reaches node i; The cumulative power consumption of the UAV leaving node i; 2) Determine the objective function of the UAV delivery network optimization model: In the modeling, the total time for UAVs to complete all orders is minimized, so that UAVs can serve each customer evenly during delivery and minimize the overall service time. The objective function expression is as follows: The above objective function is to minimize the sum of the times for all UAVs to complete orders. Designing the objective function in this way ensures that the algorithm will not focus only on a single order during the solution process; the algorithm will treat each order to be delivered evenly during the solution process, minimizing the overall delivery time; 3) Determine the constraints of UAV delivery: First is the inflow and outflow constraint for each node. For the initial position nodes i ∈ k' of UAVs, only outflow arcs exist in these nodes, and the outflow nodes can only be pick-up nodes P or the end point S; In addition to the flow constraint for the initial UAV nodes, pick-up nodes, delivery nodes, and their end points also need to be subject to flow constraints. For the final summary nodes of UAVs, only inflow arcs exist in these nodes; For pick-up nodes and delivery nodes, the flow balance constraint must be satisfied, that is, the inflow arcs are equal to the outflow arcs, and the pick-up node corresponding to the pick-up must be visited by the same UAV before the delivery node. Thus, in the UAV delivery network, an order for a customer is served by the same UAV, and it is ensured that the goods are picked up first and then delivered; 4) Constraint process for UAV battery replacement: First, constrain the time and load of UAV delivery: The UAV needs to complete the visit to the current node within the time window range of each node. When the UAV needs to perform battery replacement operation at the current node, the automatic battery replacement time is set to Δt, and Δt is a constant; the completion times of all orders delivered by the UAV are constrained within the promised delivery time. In addition, the time logic for UAV delivery between nodes is constrained; In addition to time constraints, drones need to continuously pick up and deliver goods at network nodes. Therefore, each drone has a corresponding load capacity at each node. When the drone arrives at a pick-up node, its load capacity will increase by q i , i ∈ P. When the drone arrives at a delivery node, its load capacity will decrease by q i , i ∈ D. Thus, the conditions for the maximum take-off weight of the drone are set, and the load logic of the drone at the network node is constrained; Subsequent battery replacement constraints: After a drone reaches a node, it is necessary to consider whether the remaining available battery power of the current drone can meet the flight requirements for the next leg of the journey. If the battery power required for the next flight leg of the drone is greater than the current available battery power, then the drone needs to perform a battery replacement operation at this node, and at this time, a phenomenon of sudden change in node battery power will occur. The modeling method for such battery replacement constraints is as follows: At each node, the power consumption for reaching the node and the power consumption for leaving the node are represented. When the drone does not need to perform a battery replacement operation at the current node, When the drone needs to perform a battery replacement operation at this node, Use to correspond to the power consumption before the current node; the access time of each node is matched one by one with a certain drone. Through the constraint variable This variable represents the time when the drone numbered k reaches node i, where the node can be any one of the initial position node, pick-up node, delivery node, and final rendezvous node; Step 2. Establishment of the set partitioning model 1) Symbol definition: C r : The corresponding delivery cost of the drone delivery route r; Ω: Set of all UAV delivery paths; Ω' ∈ Ω: Subset of UAV delivery paths; d ij : The flight distance between node i and node j; a ir : Whether node i is included in the UAV delivery path r; θ r = 1: The drone delivery path r is selected; b ijr = 1: the arc (i, j) ∈ A is included in the UAV delivery path r; 2) Determination of the objective function where: C r = ∑ (i,j)∈A d ij b ijr ; The objective function of formula (21) is to minimize the cost of all UAV delivery routes; 3) Determination of the constraint conditions When delivering goods by UAV, the generated delivery route must meet the requirements of completing the order, that is, the corresponding nodes K'∪P of the order must be included in the feasible delivery route of the UAV. The specific constraints are as follows: Expression (22) indicates that each initial position node and pick-up node must be included in a feasible route; Expression (23) indicates the type of variable; 4) Construct an algorithm to obtain the initial feasible route list for UAV delivery In the construction algorithm, first, the positions of the UAVs are initialized, and each UAV needs to initialize its take-off starting point; after the take-off starting point is initialized, the orders need to be allocated according to the designed construction rules, which corresponds to the step-by-step allocation of order nodes in the network nodes; during the order allocation process, the allocation needs to be carried out according to the designed construction rules. Once the design rules are violated, the corresponding order cannot be inserted into the delivery sequence of the UAV; after a UAV completes the allocation, the remaining orders are allocated one by one until all orders are allocated; Symbol definition: V = {K, P, D, S}: Node set; L: Promised delivery time; R: UAV delivery route list; K = {1, 2, 3, …, k}: UAV number set; P = {k + 1, k + 2, …, k + n}: UAV pick-up point set; D = {k + n + 1, k + n + 2, …, k + 2n}: UAV delivery node set; K' = {1, 2, 3, …, k}: UAV initial position node set; S = {k + 2n + 1}: UAV final summary node set; N = {K', P, D, S}: All node sets; A = {(i, j)|i ∈ N\{k + 2n + 1}, j ∈ N\K', i ≠ j}: Node connection arc; G = (N, A): Node graph; [a i ,b i : Service time window of node i; d ij : The flight distance of the UAV from node i to node j; t ij : Flight time of the UAV from node i to node j, s; Δt: Time consumed by the UAV for automatic battery replacement, s; σ: Full charge energy of the UAV lithium battery, kwh; In the construction algorithm, the input is the node set V = {K, P, D, S}, and the output is the initial feasible route R; the construction algorithm mainly obtains the initial UAV delivery route under the conditions of meeting the UAV capacity constraint, time window constraint and power constraint. It is set that the order of the UAV accessing the network nodes is to pick up the goods first and then go to the corresponding delivery node to complete the delivery of the order. After completing this order, the UAV can carry out the delivery of the next order; during the order insertion process, the order is first allocated to the UAV according to the distance, and the nearest order is allocated according to the following formula: The UAV selects the nearest order for insertion. During the order insertion, the order is inserted in sequence according to the corresponding pick-up node and delivery node of the order; it is necessary to check whether the inserted order meets the time window of the UAV when inserting: when the inserted order cannot be delivered within the agreed time, the order cannot be inserted into the delivery list of the UAV, otherwise it can be inserted into the delivery list of the UAV. After the order is inserted, the time window is divided into the following 4 cases: ① ② ③ ④ In case ①, the order node j meets the time window requirement, and the order can still be delivered within the promised delivery time L after being inserted. At this time, order j is inserted into the delivery list of drone k; In ②, since the remaining battery power of the drone is insufficient, the drone replaces the battery at node i, and a battery replacement time Δt is added to the delivery time of the drone. At this time, if order j still meets the time window requirement after being inserted, then order j is inserted into the delivery list of the drone; In ③, since the remaining battery power of the drone is insufficient at node j, the drone replaces the battery at node j, and a battery replacement time Δt is added to the delivery time of the drone. At this time, if order j still meets the time window requirement after being inserted, then order j is inserted into the delivery list of the drone; In ④, since the remaining battery power of the drone is insufficient at both nodes i and j, the drone needs to replace the battery at nodes i and j. At this time, a battery replacement time of 2Δt needs to be added to the delivery time of the drone; however, if order j meets the time window requirement after being inserted at this time, then order j can be inserted into the delivery list of the drone; When the drone traverses all orders, if any order cannot meet the time window requirement after being inserted at this time, then the drone of this flight has completed the order allocation. At this time, the algorithm proceeds to allocate orders for the next drone until all orders are allocated to the drones and the algorithm ends. At this time, the algorithm outputs the delivery path list of the drones, and this list contains the delivery information of each drone; Step 3: Iteratively optimize the initial feasible path of the drone delivery based on the large neighborhood search construction heuristic algorithm. In the large neighborhood search construction heuristic algorithm, first design the removal factor and the insertion factor of the algorithm; the process of this heuristic algorithm is based on the initial feasible path solution of the drone delivery obtained in step 2. By randomly selecting the orders to be removed, the selected orders are removed from the original delivery path, and then the removed orders are randomly inserted into the delivery path list of the drone in a random manner. During the insertion process, the requirements of the time window and the battery constraint need to be considered. When the inserted order cannot be delivered within the promised time, then this order cannot be inserted at this position. At this time, the algorithm needs to reselect a new insertion point until all the removed orders are reasonably reinserted into the delivery path list of the drone. At this time, the algorithm completes one iteration; through continuous iterative calculation, after the iteration is completed, the algorithm will output the optimal solution in the iterative process, and this optimal solution is the optimal delivery path of the drone; Symbol definition: R: Initial drone delivery path list solution; n: Number of orders; L: Promised delivery time; K: Number of drones; R': Current drone delivery path solution; R”: Store the drone delivery path solution list; R best : Current optimal delivery route solution; f(R'): Target value corresponding to the current delivery path; f best : Current optimal objective function value; P: Sub-optimal solution acceptance probability; P1: Random probability; q: Removal factor; L1: Storage list; M: Maximum number of iterations; Among them, the probability of accepting the sub-optimal solution is calculated by the following formula: When a suboptimal solution appears, the designed heuristic algorithm does not completely reject it, but accepts the suboptimal solution with a certain probability P; among them, the random probability P1 is generated randomly, and the randomly generated probability P1 is compared with the acceptance probability P. When the random probability P1 ≥ P, the algorithm automatically accepts the suboptimal solution; when the random probability P1 < P, the suboptimal solution will not be accepted as the current solution; The removal factor q is obtained using the following formula: The insertion rule is obtained using the following formula: Finally, the termination condition of the designed algorithm iteration is set, and the maximum number of iterations is set to M.

2. The solution algorithm for optimizing the drone delivery network for urban instant delivery according to claim 1, wherein In 3) of Step 1, the specific constraints for drone delivery are shown in the following formulas (2) - (6): Expression (2) indicates that each customer's service is served by exactly one drone; Expression (3) indicates that a group of pickup nodes and delivery nodes must be served by the same drone; Expression (4) indicates that each drone starts from only one initial position node; Expression (5) indicates that for pickup nodes and delivery nodes, the flow balance constraint must be satisfied, that is, the incoming arcs are equal to the outgoing arcs; Expression (6) indicates that each drone must finally return to the drone final aggregation node.

3. The solution algorithm for optimizing the drone delivery network for urban instant delivery according to claim 1, wherein In 4) of Step 1, the constraints for battery swapping are shown in the following formulas (7) - (20): Expression (7) indicates the time logic constraint before and after the delivery of drone k at the network node; Expression (8) indicates the load logic constraint on the drone network node; Expression (9) indicates that drone k must first pick up goods at node i before going to deliver goods to the corresponding node n + i; Expression (10) indicates the power consumption constraint for the drone to reach the network node; Expression (11) indicates the power consumption constraint for the network node after the drone swaps batteries; Expression (12) indicates the power consumption constraint for the network node before and after the drone swaps batteries; Expression (13) indicates the relationship between the power consumption when the drone leaves the node and the power consumption when it arrives at the node; Expression (14) indicates the power consumption constraint for the next network node after the drone swaps batteries; Expression (15) indicates the power consumption constraint for all drone starting nodes; Expression (16) indicates the time window constraint for drone k to visit node i; Expression (17) indicates the load constraint on the drone network node; Expression (18) indicates the battery swapping and load constraints for the drone starting node and the end point; Expressions (19) - (20) indicate the types of variables.

Images