Bus-assisted unmanned aerial vehicle cooperative distribution method considering dynamic demands of customer points

Abstract

The invention discloses a public transportation auxiliary unmanned aerial vehicle cooperative distribution method considering client point dynamic demands, introduces a public transportation auxiliary unmanned aerial vehicle with large service range coverage to carry out express distribution, and aims to reduce the terminal logistics distribution cost and improve the distribution timeliness. The method comprises the following steps: constructing a generation model which simulates a dynamic customer demand and considers a time window in combination with an actual scene, and establishing a bus auxiliary unmanned aerial vehicle cooperative distribution optimization model which considers the customer point dynamic demand based on a customer point generation mechanism and a bus unmanned aerial vehicle cooperative operation mechanism in the generation model; and aiming at the established model, designing a dynamic information processing strategy based on bus departure time, and fusing the dynamic information processing strategy into an adaptive large neighborhood search framework for solving, thereby obtaining a heuristic algorithm suitable for solving a bus-assisted unmanned aerial vehicle cooperative distribution problem considering the dynamic demand of a client point. The distribution path can be dynamically planned according to the demand points generated at any time, the logistics delivery efficiency is improved, and the distribution cost is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of path optimization, and in particular to a bus-assisted unmanned aerial vehicle cooperative distribution method considering dynamic customer point demand. BACKGROUND

[0002] With the rapid development of e-commerce industry, the demand for express delivery is increasing year by year, and due to the complexity of urban traffic, using traditional vehicles for distribution not only has high express delivery cost but also poor flexibility, resulting in a decline in express service satisfaction. Some areas consider using flexible unmanned aerial vehicles for terminal logistics distribution, but the defects of unmanned aerial vehicles are also obvious, namely, insufficient endurance and carrying capacity. Therefore, considering introducing buses, which are green, energy-saving, have better accessibility and can utilize more space, into the terminal distribution service of unmanned aerial vehicles can reduce the cost of distribution and alleviate traffic congestion and reduce environmental pollution. In the current actual logistics distribution process, customer demand points are usually generated in real time, but current research focuses on dynamic path planning of traditional express delivery modes, and there is little research on vehicle-unmanned aerial vehicle cooperative distribution considering dynamic customer point demand. Bus routes usually include multiple frequency buses, which have an inherent advantage in studying dynamic customer point demand. Combining buses with unmanned aerial vehicles for urban logistics terminal distribution will improve the efficiency and timeliness of express delivery. SUMMARY

[0003] The present application discloses a bus-assisted unmanned aerial vehicle cooperative distribution method considering dynamic customer point demand, which aims to consider dynamic customer demand generated over time, use buses and unmanned aerial vehicles for cooperative logistics delivery, and optimize the distribution path in real time to improve distribution efficiency.

[0004] To achieve the above-mentioned purpose, the technical scheme provided by the present application is:

[0005] A bus-assisted unmanned aerial vehicle cooperative distribution method considering dynamic customer point demand, characterized in that it comprises the following steps:

[0006] Step 1: Construct a generation model that simulates dynamic customer demand and considers time windows in combination with actual distribution scenarios;

[0007] Step 2: Establish a bus-assisted unmanned aerial vehicle cooperative operation mechanism considering dynamic customer point demand, and based on this, establish a bus-assisted unmanned aerial vehicle cooperative distribution optimization model;

[0008] Step 3: For the bus-assisted unmanned aerial vehicle cooperative distribution problem considering dynamic customer point demand, propose a dynamic information processing strategy that meets the bus-assisted unmanned aerial vehicle cooperative distribution problem;

[0009] Step 4: Establish an adaptive large neighborhood search solving framework, integrate the proposed dynamic information processing strategy, and solve the bus-assisted UAV collaborative distribution optimization model considering the dynamic demand of customer points.

[0010] To optimize the above technical solutions, the specific measures / limitations taken also include:

[0011] In step 1, the proposed generation model simulates dynamic customer demand and considers time windows, including:

[0012] Customer point arrival model: Considering the actual situation of current logistics distribution, the generation (arrival) of customer points generally follows a non-homogeneous Poisson distribution (NHPP), i.e., a certain number of customer points are generated per unit time, and the probability of a customer point appearing in a time interval is:

[0013] P(N(t+Δt)-N(t)=1)≈λ(t)Δt+ο(Δt)

[0014] where λ(t) represents the arrival intensity of customer points per unit time, N(t) represents the number of customer points at time t, N(t+Δt) represents the number of customer points at time t+Δt, P() represents the probability of the situation inside the parentheses, Δt represents the interval time, and ο() represents the high-order infinitesimal inside the parentheses.

[0015] Customer point location modeling: Considering the actual situation of current logistics distribution, the location (x i ,y i ) of customer point i is random, and in this invention, it is assumed to follow a uniform distribution within the delivery range A:

[0016] (x i ,y i )~U(A)

[0017] Customer demand modeling: Considering the actual situation of current logistics distribution, the demand of customer points is a non-negative real number, does not exceed the maximum load of UAV, and follows a uniform distribution:

[0018] w i ~Uniform(w min ,w max )

[0019] where w i represents the demand of customer point i, w max and w min represent the maximum and minimum demand, respectively.

[0020] Customer point time window modeling: considering the actual situation of current logistics distribution, the time window preset by the customer point is usually a flexible time window, that is, it allows delivery outside the preset time range, but will be punished in terms of cost. The time window generated by the dynamic customer point is also random, containing two random parameters: the middle value of the time window and the time window width When a customer point with a determined location and demand is generated, the middle value of its time window is uniformly distributed within the delivery range [L s , L E ], and the time window width is randomly selected from [20min, 40min, 60min, 120min].

[0021] In step 2, the bus-assisted UAV cooperative operation mechanism considering dynamic customer demand includes:

[0022] Dynamic customer point allocation mechanism: considering that it is impossible to supplement goods to the bus that has already started, the newly added customer points that have not been delivered up to the current time node can only be delivered by the UAV carried by the bus that will start after the current time node. That is, when customer point i is in the newly added customer point set at the cutoff node T k , then the UAV and bus combination serving customer point i must be in the bus and UAV combination set that has not started at the cutoff time node T k .

[0023] Bus and UAV access order coordination: the access order of the bus and the UAV carried by it at the bus stop needs to be consistent; that is, if the access order of bus k at the bus stop is O bk = [1, 2,..., s-1], then the access order of UAV k at the bus stop is either or consistent with the access order of bus k.

[0024] Bus and UAV running time coordination: the UAV's task execution time is constrained by the arrival time of the bus it is carried on, and the UAV needs to be launched and recovered without affecting the normal operation of the bus. That is, when the UAV k is launched, the bus k and the UAV k leave the launch bus stop at the same time; when the UAV is recovered, in order not to affect the normal operation of the bus, the UAV k needs to arrive at the recovery bus stop before the bus k.

[0025] UAV multiple access mechanism: after the UAV has served the first customer point, it can choose a bus stop to wait for the bus to carry it forward, or it can choose to continue to serve the next customer point. Based on the variable N​dkpq , which means that after the UAV k is launched from the bus stop d, it successively visits the customer point p and the customer point q, then when the UAV k leaves the customer point p, it at least carries the package required by the customer point q, and the demand of the customer point p and the customer point q is less than the maximum load U of the UAV max .

[0026] In step 2, the bus-assisted UAV collaborative distribution optimization model considering the dynamic demand of customer points is established, and the model assumptions include:

[0027] Bus operation assumption: all buses will not have accidents during operation, and there is enough space to provide goods and UAVs;

[0028] UAV assumption: the UAV flies at a constant speed and is fully charged at the beginning of each flight (the flight time is the maximum flight time), the UAV remains in a hovering state when serving customer points, and the service time is included in the flight time;

[0029] Package assumption: all packages are stored at the departure bus stop (the first bus stop), and the bus carries the allocated packages when it departs.

[0030] In step 2, the bus-assisted UAV collaborative distribution optimization model considering the dynamic demand of customer points is established, and the model objectives include:

[0031] Time window loss cost of customer points: the customer point is set as a soft time window, allowing the UAV to deliver outside the preset range, but will be subject to a cost penalty loss:

[0032]

[0033] wherein represents the time window loss at customer point i, represents the time when UAV k arrives at customer point i, T mki is a 0-1 variable, equal to 1 when UAV k visits customer point i after being launched from bus stop m, [a i , b i is the time window of customer point i, ξ and ζ are the penalty costs per unit of early and late service, respectively, a i and b i are the upper and lower bounds of the time window of customer point i, respectively.

[0034] UAV transportation cost: the unit time transportation cost of the UAV is C d , so the cost of a single flight is the total duration of the flight multiplied by C d .

[0035] In step 2, the established bus-assisted UAV collaborative distribution optimization model considering dynamic customer demand constraints include:

[0036] Dynamic package allocation constraints:

[0037]

[0038] The meaning of this constraint is: the dynamic customer demand generated before the time node can only be served by the combination of buses and UAVs departing after the time node, where P is the set of UAV launch bus stops, is the set of new customer points generated at the cutoff node, represents the set of buses and UAVs that have not departed from the cutoff node.

[0039] Bus-assisted UAV collaborative operation constraints:

[0040]

[0041] The meaning of this constraint is: if UAV k is launched from bus stop m and flies to bus stop j from customer point i for recovery, then the order of UAV k visiting bus stop m and bus stop j should be consistent with the bus k it carries. Where E mkij is the 0-1 variable for this action, is the order of bus k visiting stop j, is the order of bus k visiting stop m, M is a maximum value, G k is the set of served customer points at cutoff node T k , P is the set of UAV launch bus stops, Q is the set of UAV recovery bus stops, and X is the set of UAV and bus combinations.

[0042] In addition to the collaborative constraints, there are also access order constraints, load conservation constraints, time conservation constraints, maximum flight time and maximum load constraints.

[0043] In step 3, the dynamic information processing strategy for the bus-assisted UAV collaborative distribution problem is:

[0044] Dynamic information processing strategy based on bus departure time: based on the departure time of different shifts of buses in the bus line, the optimization process is divided into multiple time segments, and the number of new customer points in the segment is counted. Different processing strategies are adopted according to the number of customer points generated:

[0045] When the number of new customer points in the time segment is small (less than a fixed value α): new customer points are inserted into the current solution first, and then the entire distribution scheme is optimized;

[0046] When the number of new customer points in the time slice is large (greater than a fixed value a): the newly added customer points and the customer points that have been planned but have not yet started to be executed are counted, and the customer points covered by them are combined to form a set of customer points to be planned. The overall re-planning is performed on all customer points in the set.

[0047] The adaptive large neighborhood search framework incorporating dynamic information processing strategy in step 4 includes:

[0048] Initial solution generation module: a multi-visit initial solution is generated using the savings algorithm. This module is mainly aimed at the initial known customer point planning and the overall re-planning when there are many new customer points;

[0049] Repair operator and destruction operator module: the destruction operator removes a fixed number of customer nodes that have a large impact on the distribution cost from the current distribution scheme; the repair operator inserts the destroyed or newly added customer points into the current distribution scheme for the purpose of reducing the distribution cost;

[0050] Wherein the destruction operator includes: maximum loss removal, local maximum loss removal, random removal; the repair operator includes: regret value insertion, greedy insertion, sequential insertion, at the beginning of each iteration, the roulette method is used to select which destruction operator and repair operator to use according to the weight of the operator;

[0051] Adaptive adjustment module: according to the historical use of the operator and the quality of the generated neighborhood solution, the operator with better quality is selected to participate in the next iteration process, and the operator weight is updated as follows:

[0052]

[0053] Wherein, W r is defined as the weight of the operator after the rth update of the weight, S is defined as the score obtained by the operator in the past F iterations, U is the number of times the operator is used in the past F iterations, and η is the weight adjustment coefficient, which means the importance of the historical performance of the operator when updating the weight.

[0054] In step 4, the solution steps of the adaptive large neighborhood search framework incorporating dynamic information processing strategy include:

[0055] Step 1: For the initial known customer points, the savings algorithm + adaptive large neighborhood search two-stage algorithm is used to solve the initial distribution scheme, the savings algorithm is responsible for generating a high-quality initial solution, and the adaptive large neighborhood search algorithm calls the operator module and the adaptive adjustment module to iteratively optimize the initial solution;

[0056] Step 2: when reaching the dynamic information processing time node, the newly added customer points are counted, and different strategies are adopted according to different numbers: when the number of newly added customer points is small, the repair operator module is used to insert the newly added customer points into the current distribution scheme which has not started to execute, and fixed number of iteration optimization is carried out; when the number of newly added customer points is large, the two-stage algorithm is used to re-plan the planned customer point set.

[0057] Compared with the prior art, the beneficial effects of the present application are:

[0058] The present application aims at the deficiencies in the existing research on unmanned aerial vehicle terminal logistics distribution, and proposes a more practical and more practical bus-assisted unmanned aerial vehicle collaborative distribution method considering dynamic customer demand. In consideration of the dynamic customer demand that may occur at any time, the bus with a high degree of overlap with the terminal logistics distribution service range is introduced to assist the unmanned aerial vehicle to provide distribution service for the customer demand point, which makes up for the research gap of bus-assisted unmanned aerial vehicle collaborative distribution in considering dynamic customer demand, and proposes a dynamic information processing strategy for bus-assisted unmanned aerial vehicle collaborative distribution problem considering dynamic customer demand, and integrates it into the adaptive large neighborhood search framework to obtain a heuristic algorithm suitable for solving the collaborative distribution problem. The present application has good adaptability in actual scene, can reduce the cost of traditional vehicle distribution, reduce traffic congestion and environmental pollution, and improve the efficiency of terminal logistics distribution. BRIEF DESCRIPTION OF DRAWINGS

[0059] Figure 1 is a flowchart of the method of the present application.

[0060] Figure 2 is a schematic diagram of bus and unmanned aerial vehicle collaborative operation.

[0061] Figure 3 is a dynamic information processing schematic diagram. DETAILED DESCRIPTION

[0062] The above content of the present application will be further described in the form of examples, but this should not be understood as limiting the scope of the above subject matter of the present application to the following examples. Any technology realized based on the above content of the present application belongs to the scope of the present application.

[0063] The present application proposes a bus-assisted unmanned aerial vehicle collaborative distribution method considering dynamic customer demand, and the flowchart is as shown in Figure 1 The method comprises the following steps:

[0064] (1) combined with the actual distribution scene, a generation model of simulating dynamic customer demand and considering time window is constructed.

[0065] The generated model proposed in step (1) simulates dynamic customer demand and considers time window, which includes:

[0066] 1.1, customer point arrival model

[0067] In combination with the actual situation of current logistics distribution, the generation (arrival) of customer points generally obeys non-homogeneous Poisson distribution (NHPP), that is, the probability of one customer point appearing in a time interval is:

[0068] P(N(t+Δt)-N(t)=1)≈λ(t)Δt+ο(Δt)

[0069] where λ(t) represents the arrival intensity of customer points per unit time, N(t) means the number of customer points when the time is t, N(t+Δt) means the number of customer points when the time is t+Δt, P() represents the probability of the situation in the parentheses, Δt represents the interval time, and ο() represents the high-order infinitesimal in the parentheses.

[0070] 1.2, customer point location modeling

[0071] In combination with the actual situation of current logistics distribution, the location (x i ,y i ) of the generated customer point i is random, and in the present application, it is assumed to obey uniform distribution within the distribution range A:

[0072] (x i ,y i )~U(A)

[0073] 1.3, customer point demand modeling

[0074] In combination with the actual situation of current logistics distribution, the demand of customer points is a non-negative real number, which does not exceed the maximum load of the unmanned aerial vehicle and obeys uniform distribution:

[0075] w i ~Uniform(w min ,w max )

[0076] where w i represents the demand of customer point i, w max and w min represent the maximum demand and the minimum demand, respectively.

[0077] 1.4, customer point time window modeling

[0078] In combination with the actual situation of current logistics distribution, the preset time window of customer points is usually a flexible time window, that is, it allows distribution outside the preset time range, but will be punished in terms of cost. The time window of dynamic customer points is also random, containing two random parameters: the middle value of the time window and the time window width When generating a customer point whose location and demand are determined, the median value of its time window In the delivery range [L s ,L E ] obeys uniform distribution, and the time window width Randomly select from [20min,40min,60min,120min].

[0079] (2) Establish a cooperative operation mechanism for public transportation auxiliary drones that takes into account the dynamic needs of customer points. The schematic diagram is as follows Figure 2 As shown in the figure, a bus-assisted drone collaborative delivery optimization model is established based on this.

[0080] 2.1. The coordinated operation mechanism of public transportation auxiliary drones considering the dynamic needs of customer points includes:

[0081] 2.1.1, Dynamic customer point allocation mechanism: Considering that it is impossible to replenish the goods for the bus that has already departed, the customer points that have not yet been delivered as of the current time node can only be delivered by the drones carried by the buses that depart after the current time node. That is, when the customer point i is at the end node T k The set of newly added customer points Then the drone and bus combination serving customer point i must be at the deadline node T k A combination of buses and drones that have not yet departed among.

[0082] 2.1.2. Bus and UAV Visit Order Coordination: The bus and its carried UAV need to maintain the same visit order in the bus stops; that is, the order O of bus k visiting the bus stops is bk =[1,2,......,s-1] is known, then the order in which drone k visits the sites is either Or keep the same access order as bus k.

[0083] 2.1.3 Bus and Drone Operation Time Coordination: The drone's mission execution time is constrained by the arrival time of the bus it is carrying. The drone must be launched and recovered without disrupting the bus's normal operation. Specifically, when launching drone k, bus k and drone k must simultaneously depart the launch bus stop. When recovering drone k, drone k must arrive at the recovery bus stop before bus k to avoid disrupting the bus's normal operation.

[0084] 2.1.4, UAV multiple visit mechanism: when the UAV has served the first customer point, it can choose to wait for a bus at a bus stop to take it forward, or it can choose to continue to serve the next customer point. Based on variable N dkpq , which means that after the UAV k is launched from the bus stop d, it successively visits customer point p and customer point q, then when the UAV k leaves customer point p, it at least carries the package required by customer point q, and the demand of customer point p and customer point q is less than the maximum load U of the UAV max .

[0085] 2.2, the established bus-assisted UAV collaborative distribution optimization model considering the dynamic demand of customer points, the model assumptions include:

[0086] 2.2.1, bus operation assumption: all buses will not have accidents during operation, and there is enough space to provide goods and UAVs.

[0087] 2.2.2, UAV assumption: UAV flies at a constant speed and starts each flight with full power (flight time is maximum flight time), UAV remains hovering when serving customer points, and service time is included in flight time.

[0088] 2.2.3, package assumption: all packages are stored at the bus departure station (the first station), and the bus carries the allocated packages when it departs.

[0089] 2.3, the established bus-assisted UAV collaborative distribution optimization model considering the dynamic demand of customer points aims to minimize the cost of distributing all known customer points and dynamically generated customer points, including:

[0090] 2.3.1, customer point time window loss cost: the customer point is set as a soft time window, allowing the UAV to distribute outside the preset range, but will be subject to cost penalty loss:

[0091]

[0092] where represents the time window loss at customer point i, represents the time when UAV k arrives at customer point i, T mki is a 0-1 variable, equal to 1 when UAV k visits customer point i after being launched from bus stop m, [a i ,b i are the time window of customer point i, ξ and ζ are the penalty cost per unit of early and late service, respectively, a i and b i are the upper and lower bounds of the time window of customer point i.

[0093] 2.3.2, UAV transportation cost: the UAV transportation cost per unit time is C d Therefore, the cost of a single flight is the total duration of the flight multiplied by C d .

[0094] 2.4, the bus-assisted UAV collaborative distribution optimization model considering the dynamic demand of customer points, the constraints include:

[0095] 2.4.1, dynamic package allocation constraints:

[0096]

[0097] The meaning of this constraint is that the dynamic customer demand generated before the time node can only be served by the combination of buses and UAVs departing after the time node, where P is the set of UAV launch bus stops, is the set of new customer points generated at the cutoff node, represents the set of buses and UAVs that have not departed from the cutoff node.

[0098] Bus-assisted UAV collaborative operation constraints considering dynamic customer point demand:

[0099]

[0100] The meaning of this constraint is that if UAV k is launched from bus stop m and flies to bus stop j from customer point i for recovery, the order in which UAV k visits bus stop m and bus stop j should be consistent with the bus k it carries. Where E mkij is a 0-1 variable to describe this action, is the order in which bus k visits stop j, is the order in which bus k visits stop m, M is a maximum value, G k is the set of served customer points at cutoff node T k , P is the set of UAV launch bus stops, Q is the set of UAV recovery bus stops, and X is the set of UAV and bus combinations.

[0101] 2.4.2, in addition to the collaborative constraints, it also includes: load conservation constraints, time conservation constraints, maximum flight time and maximum load constraints, visit order constraints.

[0102] (3) For the bus-assisted UAV collaborative distribution problem considering dynamic customer point demand, a dynamic information processing strategy consistent with the bus-assisted UAV collaborative distribution problem is proposed.

[0103] In step 3, the dynamic information processing strategy consistent with the bus-assisted UAV collaborative distribution problem is:

[0104] Dynamic information processing strategy based on bus departure time: based on the departure time of different bus in the bus line, the optimization process is divided into multiple time segments, the newly added customer points in the segment are counted, as shown in the schematic diagram Figure 3 Different processing strategies are taken according to the number of customer points generated:

[0105] When the number of newly added customer points in the time segment is small (less than a fixed value a): the newly added customer points are inserted into the current solution first, and then the entire distribution scheme is optimized;

[0106] When the number of newly added customer points in the time segment is large (greater than a fixed value a): count the newly added customer points and the customer points that have been planned but have not yet started to be executed, and form a set of customer points to be planned, and then re-plan all customer points in the set.

[0107] (4) Establish an adaptive large neighborhood search solving framework, integrate the proposed dynamic information processing strategy, and solve the bus-assisted UAV cooperative distribution optimization model considering dynamic customer demand.

[0108] 4.1, the adaptive large neighborhood search framework integrated with dynamic information processing strategy includes:

[0109] 4.1.1, initial solution generation module: use the savings algorithm to generate a multi-visit initial solution. This module is mainly used for initial known customer point planning and overall re-planning when there are many newly added customer points, and the specific steps are:

[0110] Step 1: First, under the premise of meeting the maximum flight time of the UAV and the maximum load of the UAV, assign two stations before and after each customer point as launch station and recovery station;

[0111] Step 2: Randomly select two UAV single-visit paths, consider different merging directions and UAVs performing tasks, select the way that meets the basic conditions and maximally reduces the cost to merge and update the current solution and iteration number;

[0112] Step 3: Repeat Step 2 until the set iteration number is reached, then complete the solution and obtain a multi-visit initial distribution scheme;

[0113] 4.1.2, repair operator and destruction operator module: the destruction operator removes a fixed number of customer nodes that have a large impact on distribution cost from the current distribution scheme; the repair operator inserts the destroyed or newly added customer points into the current distribution scheme to reduce the distribution cost;

[0114] The destruction operator includes:

[0115] Maximal loss removal: remove the customer node that causes the most cost increase in the current distribution scheme, the specific operation is: remove each customer node one by one, record the amount of total distribution cost reduction after removal, select the customer node with the largest reduction amount for removal;

[0116] Local maximal loss removal: randomly select a UAV path, and select the customer node in the path that causes the largest cost increase for removal;

[0117] Random removal: randomly remove a fixed number of customer nodes from the current distribution scheme.

[0118] The repair operator includes:

[0119] Regret value insertion: the difference between the cost increase of the optimal position and the suboptimal position of a certain customer node is defined as the regret value, and the customer node with the largest regret value is preferentially inserted into the corresponding optimal position;

[0120] Greedy insertion: insert the customer node into the position that causes the smallest cost increase;

[0121] Sequential insertion: insert the removed customer node into the position in the order of the current solution.

[0122] At the beginning of each iteration, the roulette method is used to select which destruction operator and repair operator to use according to the weights of the operators;

[0123] 4.1.3, adaptive adjustment module: according to the historical use of the operator and the quality of the generated neighborhood solution, adaptively select better quality operators to participate in the next iteration process, and the operator weight is updated as follows:

[0124]

[0125] where W r is the weight of the operator after the rth update of the weight, S is the score obtained by the operator in the past F iterations, U is the number of times the operator is used in the past F iterations, and η is the weight adjustment coefficient, which means the importance of the historical performance of the operator when updating the weight.

[0126] 4.2, the solving steps of the adaptive large neighborhood search framework integrating dynamic information processing strategy include:

[0127] Step 1: For the initially known customer nodes, use the saving route + adaptive large neighborhood search two-stage algorithm to solve the initial distribution scheme, and the saving route algorithm is responsible for generating a high-quality initial solution, and the adaptive large neighborhood search algorithm calls the operator module and the adaptive adjustment module to iteratively optimize the initial solution;

[0128] Step 2: when reaching the dynamic information processing time node, the newly added customer points are counted and different strategies are taken according to different numbers: when the number of newly added customer points is small, the repair operator module is used to insert the newly added customer points into the current distribution scheme which has not yet started to execute, and fixed number of iteration optimization is carried out; when the number of newly added customer points is large, two-stage algorithm is used to re-plan the set of customer points to be planned.

[0129] The above is only the preferred embodiment of the present application, and does not limit the present application in any form. Any skilled person in the art, without departing from the technical solution of the present application, according to the technical essence of the present application, any simple modification, equivalent replacement and improvement of the above embodiment, etc. still belongs to the protection scope of the technical solution of the present application.

Claims

1. A public transportation-assisted UAV collaborative delivery method considering the dynamic needs of customer points, characterized by: The following steps are involved: Step 1: Based on actual delivery scenarios, build a generative model that simulates dynamic customer demand and considers time windows; Step 2: Establish a bus-assisted UAV collaborative operation mechanism that considers the dynamic needs of customer points, and based on this, establish a bus-assisted UAV collaborative delivery optimization model; Step 3: Propose a dynamic information processing strategy that meets the public transportation-assisted UAV collaborative delivery problem; Step 4: Establish an adaptive large neighborhood search solution framework, integrate the proposed dynamic information processing strategy, and solve the bus-assisted drone collaborative delivery optimization model.

2. The public transportation-assisted UAV collaborative delivery method considering the dynamic needs of customer points according to claim 1 is characterized in that: The generation model proposed in step 1 to simulate dynamic customer demand and consider time windows includes: Customer point arrival model: Based on the actual situation of current logistics distribution, customer point arrival follows a non-homogeneous Poisson distribution, that is, a certain number of customer points are generated in a unit of time; Customer point location modeling: Based on the actual situation of current logistics distribution, the locations of customer points are randomly generated and are uniformly distributed within the delivery range by default; Customer point demand modeling: Based on the actual situation of current logistics distribution, the customer point demand is a non-negative real number, a single package does not exceed the maximum load of the drone, and follows a uniform distribution; Customer point time window modeling: Based on the actual situation of current logistics distribution, the time window preset for the customer point is a flexible time window, which means that delivery outside the preset time range is allowed, but there will be a cost penalty; the time window generated by dynamic customer points is also random and contains two random parameters: the time window median value and the time window width.

3. The public transportation-assisted UAV collaborative delivery method considering the dynamic needs of customer points according to claim 1 is characterized in that: The public transportation auxiliary drone collaborative operation mechanism established in step 2 considering the dynamic needs of customer points includes: Dynamic customer point allocation mechanism: Considering that it is impossible to replenish goods on buses that have already departed, customer points that have not yet been delivered to at the current time point can only be delivered by drones carried by buses that depart after the current time point; Bus and drone visit order coordination: Buses and the drones they carry need to maintain consistent visit order at bus stops; Bus and drone operation time coordination: The drone's mission execution time is constrained by the arrival time of the bus it carries. The drone needs to be launched and recovered without affecting the normal operation of the bus. Drone multiple visit mechanism: After serving the first customer point, the drone can choose a bus stop to wait for the bus to take it forward, or it can choose to continue serving the next customer point.

4. The public transportation-assisted UAV collaborative delivery method considering the dynamic needs of customer points according to claim 1 is characterized in that: The public transportation-assisted UAV collaborative delivery optimization model established in step 2 considering the dynamic needs of customer points includes: Model assumptions: including bus operation assumptions, drone assumptions, and package assumptions; Model objective: Minimize the cost of all known customer points and dynamically generated customer points, including the time window loss cost of the customer points and the drone transportation cost; Model constraints: including bus-assisted drone collaborative operation constraints, dynamic package distribution constraints, load conservation constraints, time conservation constraints, maximum flight time and maximum load constraints, and visit order constraints.

5. The public transportation-assisted UAV collaborative delivery method considering the dynamic needs of customer points according to claim 1 is characterized in that: The dynamic information processing strategy in step 3 that meets the public transportation-assisted UAV collaborative delivery problem is: Dynamic information processing strategy based on bus departure time: The optimization process is divided into multiple time segments based on the departure time of different buses on the bus route. New customer points in each segment are counted and different processing strategies are adopted according to the number of customer points generated: When the number of newly added customer points in a time segment is small, that is, less than a fixed value α: the newly added customer points are first inserted into the current solution, and then the entire delivery plan is optimized; When the number of newly added customer points in a time segment is large and greater than a fixed value α: count the newly added customer points up to the current time node and the customer points that have been planned but not yet executed, form the customer points covered by them into a set of customer points to be planned, and re-plan all customer points in this set as a whole.

6. The public transportation-assisted UAV collaborative delivery method considering the dynamic needs of customer points according to claim 5 is characterized in that: The adaptive large neighborhood search framework incorporating dynamic information processing strategies in step 4 includes: Initial solution generation module: This module uses a mileage-saving algorithm to generate a multi-visit initial solution for planning of initially known customer points and for overall replanning when a large number of new customer points are added. Repair operator and destruction operator modules: The destruction operator removes a fixed number of customer nodes that have a significant impact on delivery costs from the current delivery plan; the repair operator inserts the destroyed or newly counted customer nodes into the current delivery plan with the goal of reducing delivery costs; Adaptive adjustment module: Adaptively selects better operators to participate in the next iteration process based on the historical operators used and the quality of the generated neighborhood solutions.

7. The public transportation-assisted UAV collaborative delivery method considering the dynamic needs of customer points according to claim 6 is characterized in that: The solution steps of the adaptive large neighborhood search framework incorporating the dynamic information processing strategy in step 4 include: Step 1: For the initial known customer points, a two-stage algorithm (mileage saving and adaptive large neighborhood search) is used to solve the initial delivery plan. The mileage saving algorithm is responsible for generating a high-quality initial solution, while the adaptive large neighborhood search algorithm calls the operator module and the adaptive adjustment module to iteratively optimize the initial solution. Step 2: When the dynamic information processing time node is reached, count the newly added customer points and adopt different strategies based on the number: When the number of newly added customer points is small, the repair operator module is used to insert the newly added customer points into the distribution plan that has not yet been executed, and a fixed number of iterative optimizations are performed; when the number of newly added customer points is large, the planned customer point set is replanned using a two-stage algorithm.

Images