Vehicle and unmanned aerial vehicle cooperative path planning method considering three-dimensional loading constraint

Abstract

The invention discloses a vehicle and unmanned aerial vehicle cooperative path planning method considering three-dimensional loading constraint. The method comprises the following steps: respectively adopting two unloading modes of first-in and second-out and unloading and then loading. Wherein the cargo first-in and second-out mode means that only cargos located on the outermost layer of the vehicle can be distributed every time, and the cargo unloading allowing and then loading mode means that cargos located on the inner layer of the vehicle are allowed to be distributed; constructing a mixed integer linear programming model including decision variables, auxiliary variables, optimization targets and constraint conditions; and designing a mixed meta-heuristic solving method with high solving efficiency, and solving a loading scheme and a distribution scheme. The method is applied to the field of collaborative path planning, the loading and distribution processes of goods are considered at the same time, the common modes of first-in and second-out and unloading and then loading in the logistics distribution process are adopted, the main path of the distribution vehicle and the path of the unmanned aerial vehicle are optimized, the logistics distribution efficiency is remarkably improved, and meanwhile the requirements of customers are met.

Description

Technical Field

[0001] The present invention relates to the technical field of collaborative path planning, and in particular to a method for collaborative path planning between a vehicle and an unmanned aerial vehicle taking into account three-dimensional loading constraints. Background Art

[0002] Modern logistics and distribution face a series of challenges, including the complex problem of planning the loading and distribution of goods at the same time. In order to efficiently deliver products from factories or distribution centers to customers, logistics companies must solve these problems. In logistics and distribution, loading and distribution are two key stages. The loading stage needs to consider the size, weight, characteristics and loading sequence of the goods to facilitate subsequent unloading operations. The distribution stage needs to determine the best route so that the delivery vehicle can effectively visit multiple customers and meet all order requirements while minimizing delivery time.

[0003] In the field of logistics and distribution, unmanned aerial vehicles (UAVs) have shown great application opportunities due to their high efficiency and low cost, and have gradually become a new research hotspot. However, although scholars have proposed better methods for multi-UAV collaborative distribution, it is difficult to complete large-scale and long-distance distribution tasks due to the limited endurance of UAVs.

[0004] The collaborative reconnaissance method of vehicles and drones can make up for the shortcomings of using only drones for reconnaissance, and has become a more popular research direction for researchers in recent years. The more classic studies on this type of problem include the Flying Sidekick Traveling Salesman Problem (FSTSP) and the Traveling Salesman Problem with Drone (TSP-D).

[0005] In the Flying Assistant Traveling Salesman Problem (FSTSP), it is assumed that the delivery drone leaves the vehicle, completes a delivery task, and then immediately returns to the vehicle. It uses Mixed Integer Linear Programing (MILP) to describe the problem, and proposes a heuristic method to solve it. Finally, the proposed model and algorithm are evaluated and demonstrated through a series of computational experiments. The above heuristic algorithm first solves the TSP path of the vehicle based on all customer points, and then assigns each eligible customer to the drone path, or reallocates it to other locations on the vehicle path. In the Traveling Salesman Problem with Drones (TSP-D), unlike FSTSP, it allows the drone to return to the node where the drone was launched, and provides a new mixed integer programming model and corresponding heuristic algorithm.

[0006] Considering the coordination between loading and delivery in logistics activities, the problem of simultaneously handling routing and loading / packaging has been widely studied in recent years in many variants. In these problems, the routing is valid when certain loading constraints are met, which can be classified as two-dimensional (2D) or three-dimensional (3D) loading constraints according to whether items can be stacked on other items. However, in the current research on logistics and delivery, the vehicle loading and delivery problem is separated from the problem of vehicle-drone collaboration, which lacks realism. Summary of the invention

[0007] In view of the above-mentioned deficiencies in the prior art, the present invention provides a vehicle and UAV collaborative path planning method that takes into account three-dimensional loading constraints, while also taking into account the loading and distribution process of goods, which can significantly improve the logistics distribution efficiency.

[0008] To achieve the above object, the present invention provides a vehicle-UAV collaborative path planning method considering three-dimensional loading constraints, comprising the following steps: Step 1: construct decision variables in the process of collaborative path planning between vehicles and drones, wherein the decision variables determine the vehicle path, the drone path and the cargo loading location; Step 2: construct the optimization goal and constraints of the vehicle-UAV collaborative path planning, wherein the optimization goal is to minimize the delivery time, and the constraints include vehicle path constraints, UAV path constraints, vehicle-UAV collaborative constraints, cargo pressure resistance constraints, vehicle loading constraints, cargo first-in-last-out constraints, and cargo unloading and reloading constraints; Step 3: Under the first-in-last-out mode and the mode where unloading and reloading are allowed, a hybrid metaheuristic solution method is used to calculate the cargo loading plan and the delivery plan. The first-in-last-out mode means that only the cargo on the outermost layer of the vehicle can be delivered each time, and the mode where unloading and reloading are allowed means that the cargo on the inner layer of the vehicle can be delivered.

[0009] Compared with the prior art, the present invention has the following beneficial technical effects: The vehicle-UAV collaborative path planning method considering three-dimensional loading constraints proposed in the present invention takes into account the loading and distribution processes of goods, and optimizes the main path of the delivery vehicle and the path of the UAV respectively according to the two unloading modes of "first in, last out" and "unloading and then loading" which are common in the logistics distribution process. It can significantly improve the logistics distribution efficiency while meeting customer needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0011] Figure 1 It is a flow chart of a method for collaborative path planning between a vehicle and an unmanned aerial vehicle considering three-dimensional loading constraints in an embodiment of the present invention; Figure 2 The position number of the cargo in the carriage in the embodiment of the present invention Schematic diagram; Figure 3 Schematic diagram of the unloading mode in the embodiment of the present invention, wherein (a) is a schematic diagram of the first-in-last-out mode, and (b) is a schematic diagram of the unloading-first-then-loading mode; Figure 4 Schematic diagram of target access sequence in an embodiment of the present invention; Figure 5 This is a schematic diagram of goods exchange on the same floor in an embodiment of the present invention; Figure 6 This is a schematic diagram of the exchange of goods with the same compression resistance grade in an embodiment of the present invention; Figure 7 Schematic diagram of the additional time cost caused by cargo loading and unloading in the embodiment of the present invention, wherein (a) is a schematic diagram of the additional time cost for loading and unloading of the delivered cargo 22, and (b) is a schematic diagram of the additional time cost for loading and unloading of the delivered cargo 34; Figure 8 1 is an example diagram of a destruction operator in an embodiment of the present invention, wherein (a) is a schematic diagram of a deletion point that is not a take-off / landing point, and (b) is a schematic diagram of a deletion point that is a take-off / landing point; Fig. 9 1 is an example diagram of a repair operator in an embodiment of the present invention, wherein (a) is a schematic diagram of a target point being inserted into a vehicle path, and (b) is a schematic diagram of a target point being inserted into a drone path.

[0012] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0013] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0014] In addition, the technical solutions between the various embodiments of the present invention can be combined with each other, but it must be based on the fact that ordinary technicians in the field can implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0015] like Figure 1 The figure shows a method for collaborative path planning between a vehicle and a UAV considering three-dimensional loading constraints disclosed in this embodiment, which mainly includes the following steps: Step 1: construct decision variables in the process of collaborative path planning between vehicles and drones, where the decision variables determine the vehicle path, drone path, and cargo loading location; Step 2: Construct the optimization objectives and constraints of the collaborative path planning of vehicles and drones. The optimization objective is to minimize the delivery time. The constraints include vehicle path constraints, drone path constraints, vehicle-drone collaboration constraints, cargo pressure resistance constraints, vehicle loading constraints, cargo first-in-last-out constraints, and cargo unloading and reloading constraints. Step 3: Under the first-in-last-out mode and the mode where unloading and reloading are allowed, a hybrid metaheuristic solution method is used to calculate the cargo loading plan and the delivery plan. The first-in-last-out mode means that only the cargo on the outermost layer of the vehicle can be delivered each time, and the mode where unloading and reloading are allowed means that the cargo on the inner layer of the vehicle can be delivered.

[0016] The planning method in this embodiment takes into account the loading and distribution process of goods at the same time, and optimizes the main path of the delivery vehicle and the path of the drone respectively, which can significantly improve the efficiency of logistics distribution while meeting customer needs.

[0017] In the vehicle and drone collaborative path planning method of this embodiment, a target point set is defined The goods required by a group of customers (i.e. a collection of customers) are loaded onto the vehicle in a certain order. The top of the carriage can be opened for loading and unloading. One vehicle and multiple drones are used as delivery tools to collaborate to complete the delivery task. The entire delivery process can be divided into two stages: cargo loading and cargo delivery.

[0018] During the cargo loading phase, the vehicle is loaded with multiple pieces of cargo of the same shape. The carriage is divided into grids according to the shape of the cargo, and the position numbers of the cargo in the carriage are determined from top to bottom. ,like Figure 2 In this problem, the cargoes have different compression resistances, and cargoes with poor compression resistance should be loaded on the upper layer as much as possible. To simplify the problem, this example uses compression resistance grades. To quantify the compressive strength of the cargo, indicating the target point The required compression level of the cargo. The stronger the compression resistance of the cargo, The bigger.

[0019] During the cargo delivery phase, the vehicle can carry multiple heterogeneous drones from the base (i.e., warehouse) and return to the warehouse after completing the delivery of all target points. When the vehicle stops at a target point, the drone can take off from the vehicle to perform a delivery mission and land on the vehicle at other target points to charge or replace the battery.

[0020] In addition, the order of goods distribution is closely related to the order of goods loading. This embodiment considers two unloading modes: "first in, last out" and "unloading before loading". Figure 3 In the "first in, last out" mode, the cargo that is pressed down cannot be unloaded, that is, only the cargo exposed in the top view of the carriage can be unloaded. When a cargo is unloaded, new cargo may be exposed. Figure 3 As shown in (a), the light-colored goods are the goods that can be delivered at present. After the delivery of goods 10 is completed, goods 22 are exposed. In the "unloading and reloading" mode, when the delivered goods are pressed down, the goods above them need to be unloaded and then loaded into the vehicle from the original position. Unloading and reloading the goods takes a certain amount of time. Figure 3 As shown in (b), when the goods 22 are delivered, the goods 10 need to be unloaded and loaded from the original position after the delivery is completed, but its position will be moved down one level.

[0021] To facilitate modeling and solving of the problem, this embodiment makes the following assumptions: The vehicle has sufficient fuel to complete the entire delivery mission; The drone only delivers one piece of cargo per flight; The delivery time of goods at each destination point is the same and is not considered in the model; The delivery process of the vehicle does not affect the take-off and landing of the drone, so the take-off and landing time of the drone is not considered; Without considering the effect of load on speed, the vehicle and the drone maintain a constant speed; The drone can only take off or land when the vehicle is stationary; Considering the safety of the drone, the vehicle should arrive at the landing point of the drone first. That is to say, the drone is not allowed to wait for the vehicle.

[0022] All goods required by each destination point are packaged into uniform size; Simplify the compression constraints on goods, and stipulate that goods with poor compression resistance cannot be placed in the lower layer with strong compression resistance.

[0023] This embodiment constructs a mixed integer linear programming model for solving the collaborative path planning of vehicles and UAVs considering three-dimensional loading constraints, which includes decision variables, auxiliary variables, optimization objectives and constraints.

[0024] Decision variables include variables that determine the vehicle path , variables that determine the drone’s path and the variables that determine where the cargo is loaded , specifically: When the vehicle moves from the target point Drive to the destination hour, ,otherwise ; When a drone From the target point Drive to the destination hour, ,otherwise ; If the target point The required cargo is in the carriage When the position ,otherwise .

[0025] In order to eliminate sub-loops and achieve the coordination between vehicles and drones, it is necessary to introduce auxiliary variables that reflect the order in which the target points are visited, including: The order in which variables are accessed , Target point With target point Comparison variables in the order they are accessed , the vehicle reaches the destination Time , drones Arrive at the destination Time , where the target point visited by the drone is Same as its take-off point, e.g. Figure 4 As shown. The auxiliary variables meet the following conditions: If the target point At the target point Delivered between ,otherwise . At the same time, the auxiliary variables meet the following conditions: ; ; ; ; ; ; ; ; in, is the number of target points, The transportation arrival point is the gathering point. is the target point set, Indicates that the base is visited before all target points. is a set containing the base and all target points, is the time when the vehicle departs from the base, is the time when the drone departs from the base, Assemble for drones.

[0026] The optimization goal in this embodiment is to minimize the delivery time, which includes the driving time of the vehicle, the driving time of the drone, the waiting time consumed by the vehicle or drone to achieve synchronization, etc. The default delivery time of the goods is the same and is not considered in this embodiment.

[0027] The constraints in the mixed integer linear programming model include vehicle path constraints, UAV path constraints, vehicle-UAV collaboration constraints, cargo pressure resistance constraints, vehicle loading constraints, cargo first-in-last-out constraints, and cargo unloading and reloading constraints.

[0028] To ensure the feasibility of the vehicle's driving path, the vehicle path constraints defined in this embodiment include: The constraints that ensure that the vehicle departs from the base and returns to the base after completing all delivery tasks are: ; ; Eliminate the constraints of the sub-loop, which is: ; Constraints to maintain consistency of progress and out-degree ; The lower bound of the time it takes for a vehicle to reach a target node is the sum of the time it takes the vehicle to depart from the previous node and the time it takes the vehicle to travel between the two nodes. Therefore, the time constraint for the vehicle to reach the target node is: ; in, For the vehicle to move from the target point Drive to the destination The time consumption, is an infinite positive number, Gather at the vehicle departure point.

[0029] The drone collection in this embodiment There are multiple heterogeneous drones in the system, and the definition of drone path constraints includes: The constraint that each drone can only take off or land once at the same point is: ; ; The delivery point of the drone is specified in the auxiliary decision variable The constraints that are consistent with the take-off point of the drone are: ; ; Different from the definition of vehicle arrival time, the UAV arrival time uses tight constraints to avoid the situation where the UAV waits for the vehicle. The vehicle arrival time includes the possible time that the vehicle waits for the UAV, which is: ; ; ; ; in, For drones From the target point Drive to the destination time-consuming; The endurance constraints of the drone are: ; in, For drones Maximum battery life; The intersection constraint of the drone paths that ensures that each drone can take off next time after landing is: ; in, Gather at the transportation arrival point.

[0030] The vehicle-drone collaboration constraints in this embodiment explain how vehicles and drones work together to complete collaborative delivery tasks, including: Regardless of whether vehicle delivery or drone delivery is used, the constraint to ensure that each target is visited once is: ; in, is the set of cargo locations in the carriage; The constraints that require the take-off and landing points of the drone to be accessible by the vehicle are: ; ; The constraint that requires the visit order of the drone take-off point in the vehicle's visit path to be before the landing point is: ; The constraints to ensure time synchronization at the take-off point of the drone are: ; ; The constraints to ensure time synchronization at the drone landing point are: ; ; The cargo compression constraints and vehicle loading constraints in this embodiment are specifically: The constraint that each location can only carry one cargo at most and all cargoes are loaded is: ; ; Tight loading restraints, which stipulate that cargo is loaded from the bottom of the compartment upwards to prevent gaps from forming at the bottom of the compartment, are: ; Compression restraints ensure that cargo with poor compression resistance is always on the upper layer or on the same layer as cargo with strong compression resistance during loading. ; in, is the set of cargo locations in the carriage, Collection of goods.

[0031] When goods are delivered, this embodiment constructs corresponding constraints for the two modes commonly used in distribution operations: "unloading and reloading are not allowed" ("first in, last out") and "unloading and reloading are allowed", including: The first-in, last-out constraint of goods links the loading order of goods with the delivery path. The goods on the lower layer can only be delivered after the goods on the upper layer have been delivered. It is: ; in, The quantity of cargo that can be loaded on each floor of the carriage; The cargo can be unloaded and reloaded. Under this type of constraint, the cargo on the lower level can be delivered before the cargo on the upper level, but it will incur additional loading and unloading time costs. The time it takes for each cargo to be unloaded and reloaded is ,use Indicates the quantity of cargo that needs to be unloaded and reloaded after arriving at a certain location.

[0032] In the specific implementation process, the time constraint of the vehicle reaching the target point in the vehicle path constraint can be replaced by the following formula: ; ; ; The following formula can be used to replace the first two formulas of the UAV arrival time constraint in the UAV path constraint, specifically: ; ; ; Among them, the symbol Indicates taking The integer part of Can be expressed The total number of layers below the location. The reason why the drone path constraint only sets one boundary is: There may be multiple items on top of the cargo , each Each corresponds to a constraint condition, The constraint that takes the longest time should be satisfied.

[0033] After completing the construction of the mixed integer linear programming model, although the mathematical programming optimizer Gurobi can be used directly to solve the problem and obtain the exact solution of the coordinated path planning of the vehicle and the drone, the solution becomes extremely slow as the scale of the problem becomes larger. Based on this, in order to speed up the solution process, this embodiment adopts a two-stage meta-heuristic solution method, which can greatly improve the solution efficiency while obtaining a relatively good and satisfactory solution. Therefore, in this embodiment, the calculation process of the cargo loading plan and the distribution plan in the cargo first-in-last-out mode and the cargo unloading and reloading mode is specifically as follows: Step 301, obtaining an initial cargo loading plan for all cargoes; Step 302, iteratively optimizing the initial cargo loading plan to obtain a cargo loading plan; Step 303, based on the cargo loading plan, optimize the distribution plan in the cargo first-in-last-out mode and the cargo unloading and reloading mode respectively.

[0034] In the process of constructing the initial cargo loading plan in step 301, cargo is loaded from the bottom of the carriage in order of compression resistance from high to low. The specific implementation process includes the following steps: Step 301.1, obtain the cargo space set in the carriage , cargo collection ,in, is the number of cargo spaces in the carriage, is the quantity of goods, and ; Step 301.2, select the cargo space at any corner at the bottom of the carriage as the current cargo space; Step 301.3, randomly select one from all currently unloaded cargoes with the highest compression resistance level and load it in the current cargo position; Step 301.3, determine whether the storage location on the same floor as the current storage location and adjacent to the current storage location is empty: If yes, use the adjacent cargo location as the current cargo location; Otherwise, the current cargo location will be the one layer above the current cargo location and adjacent to the current cargo location. Step 301.4, determine whether all goods have been loaded: If so, output the current loading plan as the initial cargo loading plan; Otherwise, return to step 301.3.

[0035] After obtaining the initial cargo loading plan, this embodiment uses the taboo search algorithm to optimize it, thereby obtaining the cargo loading plan. The specific implementation process includes the following steps: Step 302.1, obtain the maximum number of iterations Initial cargo loading plan , and let the current number of iterations 、Current solution 、Current optimal solution ; Step 302.2, use the exchange operator to After transformation, several candidate solutions are obtained. A collection of; Judge 302.3, select the best solution in the candidate solution set , to determine whether there is a : If yes, proceed to step 302.4; Otherwise, proceed to step 302.5; Step 302.4, determine Whether the amnesty rules are met: If yes, proceed to step 302.5; Otherwise, After deleting from the candidate solution set, return to step 302.3; Step 302.5, determine Is it better than : If so, = , = Then, proceed to step 302.6; Otherwise, greedily select a candidate solution from the current candidate solution set as the current solution ; Step 302.6, change the current Add to the taboo list and make , and determine Is it established: If so, output the current As a cargo loading solution; Otherwise, return to step 302.2.

[0036] In the specific implementation process of step 302.4, the conditions for satisfying the amnesty rule can be set adaptively according to the needs, for example, when the same solution is tabooed and the number of consecutive appearances exceeds a threshold, the amnesty rule is satisfied. And / or when the number of iterations in which the solution is tabooed and does not appear continuously exceeds a threshold, the amnesty rule is satisfied, etc. In addition, the taboo table in this embodiment has a certain length limit. When the number of tabooed solutions exceeds a set threshold, the taboo of some solutions set in the taboo table is lifted, for example, the better 10% is lifted.

[0037] It is worth noting that in step 302.5, and When the optimal solution is and The corresponding loading plan adopts the optimization process of subsequent steps 303.1A-step 303.4A or steps 303.1B-step 303.2B to obtain the corresponding distribution plan, and then calculates and The target value of the corresponding delivery plan (i.e. delivery time), and The solution with smaller objective value is better.

[0038] In the specific implementation process, this embodiment designs two exchange operators that meet the cargo pressure resistance constraints, namely the same-layer cargo exchange operator and the same-pressure resistance grade cargo exchange operator. Each time, one operator is randomly selected to exchange the positions of two cargoes, thereby generating a new neighbor solution. Among them: The same-layer cargo exchange operator randomly selects two cargoes on the same layer of the carriage and exchanges their positions. Figure 5 As shown, cargo 1 and cargo 2 are located on the same layer. Regardless of whether their compression resistance levels are the same, exchanging their positions will not violate the compression resistance constraints of the cargo. The same pressure resistance level goods exchange operator means randomly selecting two goods of the same pressure resistance level each time and exchanging their positions. Figure 6 As shown, goods of the same color have the same compression resistance level. Although goods 1 and goods 2 are located on different layers, the compression resistance constraints of the goods are still met after exchanging their positions.

[0039] From the perspective of picking up goods, two exchange operators are adopted: exchange of goods on the same layer and exchange of goods with the same pressure resistance level, so that any two states can be reached within a finite step length, thereby ensuring the diversity of solutions.

[0040] In the specific implementation process of step 303, the process of optimizing the distribution plan in the first-in-last-out mode is similar to the path planning problem of dynamic demand. The optimization process only considers the current deliverable target points, and does not consider the goods under pressure and the delivered target points. After each delivery, new goods may be exposed. The newly exposed goods are regarded as new demands, which are inserted into the current distribution plan with a greedy strategy, and the distribution plan is re-optimized. Therefore, the optimization process of the distribution plan in the first-in-last-out mode of goods in this embodiment is specifically as follows: Step 303.1A, obtain the target point set corresponding to the cargo currently located on the top layer of the carriage , and generate the target point set Initial delivery plan; Step 303.2A, gather at the target point The initial distribution plan is optimized based on the adaptive large neighborhood search iterative framework of the fusion annealing mechanism to obtain the target point set Current delivery options; Step 303.3A, simulate the vehicle according to the target point set The current delivery plan drives to the next destination and delivers the goods; Step 303.4A, determine whether all goods in the carriage have been delivered: If so, the output simulated vehicle route is the delivery plan; Otherwise, return to step 303.1A.

[0041] As for the optimization of the distribution plan under the mode where goods are allowed to be unloaded and then loaded, all goods can be optimized at one time. The specific implementation process includes the following steps: Step 303.1B, obtain the target point set corresponding to all the goods in the carriage , and generate the target point set Initial delivery plan; Step 303.2B, gather at the target point The initial distribution plan is optimized and the distribution plan is obtained by using an adaptive large neighborhood search iterative framework with fusion annealing mechanism.

[0042] It is worth noting that when a vehicle or drone is picking up a cargo, if the cargo is pressed down by other cargo, the cargo on top of the cargo to be delivered needs to be temporarily unloaded, picked up, and then reloaded back into the vehicle. The additional time cost incurred by unloading and reloading is calculated in units of the number of cargoes. The time spent on unloading and reloading each cargo is .like Figure 7 As shown, at the current moment, the additional time cost for loading and unloading the delivery cargo 22 is , the additional time cost of loading and unloading the delivery of goods 34 is 2 .

[0043] In the specific implementation process of step 303.1A and step 303.1B, the process of obtaining the initial delivery plan is the same. The initial delivery plan is generated based on the greedy strategy. First, the vehicle-only delivery path is generated by finding the nearest neighbor method, and then the delivery method of some target points is greedily changed. The drone is used for delivery and its take-off and landing points are determined. The specific implementation process is as follows: Step 401, based on all the target points in the current target point set, simulate the vehicle starting from the base and selecting the nearest target point for delivery in turn until all the target points are delivered, thereby generating a delivery path; Step 402: Based on the current delivery path, calculate the time cost increment caused by changing all vehicle delivery points to drone delivery, and obtain the minimum time cost increment. and its corresponding target point ; Step 403, determine Is it established: If so, set the target point Change to delivery by drone, and update the delivery route and the available time window of the drone collection; Otherwise, output the current delivery route as the initial delivery plan.

[0044] It is worth noting that in step 402, when calculating the time cost increment caused by changing all vehicle delivery points to drone delivery, if there is a vehicle delivery point where the drone has no available time window during the calculation, then the vehicle delivery point will be removed from this iteration, that is, the time cost increment of the vehicle delivery point will not be calculated.

[0045] In the specific implementation process of step 303.2A and step 303.2B, the process of obtaining the distribution plan is the same. This embodiment optimizes the initial distribution plan by designing a fusion algorithm that temporarily does not consider the loading constraints, which is called the annealing adaptive large-scale neighborhood search algorithm. When designing the algorithm, various constraints caused by loading, such as cargo location, pressure resistance, unloading method, unloading order, etc., are relaxed, and only the distribution path of the cargo is considered, that is, the problem is simplified to the MFSTSP problem. The algorithm adopts an adaptive large neighborhood search iterative framework that integrates annealing mechanism. On the one hand, it adaptively adjusts the operator scores and weights to improve the efficiency of solving complex problems. On the other hand, through the Metropolis criterion, it accepts poor solutions with a certain probability, improves the algorithm's exploration ability, and avoids falling into local optimal solutions, in order to improve the performance of the algorithm. The specific implementation process of using the annealing adaptive large-scale neighborhood search algorithm to optimize the distribution plan of goods is as follows: Step 501, obtaining the initial temperature , termination temperature , Temperature reduction coefficient Initial delivery plan , and construct two destruction operators and two repair operators; Step 502, initial current solution 、Current optimal solution , Current temperature , the scores of the four operators are all 0, and the weights of the four operators are all 0.5, that is: operator score = , and initialization, i.e., operator weights = ; Step 503: According to the operator weight Roulette wheel is used to select the destruction operator and repair operator to destroy and repair the current solution and generate a neighbor solution. ; Step 504: Determine the neighbor solution Is it better than the current solution? : If so, , and solve it in the neighbor Update when it is better than the current optimal solution ; Otherwise, based on the Metropolis criteria Is it accepted? If accepted, ,otherwise ; Step 505, based on the current solution Update operator score , and then update the cycle based on the operator weight Determine whether to score according to the operator Update operator weights; Step 506, determine Is it established: If so, output the current optimal solution , as a distribution solution; Otherwise, let Then, return to step 503.

[0046] In the process of calculating the distribution plan, this embodiment designs an adaptive operator selection method, which adjusts the probability of the operator being selected according to the optimization effect of each operator, so that the better the performance of the operator, the higher its score and weight, and the greater the probability of the operator being used. Specifically, the adaptive process mainly includes two parts: 1. Update the operator score according to the effect of the neighbor solution generated by the operator. 2. When the number of iterations reaches the specified weight update cycle, update the selection weight of the operator according to the operator score, and clear the weight score, otherwise the weight is not updated. That is, the specific implementation process of step 505 is: First, according to the current solution Update operator score Specifically: Get the current score of the currently selected destruction operator , and the current score of the previously selected repair operator : If the neighbor solves Updated to the current solution ,but , ; If the neighbor solves Updated to the current solution ,but , ; If the neighbor solves Neither updated to the current solution , and has not been updated to the current solution ,but , ; in, , , are all operator score bonus items, and ; Then, based on the operator weight update cycle Determine whether to score according to the operator Update operator weights, specifically: Get the number of inner iterations of the currently selected destruction operator or the selected repair operator ; make = After +1, judge = Is it established: If yes, update the weight of the destruction operator or the repair operator to: ,in, , is the weight of the destruction operator or repair operator before and after updating, is the current score of the destruction operator or repair operator. is the number of times the destruction operator or repair operator is used. In order to control the influencing factor of the weight change speed, the inner iteration number and operator score of the destruction operator or the repair operator are updated as follows: , ; Otherwise, no operator weight update is performed and the current and .

[0047] In this embodiment, the destruction operator is used to delete from the current solution Target point nodes ( Randomly generated within a given range), the repair operator is used to insert the deleted target point node into a different position from the previous one, generating a new neighbor solution. In view of this feature, this embodiment designs two repair operators and a destruction operator.

[0048] The destruction operator is removed from the current path. There are two ways to delete a target point node in this embodiment: 1. When the deleted node is a drone access point or a vehicle access point for non-drone takeoff / landing, it can be directly deleted from the path, such as Figure 8 (a) as shown; 2. When the deleted node is a vehicle access node for the drone to take off / land, after deletion, a greedy strategy is adopted to find the node closest to the drone among the current vehicle access points as the new take-off / landing point of the drone, such as Figure 8 (b) After clarifying the node deletion method, the following two destruction operators are designed: Random Destruction Operator: Random Selection target nodes, delete them from the current path plan Greedy destruction operator: calculate the time cost reduced when each target point is deleted in turn, and delete the target point that saves the most time until it is deleted. target node.

[0049] The repair operator reinserts the deleted node into the current path. Two insertion methods are also designed: 1. Insert the node into the vehicle path, such as Fig. 9(a); 1. Treat the node as a new UAV access node and find a suitable take-off point and landing point for it, such as Fig. 9 (b) as shown.

[0050] Then two repair operators are designed: Greedy repair operator: For the deleted Calculate the time cost of inserting each node into all possible positions, and insert the node with the smallest time increment into its optimal position until all nodes are reinserted into the path. Regret repair operator: When inserting a node, it may cause the number of available insertion positions for subsequent nodes to decrease, resulting in a greater time cost. Therefore, unlike the greedy repair operator, this operator focuses on the impact of the current inserted node on the subsequent inserted nodes. To indicate the insertion of a node The degree of "regret" is determined by inserting nodes with large regret values ​​into the path first, so that the repair operation has a more global perspective. The calculation process of the regret value is as follows:

[0051] in, , Respectively represent nodes The delivery time increments resulting from inserting the optimal and suboptimal locations are thus available Approximately represent nodes The extra time cost of inserting into a suboptimal position when the optimal position is occupied.

[0052] It is worth noting that in the stage of destroying and repairing the solution using the destruction operator and the repair operator, two solutions that do not meet the constraints may be generated: 1. The take-off point or landing point of the drone path is deleted; 2. When a target point is inserted into the vehicle path, the drone takes longer to wait for the vehicle after taking off, thereby violating the drone's endurance constraint. In summary, both types of infeasible solutions come from the drone path. When an infeasible solution is generated, a solution correction method is used to modify it into a feasible solution, specifically: Find the target point visited by the drone that makes the solution infeasible, and use the two closest vehicle access points as its take-off and landing points. If the drone path is feasible, accept this path, otherwise greedily insert the target point into the vehicle path.

[0053] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. All equivalent structural changes made by using the contents of the present invention specification and drawings under the inventive concept of the present invention, or directly / indirectly applied in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. A method for collaborative path planning between a vehicle and an unmanned aerial vehicle considering three-dimensional loading constraints, characterized in that: The steps include: Step 1, constructing decision variables in the process of collaborative path planning between vehicles and drones, wherein the decision variables determine the vehicle path, the drone path and the cargo loading location; Step 2: construct the optimization goal and constraints of the vehicle-UAV collaborative path planning, wherein the optimization goal is to minimize the delivery time, and the constraints include vehicle path constraints, UAV path constraints, vehicle-UAV collaborative constraints, cargo pressure resistance constraints, vehicle loading constraints, cargo first-in-last-out constraints, and cargo unloading and reloading constraints; Step 3: Under the first-in-last-out mode and the mode where unloading and reloading are allowed, a hybrid metaheuristic solution method is used to calculate the cargo loading plan and the delivery plan. The first-in-last-out mode means that only the cargo on the outermost layer of the vehicle can be delivered each time, and the mode where unloading and reloading are allowed means that the cargo on the inner layer of the vehicle can be delivered.

2. The method for collaborative path planning between a vehicle and an unmanned aerial vehicle considering three-dimensional loading constraints according to claim 1 is characterized in that: In step 1, the decision variables include variables that determine the vehicle path , variables that determine the drone’s path and the variables that determine where the cargo is loaded , specifically: When the vehicle moves from the target point Drive to the destination hour, ,otherwise ; When a drone From the target point Drive to the destination hour, ,otherwise ; If the target point The required cargo is in the carriage When the position ,otherwise .

3. The method for collaborative path planning between a vehicle and an unmanned aerial vehicle considering three-dimensional loading constraints according to claim 2 is characterized in that: Step 3 specifically includes: Step 301, obtaining an initial cargo loading plan for all cargoes; Step 302, iteratively optimizing the initial cargo loading plan to obtain an optimized cargo loading plan; Step 303, based on the optimized cargo loading plan, optimize the distribution plan in the cargo first-in-last-out mode and the cargo unloading and reloading mode respectively.

4. The method for collaborative path planning between a vehicle and an unmanned aerial vehicle considering three-dimensional loading constraints according to claim 3 is characterized in that: Step 301 specifically includes: Step 301.1, obtain the cargo space set in the carriage , cargo collection ,in, is the number of cargo spaces in the carriage, is the quantity of goods, and ; Step 301.2, select the cargo space at any corner at the bottom of the carriage as the current cargo space; Step 301.3, randomly select one from all currently unloaded cargoes with the highest compression resistance level and load it in the current cargo position; Step 301.3, determine whether the storage location on the same floor as the current storage location and adjacent to the current storage location is empty: If yes, use the adjacent cargo location as the current cargo location; Otherwise, the current cargo location will be the one layer above the current cargo location and adjacent to the current cargo location. Step 301.4, determine whether all goods have been loaded: If so, output the current loading plan as the initial cargo loading plan; Otherwise, return to step 301.

3.

5. The method for collaborative path planning between a vehicle and an unmanned aerial vehicle considering three-dimensional loading constraints according to claim 4 is characterized in that: Step 302 specifically includes: Step 302.1, obtain the maximum number of iterations Initial cargo loading plan , and let the current number of iterations 、Current solution 、Current optimal solution ; Step 302.2, use the exchange operator to After transformation, several candidate solutions are obtained. A collection of; Judge 302.3, select the best solution in the candidate solution set , to determine whether there is a : If yes, proceed to step 302.4; Otherwise, proceed to step 302.5; Step 302.4, determine Whether the amnesty rules are met: If yes, proceed to step 302.5; Otherwise, After deleting from the candidate solution set, return to step 302.3; Step 302.5, determine Is it better than : If so, = , = Then, proceed to step 302.6; Otherwise, greedily select a candidate solution from the current candidate solution set as the current solution ; Step 302.6, change the current Add to the taboo list and make , and determine Is it established: If so, output the current As a cargo loading solution; Otherwise, return to step 302.

2.

6. The method for collaborative path planning between a vehicle and an unmanned aerial vehicle considering three-dimensional loading constraints according to claim 5 is characterized in that: The exchange operator comprises: Same-layer cargo exchange operator: Each time, two cargoes on the same layer of the carriage are randomly selected and their positions are exchanged; The exchange operator of goods with the same pressure resistance level: randomly select two goods with the same pressure resistance level each time and exchange their positions.

7. The method for collaborative path planning between a vehicle and an unmanned aerial vehicle considering three-dimensional loading constraints according to any one of claims 3 to 6, characterized in that: Step 303 specifically includes: Under the first-in, last-out mode, the optimization process of the distribution plan is: Step 303.1A, obtain the target point set corresponding to the cargo currently located on the top layer of the carriage , and generate the target point set Initial delivery plan; Step 303.2A, gather at the target point The initial distribution plan is optimized based on the adaptive large neighborhood search iterative framework of the fusion annealing mechanism to obtain the target point set Current delivery options; Step 303.3A, simulate the vehicle according to the target point set The current delivery plan drives to the next destination and delivers the goods; Step 303.4A, determine whether all goods in the carriage have been delivered: If so, the driving route of the simulated vehicle is output as the delivery plan; Otherwise, return to step 303.1A; In the cargo unloading and reloading mode, the optimization process of the distribution plan is: Step 303.1B, obtain the target point set corresponding to all the goods in the carriage , and generate the target point set Initial delivery plan; Step 303.2B, gather at the target point The initial distribution plan is optimized based on the optimization, and the distribution plan is obtained by adopting an adaptive large neighborhood search iterative framework with a fusion annealing mechanism.

8. The method for collaborative path planning between a vehicle and an unmanned aerial vehicle considering three-dimensional loading constraints according to claim 7 is characterized in that: The process of obtaining the initial distribution plan is as follows: Step 401, based on all the target points in the current target point set, simulate the vehicle starting from the base and selecting the nearest target point for delivery in turn until all the target points are delivered, thereby generating a delivery path; Step 402: Based on the current delivery path, calculate the time cost increment caused by changing all vehicle delivery points to drone delivery, and obtain the minimum time cost increment. and its corresponding target point ; Step 403, determine Is it established: If so, set the target point Change to delivery by drone, and update the delivery route and the available time window of the drone collection; Otherwise, the current delivery route is output as the initial delivery plan.

9. The method for collaborative path planning between a vehicle and an unmanned aerial vehicle considering three-dimensional loading constraints according to claim 7, characterized in that: The process of obtaining the distribution plan is as follows: Step 501, obtaining the initial temperature , termination temperature , Temperature reduction coefficient Initial delivery plan , and construct two destruction operators and two repair operators; Step 502, initial current solution 、Current optimal solution , Current temperature , the scores of the four operators are all 0, that is: operator score = , and initialize the weights of the four operators to 0.5, that is, the operator weights = ; Step 503: According to the operator weight Roulette wheel is used to select the destruction operator and repair operator to destroy and repair the current solution and generate a neighbor solution. ; Step 504: Determine the neighbor solution Is it better than the current solution? : If so, , and solve it in the neighbor Update when it is better than the current optimal solution ; Otherwise, based on the Metropolis criteria Whether it is accepted, if it is accepted ,otherwise ; Step 505, based on the current solution Update operator score , and then update the cycle based on the operator weight Determine whether to score according to the operator Update operator weights; Step 506, determine Is it established: If so, output the current optimal solution , as the distribution plan; Otherwise, let Then, return to step 503.

10. The method for collaborative path planning between a vehicle and an unmanned aerial vehicle considering three-dimensional loading constraints according to claim 9, characterized in that: In step 505, the Update operator score Specifically: Get the current score of the currently selected destruction operator , and the current score of the previously selected repair operator : If the neighbor solves Updated to the current solution ,but , ; If the neighbor solves Updated to the current solution ,but , ; If the neighbor solves Neither updated to the current solution , and has not been updated to the current solution ,but , ; in, , , are all operator score bonus items, and ; The operator weight update cycle Determine whether to score according to the operator Update operator weights, specifically: Get the number of inner iterations of the currently selected destruction operator or the selected repair operator ; make = After +1, judge = Is it established: If yes, update the weight of the destruction operator or the repair operator to: ,in, , is the weight of the destruction operator or repair operator before and after updating, is the current score of the destruction operator or repair operator. is the number of times the destruction operator or repair operator is used. In order to control the influencing factor of the weight change speed, the inner iteration number and operator score of the destruction operator or the repair operator are updated as follows: , ; Otherwise, no operator weight update is performed and the current and .

Images