A collaborative arc routing scheduling method for single vehicle and multiple UAVs

Abstract

The present invention discloses a method for coordinated arc routing scheduling of a single vehicle and multiple unmanned aerial vehicles, including: taking the completion time of all patrol tasks and all vehicles and unmanned aerial vehicles returning to the parking lot as the minimum as the optimization goal; under the premise of satisfying the constraint conditions, the initial solution of the patrol path planning strategy is produced by a heuristic algorithm; the patrol path planning strategy is iteratively optimized by an optimization method based on a mixture of simulated annealing and large-scale neighborhood search to obtain the optimal patrol path planning strategy for vehicles and unmanned aerial vehicles. The present invention is applied to the field of task planning and scheduling, considering the asymmetry of the road network, establishing a mixed integer programming model for vehicle-machine collaboration for patrol tasks, and considering the actual patrol needs, constructing a neighborhood destruction operator that meets the characteristics of the problem under medium and large-scale conditions, and on the basis of the neighborhood destruction operator, establishing an optimization method based on a large-scale neighborhood search framework combined with simulated annealing to coordinate the arc routing problem of vehicles and unmanned aerial vehicles.

Description

Technical Field

[0001] The present invention relates to the technical field of task planning and scheduling, and in particular to a method for coordinated arc routing scheduling of a single vehicle and multiple unmanned aerial vehicles. Background Art

[0002] In recent years, the construction of transportation systems and the demand for mobility have exploded, leading to increasing congestion on urban roads and increasing pressure on traffic management. Traffic management departments usually use patrol vehicles and ground-based fixed cameras for traffic monitoring and management. However, patrol vehicles are limited by road networks and traffic conditions, and fixed cameras have shortcomings such as small observation range and poor mobility, making it difficult to effectively and quickly obtain traffic section conditions and alleviate traffic congestion.

[0003] As a new type of transportation tool, small drones have gradually attracted people's attention. Drones have the advantages of low cost and good maneuverability and are widely used in combat missions such as reconnaissance. In the field of transportation, drones have also been successfully used in tasks such as package delivery and violation identification. In addition, since drones are not restricted by roads and can carry a variety of information sensing equipment, drones can also be used in tasks such as road patrols and line inspections. However, the endurance of small drones is limited. Although the operation is flexible, the patrol distance is affected by the endurance of drones. Therefore, a better solution is to cooperate with vehicles and drones. The vehicle serves as a mobile platform for drones, charging or replacing batteries for drones during maneuvers. When the vehicle reaches the vicinity of the mission arc, the drone takes off from the car, completes the patrol of the mission arc, and returns to the car for charging. The car can continue to go to the next location to recover the drone during the drone patrol. In addition, the vehicle can also patrol other mission arcs at the same time according to the road network conditions. The combination of cars and drones greatly increases the reconnaissance range of drones, and also breaks through the limitations of vehicles limited by road networks and traffic conditions, providing new methods and new ideas for traffic patrols and reconnaissance.

[0004] As the core of the vehicle-machine collaborative scheduling problem, the collaborative path planning method has been widely studied. However, most of the previous research on the collaborative scheduling of vehicles and drones focused on the "last mile" package delivery. For example, in a certain area, a certain number of ground vehicles and drones form a system to jointly complete the express delivery tasks of all customer points in the area and form a delivery path planning scheme for vehicles and drones. In the process of solving the problem, the access time window of each customer point, the setting of the objective function, the matching relationship between the number of vehicles and drones, heterogeneous combinations, etc. can also be considered. The main feature of this problem is to solve the access point task (that is, the target of the visit is the customer point), which cannot be directly applied to the arc routing problem of traffic patrol, battlefield reconnaissance and other scenarios (that is, the target of the visit is some arc segments in the road network). The arc routing planning problem of vehicle-drone collaboration has certain similarities with the package delivery problem of vehicle-machine collaboration, but they are two completely different problems in essence. This requires targeted design that can be applied to the arc routing planning of the vehicle-machine collaborative system in this problem scenario.

[0005] However, solving the problem of coordinated arc routing between vehicles and drones is very challenging. It is mainly reflected in the following aspects: First, considering the patrol requirements, the vehicle must move along the ground road network, and the drone must traverse the corresponding task arc segment, which adds path constraints to both ground vehicles and drones. Second, the take-off and landing nodes of the drone are at any node in the vehicle path, and how to choose the appropriate take-off and landing nodes is crucial. Third, since the drone serves the arc task on the road network, compared with the point task, a direction dimension is added, which makes it more difficult to formulate the path for the drone. Finally, when the drone cooperates with the ground vehicle, it is also necessary to consider the conditions such as the take-off and landing sequence of the drone. The above specific problems make it quite challenging to solve the problem of coordinated arc routing between vehicles and drones. At present, the existing technology has conducted preliminary research on the problem of coordinated arc routing between vehicles and drones in the fields of power line inspection and urban traffic patrol. However, both studies only considered the combination of one vehicle and one drone, and the designed heuristic algorithms were only experimentally verified on relevant examples of small-scale arc routing planning problems.

[0006] From the above analysis, we can see that most of the current research on the problem of vehicle-UAV collaboration focuses on collaborative planning of point targets. Although some studies have involved the collaborative planning of arc routing between vehicles and UAVs, most of them only focus on one vehicle and one UAV. The proposed methods are only applicable to small-scale examples and require that the UAV must fly along the road network, which is out of touch with the actual situation. However, the problem of large-scale vehicle-UAV collaborative arc routing can be applied to scenarios such as urban road network monitoring and traffic patrols, and has great research and application potential. Summary of the invention

[0007] In view of the above-mentioned deficiencies in the prior art, the present invention provides a single-vehicle and multiple UAV collaborative arc routing scheduling method, and designs a fast and effective hybrid optimization method, which can solve the medium and large-scale vehicle disembarkation and UAV collaborative arc routing problems in the case of asymmetric road networks, and improve the efficiency of solving the arc routing problem in vehicle-computer collaboration.

[0008] To achieve the above object, the present invention provides a single vehicle and multiple UAV collaborative arc routing scheduling method, comprising the following steps:

[0009] Step 1: The optimization goal is to complete all patrol tasks and minimize the time required for all vehicles and drones to return to the parking lot. The objective function is constructed as follows:

[0010]

[0011] Where f(s) is the objective function value when the patrol path planning strategy of the vehicle and the UAV is s, t PC The patrol path planning strategy for the vehicle and the UAV is the completion time of the vehicle returning to the parking lot when s is the time. The patrol path planning strategy for vehicles and drones is the completion time for the last drone to return to the parking lot at s;

[0012] Step 2: Under the premise of satisfying the constraints, the initial solution of the patrol path planning strategy is generated through the heuristic algorithm;

[0013] Step 3: Use a hybrid optimization method based on simulated annealing and large-scale neighborhood search to iteratively optimize the patrol path planning strategy to obtain the optimal patrol path planning strategy for vehicles and drones.

[0014] In one embodiment, in step 2, the constraints include:

[0015] Vehicle path constraints are used to ensure the continuity of the vehicle path, ensuring that the vehicle starts from the parking lot and eventually returns to the parking lot;

[0016] Drone path constraints, including:

[0017] Each flight path of the drone has only one release node and one recovery node;

[0018] The nodes for launching and recovering drones must be nodes in the vehicle path, that is, the collaborative relationship between drones and vehicles is associated with the nodes for launching and recovering drones;

[0019] When the vehicle reaches the release point, only the drones on the vehicle can be released, and when the vehicle reaches the recovery point, only the drones not on the vehicle can be recovered;

[0020] Drone performance constraints, including:

[0021] At most d drones are in flight at the same time;

[0022] Each drone has a maximum flight time limit during a flight;

[0023] Time constraints and task constraints are used to ensure the earliest arrival time of vehicles and drones at each node, and to ensure that each task is visited at least once by drones or vehicles.

[0024] In one embodiment, in step 2, the specific method for generating the initial solution of the patrol path planning strategy by the heuristic algorithm is as follows:

[0025] Under the condition of meeting the constraints, assign the arc tasks to drones or vehicles for execution, and determine the execution direction of each arc task. For each task edge visited by a drone, find the takeoff and landing point pairs that meet the drone's flight range constraint;

[0026] Generate a vehicle path according to the order of drone takeoff and landing points, and insert all the task edges visited by the vehicle into the vehicle path, that is, form the initial solution of the patrol path planning strategy.

[0027] In one embodiment, step 3 specifically includes:

[0028] Step 3.1, obtain the initial temperature T0, the termination temperature T min , the maximum number of iterations I max , the annealing rate γ and the initial solution s0 of the patrol path planning strategy;

[0029] Step 3.2, let s = s0, s' = s0, T ← T0, i = 1, where s is the optimal solution of the patrol path planning strategy, s' is the current solution of the patrol path planning strategy, T is the current temperature, and i is the number of iterations;

[0030] Step 3.3, randomly destroy the current solution s' of the patrol path planning strategy by using a destruction operator, and repair the destroyed current solution by using a repair operator to obtain the repaired intermediate solution s * ;

[0031] Step 3.4, judge whether f(s * ) < f(s') holds:

[0032] If so, let s' = s * , and enter step 3.6;

[0033] Otherwise, enter step 3.5;

[0034] Step 3.5, let Δf1 = (f(s') - f(s *)) / f(s')、 And obtain a random variable ζ between 0 and 1, and determine whether Δf2 > ζ holds:

[0035] If so, let s' = s * , and enter step 3.7;

[0036] Otherwise, enter step 3.7;

[0037] Step 3.6, determine whether f(s * ) < f(s) holds:

[0038] If so, let s = s * , and enter step 3.7;

[0039] Otherwise, enter step 3.7;

[0040] Step 3.7, determine whether i < I max and T > T min holds:

[0041] If so, let i = i + 1, T = γT, and return to step 3.3;

[0042] Otherwise, output the current s as the optimal patrol path planning strategy for the vehicle and the drone.

[0043] In one embodiment, in step 3.3, the random destruction of the current solution s' of the patrol path planning strategy by using the destruction operator is specifically as follows:

[0044] Randomly select a task edge visited by the drone in s', and change its task direction; or

[0045] Randomly select a task edge visited by the drone in s', and delete its takeoff and landing point pair; or

[0046] Randomly select a task in s'. If the task is a drone task, delete the takeoff and landing point pair of this task from the vehicle path; if the task is a vehicle task, delete the two endpoints of this task from the vehicle path; or

[0047] Randomly change a task edge in s', change it from being visited by the drone to being visited by the vehicle, or change it from being visited by the vehicle to being visited by the drone;

[0048] Randomly select two drone task edges in s' that do not belong to the same connected task, and connect all the task edges in these two drone task edges to form a connected task; or

[0049] Randomly select two drone task edges in s', destroy and reorganize them into three drone task edges, and allocate new takeoff and landing point pairs to the three new drone task edges.

[0050] In one embodiment, in step 3.3, the repair operator is used to repair the damaged current solution, and the repaired intermediate solution s is selected. * , specifically:

[0051] The task after the destruction operator operation is reinserted into the vehicle path while meeting the constraints, and the intermediate solution of the patrol path planning strategy is obtained.

[0052] The present invention provides a method for arc routing scheduling of single vehicle and multiple UAVs, which considers the scenario of asymmetric road network in the scenario of vehicle-machine collaboration, making the problem closer to real life, more realistic and operational. A mixed integer programming model is established for vehicle-machine collaboration for patrol tasks, and considering the actual patrol needs, six kinds of neighborhood destruction operators that meet the characteristics of the problem are constructed under medium and large scale conditions. On the basis of the neighborhood destruction operators, an optimization method based on a large-scale neighborhood search framework combined with simulated annealing is established to coordinate the arc routing problem of vehicles and UAVs. At the same time, the influence of UAV parameters on the coordination efficiency is analyzed, which can effectively solve the problem of coordinated planning of arc routing between vehicles and UAVs and improve the coordination efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] 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.

[0054] Figure 1 is a flow chart of a scheduling method in an embodiment of the present invention;

[0055] Figure 2 This is a schematic diagram of variable coding in an embodiment of the present invention;

[0056] Figure 3 It is a flow chart of iterative optimization of patrol path planning strategy based on a hybrid optimization method of simulated annealing and large-scale neighborhood search in an embodiment of the present invention;

[0057] Figure 4 is a schematic diagram of the first destruction operator in an embodiment of the present invention;

[0058] Figure 5 is a schematic diagram of the second destruction operator in an embodiment of the present invention;

[0059] Figure 6 is a schematic diagram of the third destruction operator in an embodiment of the present invention;

[0060] Figure 7 is a schematic diagram of the fourth destruction operator in an embodiment of the present invention;

[0061] Figure 8 is a schematic diagram of the fifth destruction operator in an embodiment of the present invention;

[0062] Fig. 9 is a schematic diagram of the sixth destruction operator in an embodiment of the present invention;

[0063] Fig.10 is a schematic diagram of a repair operator in an embodiment of the present invention;

[0064] Fig.11 This is a comparison chart of results at different UAV speeds when the number of task edges is 10 in an example of an embodiment of the present invention;

[0065] Fig.12 This is a comparison chart of results at different UAV speeds when the number of task edges is 20 in an example of an embodiment of the present invention;

[0066] Fig.13 This is a comparison chart of results at different UAV speeds when the number of task edges is 40 in an example of an embodiment of the present invention;

[0067] Fig.14 This is a comparison chart of results under different numbers of drones when the number of task edges is 10 in an example of an embodiment of the present invention;

[0068] Fig.15 This is a comparison chart of results under different numbers of drones when the number of task edges is 20 in an example of an embodiment of the present invention;

[0069] Fig.16 This is a comparison chart of results under different numbers of drones when the number of task edges is 40 in an example of an embodiment of the present invention.

[0070] 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

[0071] 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.

[0072] It should be noted that all directional indications in the embodiments of the present invention (such as up, down, left, right, front, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0073] In addition, in the present invention, descriptions such as "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0074] In the present invention, unless otherwise clearly specified and limited, the terms "connection", "fixation", etc. should be understood in a broad sense. For example, "fixation" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection, an electrical connection, a physical connection, or a wireless communication connection; it can be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal connection of two elements or the interaction relationship between two elements, unless otherwise clearly defined. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0075] 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.

[0076] At present, the collaborative tasks between vehicles and drones are mainly limited to point target tasks such as package delivery, while the flexibility and environmental adaptability of drones can make them widely used in fields such as power transmission line inspection, remote sensing mapping, auxiliary communications, and traffic inspection. The cooperation between vehicles and drones, that is, the vehicle acts as a charging platform for the drone, can make up for the shortcomings of the limited endurance of patrol drones. Based on this, the present invention considers the asymmetry of the road network, establishes a mixed integer programming model for vehicle-machine collaboration for patrol tasks, and considers the actual patrol needs. Under medium and large scale conditions, 6 kinds of neighborhood destruction operators that meet the characteristics of the problem are constructed. On the basis of the neighborhood destruction operator, an optimization method based on a large-scale neighborhood search framework combined with simulated annealing (LNS-SA, Large Neighborhood Search-Simulated Annealing) is established to perform collaborative planning for the arc routing problem between vehicles and drones. At the same time, the influence of drone parameters on the collaborative efficiency is analyzed.

[0077] This embodiment considers that in the case of asymmetric road networks, a single vehicle and multiple drones collaborate to traverse certain arcs in the road network. It is assumed that a vehicle carries a certain number of drones starting from a parking lot, and releases / recovers drones at certain nodes (road intersections) set in the road network. The drones visit the task edges in the road network, return to the vehicle for charging before the drone's power is exhausted, and finally return to the parking lot after traversing all the task edges in the road network, so as to meet the problem of minimizing the objective function. The objective function of this problem is to minimize the completion time of visiting all patrol tasks and returning to the parking lot by planning the patrol paths of vehicles and drones. In the scenario considered in this embodiment, both drones and vehicles can complete patrol tasks, and vehicles also serve as ground mobile platforms for drones and energy supply stations for replacing drone batteries / charging drones. The battery life of each drone is limited, so the drone needs to return to the vehicle for charging after flying for a period of time. In addition, the movement of vehicles is constrained by the road network, while drones are not constrained by the road network.

[0078] In this embodiment, the following assumptions are made for the convenience of calculation:

[0079] 1. Each drone can be launched / recovered multiple times;

[0080] 2. The vehicle can launch / recover multiple drones at the same time at one node;

[0081] 3. The launch / recovery of drones must be performed at road network nodes;

[0082] 4. After the vehicle releases / recovers the drone, it can wait at the same location, or go to another node to retrieve the drone or launch another one;

[0083] 5. The power consumption of the drone is constant during flight;

[0084] 6. Not considering the vehicle's endurance;

[0085] 7. The time for launching / recovering the drone is negligible.

[0086] The specific symbols used in this embodiment are shown in Table 1 below:

[0087] Table 1 Symbols and definitions

[0088]

[0089] refer to Figure 1 In the specific implementation process, a single vehicle and multiple UAV collaborative arc routing scheduling method in this embodiment specifically includes the following steps:

[0090] Step 1: The optimization goal is to complete all patrol tasks and minimize the time required for all vehicles and drones to return to the parking lot. The objective function is constructed as follows:

[0091]

[0092] Where f(s) is the objective function value when the patrol path planning strategy of the vehicle and the UAV is s, t PC The patrol path planning strategy for the vehicle and the UAV is the completion time of the vehicle returning to the parking lot when s is the time. The patrol path planning strategy for vehicles and drones is the completion time for the last drone to return to the parking lot at s;

[0093] Step 2: Under the premise of satisfying the constraints, the initial solution of the patrol path planning strategy is generated through the heuristic algorithm;

[0094] Step 3: Use a hybrid optimization method based on simulated annealing and large-scale neighborhood search to iteratively optimize the patrol path planning strategy to obtain the optimal patrol path planning strategy for vehicles and drones.

[0095] In this embodiment, the constraints specifically include vehicle path constraints, drone path constraints, time constraints, and task constraints.

[0096] Let the binary variable x ij Indicates whether the vehicle path passes through edge a ij When x ij =1, indicating that the vehicle path passes through edge a ij Otherwise, x ij = 0. Therefore, the vehicle path constraint is:

[0097]

[0098]

[0099]

[0100] Among them, i, j, k are road network intersection nodes, PC represents the road network intersection node where the parking lot is located; constraint (2) ensures the continuity of the vehicle path, and constraints (3) and (4) ensure that the vehicle starts from the parking lot and eventually returns to the parking lot.

[0101] Let binary variable It means that the kth drone is launched from point s and recovered from point e. Based on the above definition, the drone path constraint is:

[0102]

[0103]

[0104] Among them, a s,e For the arc segment connecting points s and e, constraints (5) and (6) ensure that each flight path of the drone has only one release node and one recovery node;

[0105]

[0106]

[0107] Among them, constraints (7) and (8) respectively ensure that the nodes for launching and recovering drones must be nodes in the vehicle path. That is, the cooperative relationship between drones and vehicles depends on the nodes for launching and recovering drones, which is an indispensable and important condition for the cooperative relationship between vehicles and drones.

[0108] The order of launching and recovering drones is also very strictly limited. The conditions for launching a drone must be that it has never been launched or has been recovered since its last launch. If a drone has not been recovered at a node, then it cannot be launched from a vehicle at that node. For a vehicle, when it arrives at a launch point, there must be an idle drone on the vehicle before it can be launched.

[0109]

[0110]

[0111]

[0112] Among them, constraints (9) and (10) stipulate that when the vehicle arrives at the release point, only the drones on the vehicle can be released, and when the vehicle arrives at the recovery point, only the drones not on the vehicle can be recovered. Constraint (11) is the drone quantity constraint, ensuring that at most d drones are in flight at the same time;

[0113] Due to the limitation of drone battery capacity, each drone has a maximum flight time limit during a flight. Assuming that after replacing the battery, the maximum flight time of the drone is P, then for each drone path <s,ce i ,e>, the following constraints should be met:

[0114]

[0115]

[0116] Among them, constraints (12) and (13) give the maximum flight time limit of the drone when the vehicle and the drone arrive at the node first, respectively. When the vehicle arrives at the node first, each path of the drone should satisfy constraint (12); when the drone arrives at the node first, it should satisfy constraint (13).

[0117] The time constraints and task constraints are:

[0118] The time constraints for the vehicle and drone are as follows:

[0119]

[0120]

[0121]

[0122] Among them, constraints (14) and (15) represent the earliest arrival time of the vehicle and drone at each node, respectively. m is an infinite positive number, and constraint (16) states that each task needs to be visited at least once by a drone or vehicle.

[0123] In this embodiment, the specific process of generating the initial solution of the patrol path planning strategy by the heuristic algorithm is as follows: assigning arc tasks to drones or vehicles for execution under the condition that the constraints are met, and determining the execution direction of each arc task, and for each task edge visited by the drone, finding a take-off and landing point pair that meets the drone's range constraints;

[0124] The vehicle path is generated according to the sequence of the drone's take-off and landing points, and all the task edges visited by the vehicle are inserted into the vehicle path, forming the initial solution of the patrol path planning strategy.

[0125] In the specific implementation process, the task encoding method is first designed, and the collaborative planning of vehicles and drones is divided into two layers, the upper layer is for drone task planning, and the lower layer is for vehicle task planning. In addition, multiple neighborhood structures are designed according to the characteristics of the problem, and the solution to the problem is iteratively optimized through the LNS-SA framework.

[0126] In this embodiment, the path information of the vehicle and the UAV is encoded using two matrices, task-infor and vehicle-path. The encoding information of the two matrices is introduced below:

[0127] refer to Figure 2, task-infor is a matrix variable of size n×6, which is used to store task edge information, where n is the number of task edges. Each row of the matrix stores a task edge information, and the position in the matrix is ​​fixed. The first column stores the visitor of the task edge, 1 indicates that the task is performed by a drone, otherwise, -1 indicates that the task is performed by a vehicle. The second column stores the information of whether the task edge belongs to a connection task, 0 indicates that the task edge is an independent task edge, 1, 2, ... and other numbers represent connection edges, and multiple task edges in the same connection task are represented by the same number; the third column represents the order in which the task edge is visited. If the task is a separate task, the number in the third column is 0, if the task is a connection task, the number in the third column is a natural number greater than 0, the numbers of the same connection task are different, and the size of the natural number is used to determine the order of access to the connection task. The fourth to sixth columns are used to store the access direction of the task edge, the start node and the end node. Among them, the fourth column stores the access direction of the task edge, 0 indicates forward direction, and 1 indicates reverse direction. The fifth and sixth columns store the release node and the recovery node of the task edge respectively. If the task edge is assigned to vehicle access (the first column in the row representing the task in task-infor is -1), the start node and the end node are the two endpoints of the task edge respectively; if the task edge is assigned to a drone, the take-off and landing point pairs that satisfy the access task edge (or link edge) are stored, and the same link edge uses the same take-off and landing point pair.

[0128] The size of the vehicle-path matrix is ​​2×N. The first row stores the nodes that the vehicle needs to visit. The nodes that the vehicle needs to visit are obtained from the last two columns of the task-infor matrix. The second row stores the task identifier. The vehicle path is composed of natural number codes, where the parking lot is coded as 0, passing through the road network nodes along the way, and finally returning to the parking lot.

[0129] The above two matrices can uniquely determine the paths of vehicles and drones. According to the information of the task edge, the tasks that each drone needs to perform, the take-off and landing point pairs of each task (i.e., the starting and ending points of a drone path) and the access order of the drone tasks are determined to generate multiple drone sub-paths. Subsequently, the vehicle path is generated according to the take-off and landing point sequence of each drone task, and the task edges assigned to the vehicle are inserted into the existing vehicle path according to the restrictions of the task edge access order. The initial solution of the patrol path planning strategy can be generated in the above way.

[0130] In step 3, this embodiment combines large-scale neighborhood search with simulated annealing, uses large-scale neighborhood search as the algorithm framework, integrates the simulated annealing mechanism, and accepts degenerate solutions with a certain probability, so that the algorithm can get rid of the local optimum and avoid the problem of premature convergence. Figure 3 , step 3 specifically includes the following steps:

[0131] Step 3.1, obtain the initial temperature T0, the termination temperature T min , the maximum number of iterations I max , the annealing rate γ, and the initial solution s0 of the patrol path planning strategy;

[0132] Step 3.2, let s = s0, s' = s0, T ← T0, i = 1, where s is the optimal solution of the patrol path planning strategy, s' is the current solution of the patrol path planning strategy, T is the current temperature, and i is the number of iterations;

[0133] Step 3.3, randomly disrupt the current solution s' of the patrol path planning strategy using the disruption operator, and repair the disrupted current solution using the repair operator, and screen out the intermediate solution s after repair * ;

[0134] Step 3.4, determine whether f(s * ) < f(s') holds:

[0135] If so, let s' = s * , and enter Step 3.6;

[0136] Otherwise, enter Step 3.5;

[0137] Step 3.5, let Δf1 = (f(s') - f(s * )) / f(s'), and obtain a random variable ζ between 0 and 1, and determine whether Δf2 > ζ holds:

[0138] If so, let s' = s * , and enter Step 3.7;

[0139] Otherwise, enter Step 3.7;

[0140] Step 3.6, determine whether f(s * ) < f(s) holds:

[0141] If so, let s = s * , and enter Step 3.7;

[0142] Otherwise, enter Step 3.7;

[0143] Step 3.7, determine whether i < I max and T > T min holds:

[0144] If so, let i = i + 1, T = γT, and return to Step 3.3;

[0145] Otherwise, output the current s as the optimal patrol path planning strategy for the vehicle and the drone.

[0146] In the specific implementation process, this embodiment designs a total of 6 neighborhood destruction operators and 1 repair operator. In step 3.3, a destruction operator is randomly selected to delete the task edge from the original vehicle path, operate on the task edge, and then reinsert the operated task edge into the vehicle path through the repair operator to update the vehicle path.

[0147] The following is combined with Figure 4-10 The specific operation process of each destruction operator and repair operator is explained, and the Figure 4-10 For intuitive representation. Figure 4-10 In the figure, the solid / dashed line between two nodes does not represent the edge in the road network, but represents the shortest path between the two nodes in the road network. The solid line is the vehicle path, the dashed line represents the drone path, and the solid nodes represent the nodes involved in the operation of the selected destruction operator / repair operator. Specifically:

[0148] refer to Figure 4 , is a schematic diagram of the destruction operator in the first example. The operation process of the destruction operator is as follows:

[0149] Randomly select a task edge visited by a drone. If it is a single drone task edge, delete the take-off and landing points of the task edge from the vehicle path, change the direction of the task edge, and swap the take-off and landing points of the drone to change the access direction of the task edge. If it is a link task, directly flip the shortest path of the link task and delete the take-off and landing points of the link task from the vehicle path.

[0150] refer to Figure 5 , is a schematic diagram of the destruction operator in the second step, and the operation process of the destruction operator is:

[0151] Randomly select a UAV task (a single task edge or a connected task containing tasks), delete the take-off and landing point pairs of the selected task in the vehicle path, and randomly reselect a set of take-off and landing point pairs from the take-off and landing point pair set of the task;

[0152] refer to Figure 6 , is a schematic diagram of the destruction operator in the third example. The operation process of the destruction operator is as follows:

[0153] Randomly select a task. If it is a UAV task, delete the take-off and landing point pair of the UAV task from the vehicle; if it is a vehicle task, directly delete the two endpoints of the task from the vehicle path.

[0154] refer to Figure 7 , is a schematic diagram of the destruction operator in No. 4, the operation process of the destruction operator is:

[0155] Randomly select a single task edge for neighborhood operation: If it is a drone task edge, determine whether the task can be assigned to the vehicle. If so, delete the drone path that visits the task edge and assign the task edge to the vehicle. If it is a vehicle task edge, determine whether the task can be assigned to the drone. If so, randomly select a take-off and landing point pair from the take-off and landing point pair set of the task edge to generate a new drone path. In addition, delete the two endpoints / drone take-off and landing point pairs of the selected task edge from the vehicle path;

[0156] refer to Figure 8 , is a schematic diagram of the destruction operator in No. 5. The operation process of the destruction operator is:

[0157] Randomly select two UAV task edges that do not belong to the same connected task, connect all the task edges in these two UAV tasks to form a connected task, and delete the take-off and landing point pairs of the original two UAV tasks in the vehicle path;

[0158] refer to Fig. 9 , is a schematic diagram of the destruction operator in No. 6. The operation process of the destruction operator is:

[0159] Randomly select two drone task edges that do not belong to the same connection task, destroy the connection tasks to which the two task edges belong, and form a new connection task with the selected two task edges; the remaining task edges of the original two connection tasks form two new connection tasks. Select take-off and landing point pairs for the three new tasks respectively to form their own drone paths, and then delete the take-off and landing point pairs of the original two drone tasks in the vehicle path;

[0160] refer to Fig.10 , is a schematic diagram of the operation of the repair operator. The operation process of the repair operator is:

[0161] The task after the destruction operator is reinserted into the vehicle path, and several intermediate solutions of patrol path planning strategies are obtained according to the different insertion positions. One of them is randomly selected as the output intermediate solution s of the repair operator. * .

[0162] The present invention is further described below with reference to specific examples.

[0163] This example designs a set of experiments to compare with LNS-SA, SA (simulated annealing method), VNS (variable neighborhood search method), and TL (tabu search method) to explore the effectiveness of the LNS-SA method in the present invention, and analyzes the two main considerations of drones, drone speed and drone number.

[0164] All programs in this example are run on a computer with a 64-bit Windows 10 operating system, a 2.90GHz Intel Core i7-10700F CPU, and 32.00GB of memory. This example uses three data sets for testing experiments. Table 2 briefly describes each example.

[0165] Table 2 Scale of the example

[0166]

[0167] The relevant parameters of the vehicle and the UAV are set according to the typical situations in actual applications, as shown in Table 3. The experiment sets up 3 vehicles, each of which carries 3 UAVs. The vehicle speed is set to 30km / h, and the UAV speed is set to 35km / h. It is assumed that the maximum flight time of the UAV is 40min and the maximum driving time of the vehicle is 2 hours.

[0168] Table 3 Vehicle and UAV related parameters

[0169]

[0170] The parameter configurations involved in the LNS-SA method are shown in Table 4. For each case, the maximum number of iterations is set to 90000. In the simulated annealing mechanism, the initial temperature is set to 1000, the termination temperature is set to 0.1, and the cooling rate is 0.99. The maximum number of iterations (Markov chain length) of the isothermal process is set to 300.

[0171] Table 4 LNS-SA parameter settings

[0172]

[0173] In order to compare the overall performance of the LNS-SA method with the other three comparison methods, each method was run 10 times and the average value was taken. Table 5-7 shows the average performance of LNS-SA and the other three comparison methods on 27 instances. "Mean (h)" and "CPU Time (s)" are the averages of the patrol total time and CPU running time for 10 times, respectively. For each instance, the average "Mean" of the results obtained by the four methods is compared. At the same time, in order to further compare the performance difference between the LNS-SA proposed in the present invention and other methods, an indicator Gap is used. The calculation formula of this indicator is: Where X represents a comparison method, Mean X and They represent the average values ​​of the solution results of method X and method LNS-SA respectively. The value of Gap represents the degree of deviation between the optimization results of LNS-SA and other comparison methods.

[0174] As can be seen from Tables 5-7, except for the 1st and 3rd instances, the average values ​​of the results obtained by the LNS-SA method are better than those of the other three comparison methods. The overall result of the VNS method is the worst. For each road network scenario (C1, C2, and C3), the optimization effect of LNS-SA is more obvious than that of the comparison method. However, on larger-scale examples, especially those with 40 task edges, LNS-SA requires longer running time. This is because LNS-SA needs to visit different neighborhoods and a sufficiently large search space, which requires more computing time.

[0175] Table 5 Comparison of C1 example methods

[0176]

[0177] Table 6 Comparison of C2 instance methods

[0178]

[0179] Table 7 Comparison of C3 example methods

[0180]

[0181]

[0182] In order to verify the impact of drone speed on experimental results, the following experiments were designed. These experiments were all based on three different scales of instances in the road network scenario C2, namely C2_E1_A, C2_E2_A and C2_E3_A. As with all the above experiments, each example was repeated 10 times and the average value of the results was taken. The drone speed was selected in the range of 25km / h to 45km / h, and a value was taken every 5km / h for comparison experiments. The experimental results are shown in Tables 8 and Figure 11-13 .

[0183] Table 8 Comparison of results at different UAV speeds

[0184]

[0185] Figure 11-13 The following are the experimental results of three examples when the number of task edges is 10, 20 and 40. Figure 11-13 It can be seen that as the speed of the UAV increases, the total time to complete the patrol mission does not change much overall. This is mainly because the scale and size of the example are large, and the total time to complete the patrol mainly depends on the vehicle speed.

[0186] As for the effect of different numbers of drones on the method, three different scale examples, C2_E1_A, C2_E2_A and C2_E3_A, were also used to verify the effect of the number of drones carried by the vehicle on the method. Each experiment was repeated 10 times and the average value was taken. The number of drones was increased from 2 to 6 for comparative experiments. All experimental results are recorded in Tables 9 and Figure 14-16 middle.

[0187] Table 9 Solution results under different numbers of drones

[0188]

[0189]

[0190] Figure 14-16 The following are the experimental results of three examples when the number of task edges is 10, 20 and 40. Figure 14-16 As can be seen from the figure, the number of drones has a significant impact on the performance of the method. Figure 14-16 As shown in the figure, as the number of drones increases, the total patrol time first decreases and then increases, and reaches the minimum patrol time when the number of drones is 3. From this result, it can be inferred that the increase in the number of drones makes the coordination between vehicles and drones more difficult, and due to the scale of the example and the endurance of drones, the vehicle cannot release too many drones at the same time, which makes the redundant drones not play a big role. Therefore, in this example, the number of drones is more appropriate to be 3.

[0191] In summary, based on the needs of road patrol problems and the development of drone technology, the present invention proposes a road patrol planning method based on the collaboration of a single vehicle and multiple drones, taking into account the reality of asymmetric road networks. The present invention first models and analyzes the vehicle-machine arc routing problem, constructs a neighborhood structure that meets the characteristics of the problem based on the model, and proposes a solution method based on the LNS-SA framework, which can effectively solve the vehicle-drone collaborative arc routing problem. A large number of simulation experiments have demonstrated the effectiveness of the method. At the same time, the present invention analyzes the influence of drone parameters on the efficiency of vehicle-machine collaboration under different scale instances. The experimental results show the superiority of LNS-SA on medium-scale and large-scale instances. As the scale of the example increases, the calculation time of LNS-SA also increases. The next step should consider designing a more efficient and more widely applicable optimization method, such as multi-vehicle and multi-machine optimization method, to solve this problem.

[0192] 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 single vehicle and multiple UAV collaborative arc routing scheduling method, It is characterized in that The steps include: Step 1: The optimization goal is to complete all patrol tasks and minimize the time required for all vehicles and drones to return to the parking lot. The objective function is constructed as follows: Where f(s) is the objective function value when the patrol path planning strategy of the vehicle and the UAV is s, t PC The patrol path planning strategy for the vehicle and the UAV is the completion time of the vehicle returning to the parking lot when s is the time. The patrol path planning strategy for vehicles and drones is the completion time for the last drone to return to the parking lot at s; Step 2: Under the premise of satisfying the constraints, the initial solution of the patrol path planning strategy is generated through the heuristic algorithm; Step 3: Use a hybrid optimization method based on simulated annealing and large-scale neighborhood search to iteratively optimize the patrol path planning strategy to obtain the optimal patrol path planning strategy for vehicles and drones.

2. The single vehicle and multiple UAV collaborative arc routing scheduling method according to claim 1, It is characterized in that In step 2, the constraints include: Vehicle path constraints are used to ensure the continuity of the vehicle path, ensuring that the vehicle starts from the parking lot and eventually returns to the parking lot; Drone path constraints, including: Each flight path of the drone has only one release node and one recovery node; The nodes for launching and recovering drones must be nodes in the vehicle path, that is, the collaborative relationship between drones and vehicles is associated with the nodes for launching and recovering drones; When the vehicle reaches the release point, only the drones on the vehicle can be released, and when the vehicle reaches the recovery point, only the drones not on the vehicle can be recovered; Drone performance constraints, including: At most, there are d drones in flight at the same time; Each drone has a maximum flight time limit during a flight; Time constraints and task constraints are used to ensure the earliest time that vehicles and drones arrive at each node, and to ensure that each task is visited at least once by a drone or vehicle.

3. The single vehicle and multiple UAV collaborative arc routing scheduling method according to claim 2, It is characterized in that In step 2, the initial solution of the patrol path planning strategy is generated by the heuristic algorithm, specifically: Assign arc tasks to drones or vehicles to execute when the constraints are met, and determine the execution direction of each arc task. For each task edge visited by the drone, find the take-off and landing point pair that meets the drone's range constraints; The vehicle path is generated according to the sequence of the drone's take-off and landing points, and all the task edges visited by the vehicle are inserted into the vehicle path, forming the initial solution of the patrol path planning strategy.

4. The single vehicle and multiple UAV collaborative arc routing scheduling method according to claim 3, It is characterized in that Step 3 specifically includes: Step 3.1, obtain the initial temperature T 0 , termination temperature T min , maximum number of iterations I max , annealing rate γ and the initial solution s of patrol path planning strategy 0 ; Step 3.2, let s = s 0 s′=s 0 、T←T 0 , i = 1, where s is the optimal solution of the patrol path planning strategy, s′ is the current solution of the patrol path planning strategy, T is the current temperature, and i is the number of iterations; Step 3.3: Use the destruction operator to randomly destroy the current solution s′ of the patrol path planning strategy, and use the repair operator to repair the destroyed current solution, and select the repaired intermediate solution s * ; Step 3.4, determine f(s * )<f(s′) holds true: If so, let s′=s * , and proceed to step 3.6; Otherwise, go to step 3.5; Step 3.5, let Δf 1 =(f(s′)-f(s * )) / f(s′), And obtain the random variable ζ of 0 to 1, and judge Δf 2 > Is ζ established? If so, let s′=s * , and go to step 3.7; Otherwise, go to step 3.7; Step 3.6, determine f(s * )<f(s) holds true: If so, let s = s * , and go to step 3.7; Otherwise, go to step 3.7; Step 3.7, determine if i < I max And T>T min Is it established: If yes, set i=i+1, T=γT, and return to step 3.3; Otherwise, output the current s as the optimal patrol path planning strategy for vehicles and drones.

5. The single vehicle and multiple UAV collaborative arc routing scheduling method according to claim 4, It is characterized in that In step 3.3, the destruction operator is used to randomly destroy the current solution s′ of the patrol path planning strategy, specifically: Randomly select a task edge in s′ visited by the drone and change its task direction; or Randomly select a task edge in s′ that is visited by a UAV, delete its take-off and landing point pair, and reselect the take-off and landing point pair of the task; or Randomly select a task in s′. If the task is a UAV task, delete the take-off and landing point pair of the task from the vehicle path; if the task is a vehicle task, delete the two endpoints of the task from the vehicle path; or Randomly change a task edge in s′, changing it from being visited by a drone to being visited by a vehicle, or changing it from being visited by a vehicle to being visited by a drone; Randomly select two drone task edges in s′ that do not belong to the same link task, and link all the task edges in these two drone task edges to form a link task; or Randomly select two UAV task edges in s′, destroy and reorganize them into three UAV task edges, and assign new take-off and landing point pairs to the three new UAV task edges.

6. The single vehicle and multiple UAV collaborative arc routing scheduling method according to claim 4, It is characterized in that In step 3.3, the repair operator is used to repair the damaged current solution and select the repaired intermediate solution s * , specifically: The task after the destruction operator operation is reinserted into the vehicle path under the condition of meeting the constraints, and an intermediate solution of the patrol path planning strategy is obtained.

Images