Multi-category commodity parking lot inventory taking target coverage path planning method based on RFID unmanned aerial vehicle

Abstract

The invention discloses a multi-category commodity parking lot inventory target coverage path planning method based on an RFID unmanned aerial vehicle, and the method comprises the steps: firstly, building a mathematical expression of a parking lot environment and a mathematical model of an RFID reader-writer; secondly, clustering vehicle positions by using a full-link clustering algorithm, and calculating a geometric center of each cluster to obtain a target point to be covered; then, an improved particle swarm optimization algorithm is used for initial target coverage path planning to obtain a target coverage sequence; collision detection is carried out between adjacent target points, local obstacle avoidance path planning is carried out on a collision path segment by using an A * algorithm, and obstacle avoidance path points are obtained; and finally, integrating the global target coverage path and the local obstacle avoidance path to obtain a final target coverage path, and completing inventory taking path planning. According to the method, an RFID unmanned aerial vehicle flight path capable of realizing efficient target coverage and avoiding collision can be planned according to the positions of multiple types of commercial vehicles and the obstacle information in the vehicle yard, and the accuracy and the operation efficiency of a checking task are improved.

Description

TECHNICAL FIELD

[0001] The application designs a multi-category commodity parking lot target coverage path planning method based on an RFID unmanned aerial vehicle, which provides a new method for target coverage path planning according to the positions of multi-category commodity vehicles and the obstacle information in the parking lot. It belongs to the field of unmanned aerial vehicle automatic control. BACKGROUND

[0002] In modern logistics and vehicle warehouse management, regular inventory is a key step to ensure inventory accuracy. The traditional manual inventory method is low in efficiency, large in error and high in cost. With the development of unmanned aerial vehicle technology and RFID identification technology, the use of RFID-based unmanned aerial vehicle inventory can realize non-contact rapid identification of vehicles and other goods in the warehouse area. However, in actual application, the efficiency and accuracy of the inventory task are limited by the identification directionality, identification distance, obstacle shielding and vehicle distribution of the RFID reader. Therefore, reasonable planning of the unmanned aerial vehicle inventory path becomes the key. The existing path planning methods are mostly used for continuous area coverage, lack of effective coverage and obstacle avoidance of discrete target points (such as vehicles), and have problems such as high computational complexity and inability to guarantee complete obstacle avoidance in actual application. Therefore, a target coverage path planning method with better performance and higher reliability needs to be proposed.

[0003] The application "a multi-category commodity parking lot target coverage path planning method based on an RFID unmanned aerial vehicle" proposes a low-complexity unmanned aerial vehicle path planning method that can balance target coverage and obstacle avoidance. This method models the parking lot environment and RFID mathematically, and uses an A*-IPSO double-layer optimization framework to realize path generation and optimization, significantly improving obstacle avoidance efficiency and shortening path length. SUMMARY

[0004] (1) Purpose: The purpose of the application is to provide a target coverage path planning method for RFID unmanned aerial vehicles in vehicle inventory tasks. This method can plan an RFID unmanned aerial vehicle flight path that can achieve efficient target coverage and avoid collision according to the positions of multi-category commodity vehicles and the obstacle information in the parking lot, improving the accuracy and efficiency of the inventory task.

[0005] (2) Technical solution: the main content and steps of the "multi-category commodity parking lot target coverage path planning method based on RFID unmanned aerial vehicle" are as follows: first, the mathematical expression of the parking lot environment and the mathematical model of the RFID reader are established, and the target coverage path planning is carried out based on the model; second, the full link clustering algorithm is used to cluster the vehicle position, and the geometric center of each cluster is calculated to obtain the target point to be covered; then the improved particle swarm optimization algorithm is used to plan the initial target coverage path to obtain the target coverage order; then collision detection is carried out between adjacent target points, and the A* algorithm is used for local obstacle avoidance path planning for the path segment that collides, to obtain the obstacle avoidance path point; finally, the global target coverage path and the local obstacle avoidance path are integrated, and the final target coverage path is obtained, and the parking lot path planning is completed.

[0006] The "multi-category commodity parking lot target coverage path planning method based on RFID unmanned aerial vehicle" has the following specific steps:

[0007] Step one: model the parking lot environment. The vehicle and obstacle information are converted into geometric representation in two-dimensional space, and the identification model of the RFID reader is established, and the target coverage path planning mathematical model with the shortest path as the optimization target is constructed.

[0008] Step two: set the clustering threshold based on the RFID identification radius, use the full link clustering algorithm to cluster the vehicle position, calculate the geometric center of each cluster, and obtain the target point to be covered.

[0009] Step three: according to the mathematical model of target coverage path planning established in step one, use the improved particle swarm optimization algorithm to sort the paths of all target points and plan the initial target coverage path. The algorithm improves the quality of the initial solution by introducing the nearest neighbor search strategy, and adds mutation operation in the iteration process to enhance the global search ability and avoid falling into local optimum.

[0010] Step four: collision detection is carried out on the path segment of adjacent target points. For the path segment that detects the existence of obstacles, the A* algorithm is used to generate a local obstacle avoidance path. The obstacle avoidance path point is integrated with the original target point to form a new path point set. Based on the integrated path point set, the improved particle swarm optimization algorithm is executed again to generate the final global target coverage path, thereby completing the path planning of the unmanned aerial vehicle inventory task.

[0011] In step one, the mathematical representation of the vehicle information in the parking lot is as follows:

[0012]

[0013] wherein, represents the vehicle, represents the jth obstacle, represents the jth obstacle,

[0014] We simplify the obstacles in the parking lot as circles in the two-dimensional plane for convenience of calculation, and the mathematical representation of the obstacles is as follows:

[0015]

[0016] wherein, represents the jth obstacle, represents the jth obstacle, represents the radius of the jth obstacle.

[0017] The identification range of the RFID reader is a conical region, and since the vehicles are in a two-dimensional plane, we project the conical region to a circular region in the two-dimensional plane, and the vehicles located in the circular region are considered to be successfully identified, and thus the identification model of the RFID reader is mathematically represented as follows:

[0018]

[0019]

[0020] wherein, represents the position of the vehicle j, represents whether the vehicle is scanned at the target point i, if = 1, it represents that the vehicle j is scanned. represents the Euclidean distance between two target points The set represents all the scanned vehicles.

[0021] Based on the above environment model, a mathematical model of target coverage path planning is established, and the target coverage path planning is regarded as a variant of the traveling salesman problem, with the shortest path as the optimization objective, and the mathematical model is established as follows:

[0022]

[0023]

[0024]

[0025]

[0026] wherein, the objective function (5) defines the shortest path, and represents the height of the vehicle. This constraint ensures that the UAV does not collide with the vehicle during flight. (7) guarantees that all vehicles can be scanned by RFID, improving the accuracy of the inventory. (8) ensures that the UAV can avoid obstacles during flight.

[0027] wherein the all-linkage clustering based on the RFID identification range in step two defines the RFID identification range as a clustering threshold. Each vehicle is initialized as an independent cluster, and all clusters are traversed in each cycle to calculate the all-linkage distance between them, i.e. the maximum distance between any two points in two clusters. If the distance between clusters is less than the threshold, clustering is stopped. After clustering is completed, the geometric center of each cluster is calculated to obtain the position of the target point to be covered.

[0028] wherein the improved particle swarm optimization algorithm in step three introduces a nearest neighbor search to initialize the initial position of the particle to obtain a high-quality initial solution. In view of the problem that the traditional particle swarm optimization algorithm is prone to fall into local optimum, a mutation operation is introduced in the particle swarm evolution process to help the algorithm jump out of the local optimum. The update of the speed and position of the particle is as follows:

[0029]

[0030]

[0031] wherein is an inertia coefficient, is a cognitive coefficient, is a social coefficient, and are random samples uniformly distributed in the range of [0, 1]. is the current speed of the particle, is the current position of the particle, represents the individual historical optimal position of the particle, represents the global optimal position of the particle.

[0032] The quality of the initial particle solution has an important influence on the performance of path planning. The present application introduces a nearest neighbor search strategy in the particle initialization stage to generate a high-quality initial path. Each time the search is performed, a starting point is randomly selected, and according to the greedy strategy, the point closest to the current target point and not covered is selected as the next target point until all target points are covered. A random exchange operation is added in the evolution process to enhance the global search ability, thereby reducing the risk of falling into a local optimum. In view of the problem that the particle swarm optimization algorithm is prone to fall into a local minimum, a simple exchange operation is introduced in the evolution process for random variation, and the order of two target points is randomly exchanged with a certain probability, thereby helping the population to jump out of the local minimum.

[0033] The local obstacle avoidance path planning in step four uses a geometric method to perform collision detection between adjacent path points in the initial target coverage path generated in step three using the improved particle swarm optimization algorithm. The A* algorithm is used to plan an obstacle-free path that avoids the collision. The obstacle avoidance path points are combined with the initial path points to form a new set of path points, and the improved particle swarm optimization algorithm is used again for secondary path planning to obtain a higher-quality target coverage path.

[0034] Technical breakthrough and performance advantages of the present application:

[0035] The technical breakthrough of the present method is that the present application proposes a multi-category commodity parking lot target coverage path planning method based on an RFID unmanned aerial vehicle. This method combines global path planning and local path adjustment mechanisms to generate a flight path that covers all vehicles to be checked and effectively avoids obstacles in the parking lot. Unlike traditional path planning methods that focus on continuous area coverage, the present application models the vehicle checking problem as a discrete target coverage path planning problem and uses an A*-IPSO double-layer optimization framework to solve the shortest path. To overcome the problem of traditional particle swarm algorithms that easily fall into local optimal solutions, the present application introduces a nearest neighbor search strategy to obtain a high-quality initial solution and improves the global search ability of the algorithm through mutation operations during the particle iteration process. In terms of local obstacle avoidance, unlike the traditional method of replacing the collision path length with the obstacle avoidance path length, the present application includes the obstacle avoidance path points generated by the A* algorithm in the overall path planning process, effectively improving the feasibility and global optimality of the path.

[0036] In terms of performance advantages, after clustering, the number of path points and the path length are reduced by about 70% and 46%, respectively. Considering the obstacle avoidance nodes generated by the A* algorithm, the present application can achieve 100% obstacle avoidance. On the basis of ensuring full coverage, the flight path length is shortened by about 4.5% compared to existing representative methods. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 Target coverage path of the present application;

[0038] Figure 2 Target coverage path of the present application after removing the clustering operation;

[0039] Figure 3 Target coverage path of the present application after removing the A* algorithm;

[0040] Figure 4 Ablation experiment of the improved particle swarm optimization algorithm (IPSO);

[0041] Figure 5 Target coverage path planned using other methods.

[0042] Symbol explanation:

[0043] For the coordinates of vehicles in the yard

[0044] For the coordinates of obstacles in the yard

[0045] For the flight height of the UAV

[0046] For the flight speed of the UAV

[0047] For the maximum identification angle of RFID

[0048] For the identification radius of RFID

[0049] For the maximum identification height of RFID

[0050] For the initial target point set

[0051]

[0052]

[0053] For the indicator function, indicating whether the vehicle is scanned

[0054] For the set of all scanned vehicles

[0055] For the Euclidean distance between two points

[0056] For the obstacle avoidance path point set

[0057] For the final target point set

[0058] For the target point coverage sequence

[0059] For the total length of the path

[0060] For the vehicle height

[0061] For the scanning coverage range of the jth target point

[0062] For the position of the kth generation particle

[0063] velocity of the kth particle

[0064] optimal position of the particle itself history

[0065] global optimal position

[0066] inertia coefficient

[0067] cognition coefficient

[0068] social coefficient

[0069] uniformly distributed random number

[0070] uniformly distributed random number DETAILED DESCRIPTION

[0071] The following further describes the design method of each part in the present application:

[0072] The method for counting vehicles using a UAV equipped with an RFID reader has the following specific steps:

[0073] Step 1: Establish a mathematical model

[0074] The known vehicles and obstacle information in the parking lot are mathematically represented. The vehicles are represented as wherein, represents the horizontal and vertical coordinates of the jth vehicle. The obstacles are simplified as circular regions in a two-dimensional plane, represented as wherein, represents the center coordinates and radius of the jth circular obstacle. The warehouse inventory path planning is modeled as target coverage path planning, with the shortest path as the optimization target, and the mathematical model is established as follows:

[0075]

[0076]

[0077]

[0078]

[0079]

[0080] wherein, the objective function (11) defines the shortest path, and the ​represents the height of the vehicle. This constraint ensures that the UAV does not collide with the vehicle during flight. (7) ensures that all vehicles can be scanned by RFID to improve the accuracy of the inventory. (8) ensures that the UAV can avoid obstacles during flight.

[0081] Step two: clustering vehicles to determine the target point positions to be covered

[0082] To achieve effective coverage based on the identification range of RFID, clustering of vehicles is required to reduce the number of target points in path planning and improve planning efficiency. The clustering threshold is set to the effective identification radius of the RFID reader , ensuring that the UAV can complete the inventory of all vehicles in the cluster above the target point. The specific clustering process is as follows:

[0083] 1. Initially, each vehicle is considered as an independent cluster;

[0084] 2. In the clustering process, all current clusters are traversed, and the full connection distance between any two clusters is calculated, i.e., the maximum Euclidean distance between any two points in the two clusters;

[0085] 3. If the full connection distance between two clusters is less than the set threshold R, they are merged into a new cluster;

[0086] 4. Repeat the above merging operation until the full connection distance between any two clusters is not less than , then the clustering process terminates;

[0087] 5. For each final clustering cluster, calculate the geometric center of all vehicle coordinates inside it as the target point to be covered in the UAV flight task.

[0088] Step three: using an improved particle swarm optimization algorithm to solve the initial reference path

[0089] To generate an initial reference path that covers all target points, the improved particle swarm optimization algorithm is used to optimize the coverage order of the target points. The particle update mechanism of this algorithm is as follows:

[0090]

[0091]

[0092] where is the inertia coefficient, is the cognitive coefficient, is the social coefficient, and are random samples uniformly distributed in the range [0, 1]. is the current speed of the particle, is the current position of the particle, represents the best historical position of the individual particle, represents the global optimal position of the particle.

[0093] To address the problem of traditional particle swarm optimization algorithms being prone to getting stuck in local optima, the improved particle swarm optimization algorithm introduces a nearest neighbor search to initialize the initial positions of particles to obtain a high-quality initial solution. Mutation is introduced during the particle swarm evolution process to help the algorithm escape local optima. During each search, a random starting point is selected. Following a greedy strategy, the next target point is chosen, the closest uncovered point to the current target point, until all targets are covered. During the particle swarm evolution process, mutation is introduced to randomly swap the order of two target points with a certain probability, thereby breaking the local convergence of the particle swarm and helping the population escape the local minimum. Ultimately, after 1000 iterations, the path obtained by the particle swarm is the desired initial target coverage path.

[0094] Step 4: Perform collision detection on the initial path, integrate the A* obstacle avoidance path points with the original target points, and solve the final target coverage path, i.e., the inventory path.

[0095] For any pair of adjacent target points in the initial path The line between them is used as a path segment to determine whether the path segment intersects with an obstacle. If a collision occurs, the starting point is and end point The A* algorithm is used as the starting and target positions of the local path planning to search for a collision-free path in the modeled two-dimensional obstacle map. The path is composed of a series of intermediate obstacle avoidance path points to guide the drone around obstacles. The obstacle avoidance path points generated by the A* algorithm are inserted into the original path point sequence, replacing the original path segment where the collision occurred. After all local obstacle avoidance paths are replaced, all target points and newly added obstacle avoidance path points are integrated to form an updated path point set. Subsequently, the improved particle swarm optimization algorithm (IPSO) is re-invoked based on this path point set to perform a global optimization of the coverage order to generate the final target coverage path.

[0096] Experimental charts and related data:

[0097] The reliability and feasibility of the present invention have been verified after simulation.

[0098] The effectiveness of the present invention will be demonstrated below from the aspects of the target coverage path of the present invention, the target coverage path after removing the clustering operation and the A* algorithm of the present invention, the ablation experiment of the improved particle swarm optimization algorithm, and the comparison of target coverage path planning with other methods.

Claims

1. A method for planning a target coverage path for inventorying multiple categories of goods in a yard based on RFID drones, characterized by: First, a mathematical representation of the parking lot environment and a mathematical model of the RFID reader are established. Target coverage path planning is performed based on this mathematical model. Second, a full-link clustering algorithm is used to cluster vehicle locations, and the geometric center of each cluster is calculated to obtain the target point to be covered. Then, an improved particle swarm optimization algorithm is used to perform initial target coverage path planning to obtain the target coverage order. Next, collision detection is performed between adjacent target points, and the A* algorithm is used to perform local obstacle avoidance path planning on the path segments where collisions occur, obtaining obstacle avoidance path points. Finally, the global target coverage path and the local obstacle avoidance path are integrated to obtain the final target coverage path, completing the inventory path planning. The specific steps are as follows: Step 1: Model the parking lot environment; convert vehicle and obstacle information into geometric representations in two-dimensional space, establish an RFID reader recognition model, and construct a target coverage path planning mathematical model with the shortest path as the optimization goal; Step 2: Set the clustering threshold based on the RFID identification radius, use the full-link clustering algorithm to cluster the vehicle positions, calculate the geometric center of each cluster, and obtain the target point to be covered; Step 3: Based on the mathematical model of target coverage path planning established in step 1, an improved particle swarm optimization algorithm is used to sort all target points and plan the initial target coverage path; The algorithm improves the quality of the initial solution by introducing the nearest neighbor search strategy and adds mutation operation in the iterative process to enhance the global search capability and avoid falling into the local optimum. In step 4, collision detection is performed on path segments adjacent to target points. For path segments where obstacles are detected, the A* algorithm is used to generate a local obstacle avoidance path. The obstacle avoidance path points are integrated with the original target points to form a new path point set. Based on the integrated path point set, the improved particle swarm optimization algorithm is executed again to generate the final global target coverage path, completing the path planning for the drone inventory task.

2. The method for planning target coverage paths for multi-category inventory in a yard based on RFID drones according to claim 1 is characterized by: The mathematical representation of the vehicle information in the parking lot described in step 1 is as follows: in, Indicates the a car, Indicates the The horizontal and vertical coordinates of the vehicle; The obstacles in the parking lot are uniformly simplified into circles in a two-dimensional plane. The mathematical representation of the obstacles is as follows: in, represents the jth obstacle, represents the horizontal and vertical coordinates of the j-th obstacle, represents the radius of the j-th obstacle; The recognition range of the RFID reader is a conical area. Projecting the conical area onto a two-dimensional plane forms a circular area. Vehicles within this circular area are considered to be successfully identified. The mathematical representation of the RFID reader's recognition model is as follows: in, represents the position of vehicle j, Indicates whether the vehicle is scanned at the target point i. If =1, it means vehicle j is scanned; Represents two target points The Euclidean distance between sets Indicates all scanned vehicles; A mathematical model for target coverage path planning is established. The target coverage path planning is considered as a variant of the traveling salesman problem. The shortest path is taken as the optimization goal. The mathematical model is established as follows: Among them, the objective function formula (5) defines the shortest path, and the formula (6) represents the height of the vehicle; this constraint ensures that the UAV will not collide with vehicles during flight; formula (7) ensures that all vehicles can be scanned by RFID, improving inventory accuracy; formula (8) ensures that the UAV can avoid obstacles during flight.

3. The method for planning a target coverage path for inventorying multiple categories of goods in a yard based on RFID drones according to claim 1 is characterized by: In the full-link clustering based on RFID identification range described in step 2, the RFID identification range is defined as the clustering threshold. Each vehicle is initialized as an independent cluster. All clusters are traversed in each loop, and the full-link distance between them is calculated, that is, the maximum distance between any points in two clusters. Clustering is stopped if the distance between clusters is less than the threshold. After clustering is completed, the geometric center of each cluster is calculated to obtain the location of the target point to be covered.

4. The method for planning a target coverage path for inventorying multiple categories of goods in a yard based on RFID drones according to claim 1 is characterized by: The improved particle swarm optimization algorithm described in step 3 introduces the nearest neighbor search to initialize the initial position of the particles to obtain a high-quality initial solution. The particle velocity and position are updated as follows: in is the inertia coefficient, is the cognitive coefficient, is the social coefficient, and is a random sample uniformly distributed in the range [0,1]; is the current velocity of the particle, is the current position of the particle, represents the best historical position of the individual particle, represents the global optimal position of the particle; In the particle initialization stage, a nearest neighbor search strategy is introduced to generate high-quality initial paths; Each time the search is performed, a starting point is randomly selected, and according to the greedy strategy, the point closest to the current target point and not covered is selected as the next target point until all target points are covered; random exchange operations are added during the evolution process to enhance the global search capability, and simple exchange operations are introduced during the evolution process for random mutation, randomly exchanging the order of two target points with a certain probability, thereby helping the population jump out of the local minimum.

5. The method for planning target coverage paths for multi-category inventory in a parking lot based on RFID drones according to claim 1 is characterized by: The local obstacle avoidance path planning described in step 4 uses a geometric method to detect collisions between adjacent path points in the initial target coverage path generated in step 3 using the improved particle swarm optimization algorithm. The A* algorithm is used to plan an obstacle-free path that can bypass obstacles for the path segment where the collision occurs. The obstacle avoidance path points are combined with the initial path points to form a new path point set, and then the improved particle swarm optimization algorithm is used again for secondary path planning to obtain a higher quality target coverage path.

6. The method for planning a target coverage path for inventorying multiple categories of goods in a yard based on RFID drones according to claim 1 is characterized by: Step 2: Cluster the vehicles and determine the locations of the target points to be covered. The specific clustering process is as follows: S1, initially, each vehicle is considered as an independent cluster; S2, during the clustering process, traverses all current clusters and calculates the full connection distance between any two clusters, that is, the maximum Euclidean distance between any two points in the two clusters; S3, if the full connection distance between two clusters is less than the set threshold R, they are merged into a new cluster; S4, repeat the above merging until the full connection distance between any two clusters is not less than , the clustering process terminates; S5, for each final cluster, calculate the geometric center of all vehicle coordinates within it as the target point to be covered in the UAV flight mission.

7. The method for planning target coverage paths for multi-category inventory in a yard based on RFID drones according to claim 1 is characterized by: Step 4: Perform collision detection on the initial path, integrate the A* obstacle avoidance path points with the original target points, and solve the final target coverage path, i.e., the inventory path. For any pair of adjacent target points in the initial path The line between them is used as a path segment to determine whether the path segment intersects with an obstacle. If a collision occurs, the starting point is and end point The A* algorithm is used as the starting and target positions of the local path planning to search for a collision-free path in the modeled two-dimensional obstacle map. The path consists of a series of intermediate obstacle avoidance path points to guide the drone around obstacles. The obstacle avoidance path points generated by the A* algorithm are inserted into the original path point sequence to replace the original collision path segment. After all local obstacle avoidance paths are replaced, all target points and newly added obstacle avoidance path points are integrated to form an updated path point set. Subsequently, the improved particle swarm optimization algorithm (IPSO) is re-called based on this path point set to perform global optimization on the coverage order to generate the final target coverage path.

Images