A path planning and task scheduling method based on single-vehicle UAV under crowd sensing
By constructing a weighted connected complete graph and a randomized optimization method for path planning, the joint path planning and task scheduling of a single vehicle and a UAV are optimized, which solves the problem of UAV flight distance limitation, improves the efficiency of crowd sensing tasks and reduces costs.
Patent Information
- Application Number
- CN202111472534.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-12-06
- Publication Date
- 2025-09-12
- Estimated Expiration
- 2041-12-06
AI Technical Summary
Drones are limited by their flight distance in swarm perception tasks, resulting in low mission execution efficiency and high costs. The question is how to improve mission completion efficiency and reduce costs through joint path planning and task scheduling between a single vehicle and drones.
A weighted connected complete graph is constructed, and the randomized optimization method for path planning and the joint path planning algorithm are combined to optimize the task scheduling of single vehicles and UAVs, ensuring that UAVs can efficiently perform tasks within the maximum flight distance and reducing the total cost through the collaborative work of vehicles and UAVs.
It has achieved the efficient completion of group intelligence perception tasks in complex environments, reduced task execution costs, and improved task completion efficiency.
Abstract
Description
Technical Field
[0001] This invention belongs to the technical field of path planning and task scheduling for vehicles and drones in crowd-sensing scenarios, and specifically relates to a joint path planning and task scheduling method for a single vehicle and a drone. This method primarily utilizes a single vehicle and the drones deployed on it to improve the efficiency of crowd-sensing task completion while reducing task execution costs through joint optimized scheduling of the vehicle and drones. Background Art
[0002] In recent years, with the rapid development of drone technology, drones can be equipped with small sensors and remotely controlled to complete complex data perception tasks. Furthermore, compared to vehicles, drones can, to a certain extent, break through the mobility limitations of roads and explore more complex and dangerous environments. Therefore, the path planning and task scheduling problems based on drones in crowd sensing scenarios have attracted widespread attention from academia and industry, and have achieved rapid development.
[0003] Drones have the following advantages in crowd sensing scenarios: Drones do not need to consider road conditions when flying, and can explore more complex and even dangerous environments than vehicles; drones do not require human intervention during movement, which can save labor costs; drones have very low energy consumption compared to vehicles and mainly use electricity. However, drones also have many drawbacks. Their maximum flight range is shorter than that of vehicles, and after flying for a period of time, they must return to a transfer station to recharge their batteries before continuing to perform subsequent tasks. This limits the efficiency of using drones to perform crowd sensing tasks. This patent studies how to jointly dispatch single vehicles and drones to collaborate in performing crowd sensing tasks to minimize the cost of completing the task and improve the efficiency of completing the task. Summary of the Invention
[0004] This invention belongs to the technical field of single-vehicle UAV path planning and task scheduling in crowd sensing scenarios. Specifically, it relates to a method for joint path planning and task scheduling based on a single vehicle and UAVs (hereinafter referred to as "this method" unless otherwise stated). This method utilizes a single vehicle and a number of UAVs deployed on it to collaboratively complete crowd sensing tasks within a specific area, improving efficiency and reducing costs. The main components of this method include: a method for constructing a weighted connected complete graph based on the sensing area and target locations; a stochastic optimization method for path planning; and an algorithm for joint path planning and task scheduling for the single vehicle and UAVs.
[0005] This method involves the following environment: within a given perception area, there are n (n³1) target mission locations. Upon arrival at (or above) a mission location, a vehicle or drone can perform data collection using its sensors. Due to the complex environment, some sub-areas within the area are inaccessible to vehicles. If the target mission location falls within one of these sub-areas, the vehicle considers the mission inaccessible and can therefore only complete it using a drone. Each drone is initially deployed on a vehicle and begins its mission from a transfer station (denoted as D). During this time, the drone can separate from the vehicle to independently perform its mission at another mission location, then rendezvous with the vehicle at a given location. While the drone is flying independently, the vehicle can continue driving, allowing both drones to operate simultaneously. The vehicle continuously moves between two adjacent mission locations, while the drone can only take off and land from the vehicle while the vehicle is stationary. Each drone is assumed to have a maximum flight range. When the drone and vehicle rendezvous at the designated location, the first drone to arrive must wait for the other, after which the drone can recharge by swapping batteries. Each mission only needs to be completed once, either by the vehicle or the drone.
[0006] A feasible path plan consists of a vehicle route and one or more drone flight routes. A vehicle route starts and ends at a transfer station, with one or more vehicle-accessible mission locations in between. The starting and ending points of a drone flight route (i.e., the rendezvous point with the vehicle) must both be vertices on the vehicle route, with the starting points being no later than the rendezvous points on the vehicle route. Each drone flight route must satisfy the drone's maximum flight distance constraint. For a subsegment (u, v, w) on a route (whether a vehicle route or a flight route), if v is removed, u will be directly connected to w. If u is the starting point of a flight route, removing u will render the entire route infeasible. Similarly, if w is the end point of a flight route, removing it will also render the route infeasible.
[0007] 1. Method for constructing a weighted connected complete graph based on the sensing area and target location:
[0008] First, for a given set of n target mission locations, for any two points u and v, if they are both within the vehicle-accessible area, the distance between them is the distance between them that can be traveled by the vehicle as the edge weight between the two points, denoted as d(u, v). For a target location u' in the vehicle-unreachable area, the edge weight d(u', x) between it and any other point x (x ≠u') is set to a sufficiently large value M, such as M = max{weights of all edges between two points that can be reached by the vehicle}*n. For drones, any target location is accessible, so the distance between any two points u and v that can be traveled by the drone is the edge weight between them, denoted as d'(u, v). For both drones and vehicles, the weight of the edge (u, v) is the same as the weight of the edge (v, u). The weights d(u, v) and d'(u, v) are used to calculate the time and energy costs of moving along the edge (u, v), respectively, for vehicles and drones.
[0009] 2. Path planning random optimization method:
[0010] For a given path plan S, a method is used to construct a new path plan S'. The specific process is as follows:
[0011] 2.1 Randomly select k mission target points and remove them from the current path plan S. That is, neither the vehicle nor the UAV will visit these target points. If the removal of some target points results in an invalid UAV flight path, all target points on the flight path are removed. Based on this, a new path plan S' is obtained without some mission target points.
[0012] 2.2 For a target point p that was removed in step 2.1, try to reinsert it into the path plan S'. There are three insertion options: i) If p is a target point reachable by the vehicle, try to insert it into every possible position in the vehicle's access path; ii) Try to insert it into any position in each UAV's flight path; iii) If there is an idle UAV on the vehicle, try to dispatch an idle UAV to visit the target point p alone. The starting point and the meeting point with the vehicle of the selected flight path must be vertices on the vehicle's route, and the starting point on the vehicle's route must be no later than the meeting point. Traverse all possible options for these three insertion methods. For each path plan obtained after insertion, check its feasibility, that is, whether it meets the UAV's maximum flight distance constraint. If feasible, calculate its cost. Finally, use the feasible insertion method with the lowest cost to reinsert the target point p into the path plan S' of the single vehicle UAV.
[0013] 2.3 Repeat step 2.2 until all the eliminated task points are reintegrated into the planned route, and a new path plan S' is obtained.
[0014] 3. Algorithms for joint path planning and task scheduling for a single vehicle-UAV:
[0015] Here are the steps:
[0016] 3.1 Randomly generate a single-vehicle path plan on the weighted connected complete graph to visit all mission targets. Then, remove the targets on the path that are unreachable by the vehicle from the path plan and reincorporate them into the path plan using the vertex insertion method described in steps 2.2-2.3 to obtain an initial vehicle-drone path plan S.
[0017] 3.2 Initialize the integer k to 1, T to 0, R to 0, and use S_best to record the path planning solution with the lowest current cost, with its initial value being S;
[0018] 3.3 Optimize S according to the path planning stochastic optimization method described in step 2 to obtain a new vehicle-drone path plan S', and R increases by 1;
[0019] 3.4 Calculate the cost of S' and compare it with S_best. If it is lower than S_best, update S_best to S' and set T to 0.
[0020] 3.5 Compare the cost of S' with that of S. If it is lower than S, update S to S' and set k to 1 and R to 0. Otherwise, if k is less than a preset threshold k_max, increase k by 1.
[0021] 3.6 If R is greater than a preset threshold R_max, then T increases by 1, R is set to 0, k is set to 1, and step 3.1 is executed to update S to a new random path plan;
[0022] 3.7 If T is greater than a preset threshold T_max, the calculation process ends and S_best is output as the final single-vehicle UAV path planning calculation result. Otherwise, return to step 3.3.
[0023] The cost of a route planning includes the fuel cost associated with the driving distance when the vehicle moves, the electricity cost associated with the flight distance when the drone flies, and the labor cost associated with the working time of hiring a driver to drive the vehicle.
Claims
1. A path planning and task scheduling method based on a single-vehicle UAV under swarm intelligence perception, characterized by: Using a single vehicle and a number of drones deployed on it to collaboratively complete swarm sensing tasks within a certain area, the approach improves efficiency and reduces costs. This approach primarily involves constructing a weighted connected complete graph based on the sensing area and target locations, employing a randomized optimization method for path planning, and optimizing algorithms for joint path planning and task scheduling for single-vehicle drones. The method for constructing a weighted connected complete graph based on the sensing area and the target location is as follows: First, for a given n target mission locations, for any two points u and v, if they are both in the vehicle-reachable area, the distance between them is the distance between them that can be traveled by the vehicle as the edge weight between the two points, denoted as d(u,v). For the target location u' in the vehicle-unreachable area, the edge weight d(u',x) between it and any other point x is set to a sufficiently large value M, such as M = max{weights of all edges between two points that can be reached by the vehicle}*n. For drones, any target location is accessible, so the distance between any two points u and v that can be traveled by the drone is the edge weight between the two points, denoted as d'(u,v). Regardless of whether it is a drone or a vehicle, the weight of the edge (u,v) is the same as the weight of the edge (v,u). The weights d(u,v) and d'(u,v) are used as the basis for calculating the time and energy costs of the vehicle and drone moving along the edge (u,v), respectively. The path planning random optimization method is as follows: 1) Randomly select k mission target points and remove them from the current path plan S. That is, neither the vehicle nor the UAV will visit these target points. If the removal of some target points leads to an invalid UAV flight path, all target points on the flight path are removed together. Based on this, a new path plan S' is obtained without some mission target points removed. 2) For a target point p that was removed in step 1), try to reinsert it into the path plan S', including three options: i) If p is a target point reachable by the vehicle, try to insert it into every possible position in the vehicle's access path; ii) Try to insert it into any position in each UAV's flight path; ii) If there is an idle UAV on the vehicle, try to dispatch an idle UAV to visit the target point p alone. The starting point and the meeting point with the vehicle of the selected flight path must be vertices on the vehicle's route, and the starting point on the vehicle's route must be no later than the meeting point. Traverse all possible options of these three insertion methods. For each path plan obtained after insertion, check its feasibility, that is, whether it meets the UAV's maximum flight distance constraint. If feasible, calculate its cost. Finally, use the feasible insertion method with the minimum cost to reinsert the target point p into the path plan of the single vehicle UAV. 3) Repeat step 2) until all the eliminated task points are reintegrated into the planned route, and a new path plan S' is obtained; In the above steps, step 1) is called the random vertex culling method, and steps 2)-3) are called the vertex insertion optimization method.
2. The path planning and task scheduling method based on a single-vehicle UAV under crowd sensing according to claim 1 is characterized by: The optimization algorithm for joint path planning and task scheduling for a single vehicle-UAV is as follows: i. Randomly generate a single-vehicle path plan on the weighted connected complete graph to visit all mission destination points, then remove destination points on the path that are unreachable by the vehicle from the path plan and reincorporate these destination points into the path plan using the vertex insertion optimization method to obtain an initial single-vehicle UAV path plan S; ii. Initialize the integer k to 1, T to 0, and R to 0, and use S_best to record the path planning solution with the lowest current cost, with its initial value S; iii. Optimize S according to the path planning random optimization method to obtain a new vehicle drone path planning S', At this time, R increases by 1; iv. Calculate the cost of S' and compare it with S_best. If it is lower than S_best, update S_best to S' and set T to 0. v. Compare the cost of S' with S. If it is lower than S, update S to S' and set k to 1 and R to 0. Otherwise, if k is less than a preset threshold k_max, increase k by 1. vi. If R is greater than a preset threshold R_max, then T is increased by 1, R is set to 0, k is set to 1, and step i) is executed to update S to a new random path plan; vii. If T is greater than a preset threshold T_max, the calculation process ends and S_best is output as the final single-vehicle UAV path planning calculation result. Otherwise, return to step iii).