Research On The Coordinative Planning For Unmanned Aerial And Ground Vehicles In Multi-point Dynamic Aggregation Based On Messenger Mechanism

Coordination between unmanned aerial vehicles(UAVs) and unmanned ground vehicles(UGVs) is a novel and challenging research topic, whose great value of application has attracted vast attention of worldwide researchers. Due to the strong complementarities between UAVs and UGVs in many aspects such as moving speed, payload, communication ability, inspection capability, and so on, coordination between UAVs and UGVs can extend the application fields of UAVs and UGVs, and enhance the effectiveness of implementing surveillance, search, rescue tasks and so on.This paper focuses on the hetergeneous coordination between UAVs and UGVs. The potential coordination patterns between UAVs and UGVs are first studied, and a generalized optimization framework is developed to allow the decision-making problems for different types of UAGVSs to be described in a unified way. In addition, the involved key issues for UAGVSs which are used to accomplish the multi-point dynamic aggregation task(MPDAT) are studied intensively. In the task, UAVs are taken as messengers, and UGVs are taken as mobile actuators. UAVs collect task information by flying over the positions of task points periodically, and pass the information to UGVs. UGVs make their decisions according to the newest information received from UAVs to improve the efficiency of task completion. In the paper, three key problems involved in the UAGVS are studied intensively, including path planning of UAVs flying over all UGVs and tasks, task allocation and motion coordination for UGVs under the guidance of UAVs. Efficient methods are designed for each problem, and the performance of the whole system is optimized from the planning level. Finally, the implementability and efficiency of using UAGVS to perform MPDAT is verified by integrated simulation.The main work and comtribution of the paper can be summarized in the following four aspects:(1) The coordination patterns between UAVs and UGVs are analyzed deeply, and a taxonomy for classification of UAGVSs from three dimensions(i.e., functional role, task goal, and decision-making) is proposed. Based on the taxonomy, a generalized decision-making framework is developed to allow the decision-making problems for different types of UAGVSs to be described in a unified way from the perspective of optimization. In the meantime, the computational complexity and real-time problems involved in the decision-imaking are also analyzed. This research provides a general idea and strategy for the building and decision-making of a UAGVS. Finally, the mathematical model of MPDAT is formulated, and the advantages and involved key issues of using UAGVS to perform MPDAT are summarized.(2) The path planning model for curvature-constrained UAVs acting as messenger in MPDAT, namely Dubins traveling salesman problem with neighborhood(DTSPN),is constructed. Dubins paths in the sense of terminal heading relaxation are developed, whichsimplifies the computation of the shortest path between two points with curvature constraint and removes the dependence of Dubins tour on UAV headings. In addition, a boundary-based encoding strategy is also adopted, which greately reduces the number of decision variables, the size of search space as well as the optimization hardness of DTSPN. Based on the two strategies, an efficient algorithm which incorporates the genetic algorithm and an approximate gradient algorithm is proposed. Computational experiments show that the algorithm proposed in this paper, compared with other algorithms, achieves a better tradeoff between solution quality and time cost, and it can generate better solutons within the same computational time.In addition, the DTSP(Dubins Traveling Salesman Problem) in wind fields is considered in view of the great impact of wind on the path of small fixed-wing UAVs. By introducing the notion of virtual moving target, the Dubins path planning problem from an initial configuration to a target in wind field is converted into the problem of finding the minimal root of a transcendental equation. The uniqueness of the root of the transcendental equation is verified under certain conditions. Then, the Dubins tour is derived by using the differential evolution(DE) algorithm to optimize the visiting sequence of waypoints which is encoded by randomkey. This research provides an efficient computation method for the curvature-constrained path planning between two points in wind field, which is of great practical significance.(3) The multi-UGV is a dynamically decoupled system with coupled motion constraints. Based on the characteristic, a distributed model based on the idea of receding horizon control(DRHC) for the motion coordination of multiple UGVs is established. So, the large-scale centralized optimization problem for the whole system is decomposed into a set of small-scale local optimization problems for each UGV subsystem, which greatly reduces the scale of the optimization problem. To solve the DRHC problem, an improved sequential decision-making strategy based on cooperation mechanism is developed. Each UGV not only makes its own decision, but also generates feasible modifications to the decisions of other UGVs which are parameterized using a smaller number of variables. This strategy achieves global optimization in a space with reduced dimension. Simulation results demonstrate that, compared with centralized decision-making and typical sequential decision-making, the proposed method achieves a better tradeoff between the solution quality and time cost, and it can generate shorter paths for UGVs which achieving under the conflict resolution.(4) In view of the dynamic characteristic of MPDAT, the coordination between UAVs and UGVs with time-varying status of targets is analyzed. An efficient task allocation strategy based on heuristic rules for UGVs is proposed, and rapid response of UGVs to the change of task status can be achieved. In addition, the dynamic path planning problem for UAVs with the position constraints of moving UGVs is discussed for the first time, and the corresponding path planning strategies for UAVs are designed in two different scenarios. One is that UAVs know the motion parameters of UGVs in advance, and the other is that UAVs have no information on the motion parameters of UGVs in advance. For the first scenario, the reliablity of the paths of UAVs acting as messenger for moving UGVs is verified in theory. For the second scenario, a search pattern is proposed and verified.