Research On Cooperative Target Assignment And Path Planning Formulti-Unmanned Aircraft System

The multi-unmanned aircraft system(MUAS) cooperative mission planning for combat is a condition-complicated multi-models optimization problem, which can control and optimize the system of heterogeneous UAVs. It is composed of target assignment, path planning and task execution. The challenge of this problem is not only to consider the single UAV capability, constraints and damage probability etc, but also needs to think about the cooperative with other UAVs and avoid conflicts for each other. Moreover, the heterogeneous targets assignment models, real environment of 3D battlefields, random no-fly-zones and radar threats, the sequence of UAVs attacked targets, the cooperative representation of paths and environmental changes during the execution of tasks, are all increase the difficulty of the collaborative planning. Therefore, the key to solve MUAS cooperative is how to implement effective target assignment strategies, plan feasible smooth paths, and adapt to the local dynamics environmental changes.This paper focus on how to solve the difficulties of MUAS cooperative target planning, and how to plan a safe and less cost path for each UAV, which is based on the complex 3D terrain and multi-constraints. A unified model of targets assignment and flight path planning is established based on the optimization theory. The consistent allocation cost matrix is used to process the targets assignment; and the spatial fuzzy sets and cooperative correlations are applied to represent the key way point. Then the targets assignment, path planning and online re-planning of multi UAVs cooperative are mainly studied, as well as it also sets up an efficient task space representation, and improves several optimization methods to solve those problems. Meanwhile, the simulation experiments are explored to verify the feasibility.Firstly, a uniform model has been built for all kinds of UAVs target s assignment with complexity cooperative constrains. According to the relationship between UAVs and targets, the unified framework is extended by the balance and unbalance assignment cases. Then a method based on space vertical section to indicate the cost of flight path length is proposed, which applies the cost matrix of flight path length to optimize the target assignment algorithm. Meanwhile, the collaborative constraints and the violation value of constraints are also studied, so that it can improve the accuracy of target assignment alogrithm.This model unifies the processing of the single UAV to single task, multi UAVs to single task and swarm routing tasks(multi-UAVs travel to multi-tasks). In particular, the cost matrix of swarm routing tasks is constructed to ensure that has same form as others, so that can use a consistent algorithm solving all the assignment problems. So it overcomes the shortcomings of the non uniform model and the less universality.Secondly, to sovle the difficulites problems of the inconsistency of the methods, the larger size of tasks and the planning time significantly increased for the multi-targets assignment of discrete combination optimization and multi models, a novel assignment algorithm with a unified gene coding strategy has been improved, as well as the algorithm is based on a discrete mapping differential evolution(DMDE) in three-dimensional environments. During the solving process, the flight path cost is used to indicate the assignment relationship between the UAV and the target, which turns the optimization problem from discrete space to continuous space by the mapping and inverse mapping rules. Meanwhile, the dynamic crossover rate combines with the hybrid evolution strategy, so that a balance between exploration and exploitation is achieved to avoid falling into a local optimal. The proposed DMDE algorithm with the unified gene coding strategy not only effectively solves the cooperative multi-target assignment problem, but also improves the accuracy of the multi-target assignment. It is also suitable for solving the assignment problem of large scale.Thirdly, in order to sovle the problems of searching space larger and cooperative difficulties of MUAS path planning in 3D environment, the algorithm of MUAS cooperative path planning and smoothing has been researched. The representation of spatial fuzzy set and artificial cultural algorithm are introduced to solve this problem. Firstly the relationship of fuzzy degree of membership between the space points and the path planning is established to represent the space points on 3D grid, so that the key way-points on path should be more attention. Then the belief set of cultural algorithm is constructed by the comprehensive knowledge, which effectively prunes the search space of MUAS path planning. Moreover, the multi-objectives differential evolut ion is used to plan the Pareto optimal paths in the population set of cultural algorithm, which satisfies the constraints of MUAS cooperative. Finally, cultural algorithm exchanges the shared information between the belief set and the population set, so that it keeps the population diversity, accumulates the knowledge and revises the searching direction. This approach enhances the efficiency of key way-points selected, and explores more unknown space to avoid the search fall into local optimization. So it contributes to planning feasible paths for MUAS cooperative quickly.In addition, the method of curve fitting is used to smooth those cooperative paths, which can get safe and available flight paths. The paths are smoothed by the three B-spline curves in 3D space, because the key way-points of paths can be used as the control points of spline curve. Then, in order to avoid the deviations on the smoothing paths, it can inverse find out the spline curve control points that control the smooth path to pass through the key way-points.Finally, a novel path re-planning algorithm with multi Q-learning based on a cooperative fuzzy C means clustering is proposed, which is to solve the failures of tasks or the conflicts of cooperations when the unknown dynamic environment has changed. This approach uses fuzzy degree of membership and co-correlation of way-points to construct a fuzzy cooperative matrix, which can reflect the dynamic changes of planning space by the updating of matrix. Then the cooperative fuzzy C-means clustering is performed with the fuzzy cooperative matrix in every steps of planning space. Furthermore, the task classification of clustering is used as the state space, and the dynamic fuzzy cooperative matrix is used as the reward function of Q-learning. So a multi Q-learning algorithm is proposed to dynamic re-planning the optimization paths in the on-line state space for each task. The method can adapt to the changes of dynamic environment, reduces the task planning space and improves the search efficiency of the learning algorithm, so it can produce effective cooperative re-planning paths for MUAS.