Full-Cycle Mission Planning And Online Planning Of Heterogeneous UAV Swarm Under Complex Mission/Environment

Heterogeneous UAV swarm mission planning technology runs through the entire operation cycle of the swarm as an action guide.It has different connotations and faces different challenges under different operation stages,mission requirements and environmental conditions.First of all,in the mission planning stage,the modularization,isomerization,configurability of airborne resources and the development of wide-speed-domain UAV platforms make the cluster mission planning problems at this stage not only involve task allocation,timing scheduling,track planning,etc.The main planning sub-problems,new types of sub-problems such as airborne resource allocation,flight speed domain switching,etc.cannot be ignored.The mission planning problem coupled with multiple sub-problems has a large-scale and complex decision space,and it is urgent to design an efficient optimization solution method.Efficient optimization of planning problems in largescale complex decision spaces.Secondly,the heterogeneous U AV swarm will face complex factors such as unknown environment,complex resource requirements of the target,uncertain target state,and inconsistency of information b etween machines during the operation stage.Efficient collaboration under dynamic and uncertain conditions.Finally,when heterogeneous UAV swarms operate autonomously in complex environments such as cities and mountainous areas,it is often necessary to perform fast task allocation and trajectory planning within a limited time.Distributed mission planning methods with fast convergence characteristics are urgently needed.Synchronous task assignment and online trajectory planning in complex operating environments to ensure the operational efficiency and flight safety of clusters in complex environments.Based on this,this paper conducts research on the mission planning problem of heterogeneous UAV swarms in complex tasks/operation environments.The main wo rk and research results are as follows:(1)The typical scenarios of heterogeneous UAV cooperative operation are described,and the task types,related constraints and optimization objectives involved in the mission planning problem are analyzed.Formal models of multisub-problem coupling mission planning,coalition formation problem,continuous reconnaissance problem,synchronous task assignment and trajectory planning problem are constructed to support the subsequent instantiation representation of problems.A simplified UAV platform model,track estimation method and communication topology model are given,which lays the foundation for subsequent research.(2)Aiming at the mission planning problem of heterogeneous UAV swarms in the mission planning stage,a multi-sub-problem coupled mission planning problem model is proposed,which simultaneously considers task allocation,timing scheduling,trajectory planning,airborne resource allocation,and flight speed domain switching.The multi-sub-problem coupled mission planning leads to an increase in the complexity and scale of the understanding space,in order to achieve efficient optimization of multi-sub-problem coupled planning problems in a largescale complex decision space.A mission planning method based on two-level adaptive variable neighborhood search is proposed,and the corresponding two-level adaptive perturbation and variable neighborhood descent procedures are designed according to the characteristics of the problem,which realizes the intellige nt guidance of the search direction in the iterative process.A mission planning method guided by domain knowledge and optimization experience is proposed.This method ensures the diversity of solutions through the multi-optimization solution set synchronous search mechanism,and simultaneously designs the knowledge-guided search and experience-guided search links.Knowledge and optimization experience guide the search direction collaboratively in the iterative process.The simulation shows that the two proposed methods can effectively deal with the heterogeneous UAV mission planning problem considering the coupling of multiple sub-problems,and at the same time have good performance for small,medium and large-scale mission planning problems.The advantages are more prominent in mission planning problems with large decision spaces.(3)Aiming at the problem of online mission planning of heterogeneous UAV swarms in the swarm operation stage,the research takes the swarm coordinated regional reconnaissance strike as the mission background.Aiming at the problem of distributed coalition formation of heterogeneous UAV swarms in unknown dynamic environments,considering that the swarms do not have any prior information about any prior information,targets with complex resource requirements will dynamically appear in the scene,and at the same time load a variety of airborne resources.UAV swarms need to quickly form mission alliances to respond to target resource requirements and mission time limits.Accordingly,the task processing automaton of the alliance is constructed based on the finite state machine to support the autonomous task coordination of the cluster,and a distributed coalition formation method based on the upper bound confidence interval tree search is proposed,which solves the problem of the distributed formation of the alliance.It can be implemented in a distributed manner,and can be terminated at any time.The simulation shows that it can form an alliance with a platform and a low resource occupancy rate within a limited time,and effectively deal with the problem of coalition formation under large-scale clusters,high task frequency,and large resource requirements..Aiming at the problem of autonomous reconnaissance and decision-making in heterogeneous clusters,considering that limited UAVs will continuously reconnaissance and surveillance of a large number of targets,the target status is not completely observable and the information between aircrafts is inconsistent,a reconnaissance autonomous decision-making method based on deep reinforcement learning is proposed.,the corresponding state,action space,network structure and reward function are designed,and the network is trained by the proximal policy optimization method.Simulations show that the proposed reconnaissance autonomous decision-making algorithm can achieve fair,highfrequency continuous reconnaissance of multiple targets under the condition of target position change and information inconsistency between aircraft.The algorithm can effectively adapt to the random changes of the target,the number of UAVs and the position,which shows that the algorithm has good elastic response and robustness.(4)Aiming at the online task assignment and trajectory planning of heterogeneous UAV swarms under complex tasks/environmental conditions in the swarm operation stage.Considering that there are dense threats and obstacles in the scene,and the task has specific terminal constraints,the UAV should carry out the online task allocation and trajectory planning of the cluster under the premise of satisfying the state,control,and terminal constraints,so as to maximize the utility of the team.Based on this,a distributed mission planning algorithm with no center and fast convergence is proposed.The algorithm is regarded as an inner and outer ring structure.For the outer ring,a centerless auction method based on the optimal task selection mechanism is proposed to solve the inter-machine tasks.For the allocation problem,UAVs bid according to the principle of maximizing marginal utility,and achieve consistent allocation among groups through local coordination.Aiming at the inner loop,an online trajectory planning method based on sequential second-order cone-convex optimization is proposed,a convex-conical threat model is established,and constraints such as state,control,and threat are made convex.The algorithm has the characteristics of rapid convergence.It can effectively support the fast calculation of the marginal utility of the outer ring and the generation of collision-free trajectory.The simulation shows that the proposed algorithm can effectively solve the problem of online task assignment and trajectory planning of heterogeneous UAV swarms,which has certain engineering prac tical significance.