Research On Distributed Multi-UAV Collaborative Search Strategy Based On Probabilistic Gragp Fusion

With the development of drone technology,there are more and more applications that use drones to perform search tasks in complex environments.For example,personnel inspection in public places,lost aircraft search,post-disaster target detection,etc.For a single drone,the battery life is short,and the sensor carrying capacity is limited.This thesis designs a multi-UAV collaborative search strategy,rasterizes the search map,and uses a distributed structure to design a multi-UAV search system.Each drone has the ability to make decisions independently,maintain an updateable information map,and share information.In this thesis,the selforganizing search behavior of the drone is guided by the search probability map independently maintained by the drone.The Bayesian formula is established according to the detection probability and false alarm probability of the sensor empirical model to update the search map information maintained by the drone.In this thesis,the multi-UAV target search problem is divided into two scenarios: the number of search targets is clear and ambiguous.The task goal of searching for a unclear target number scenarios is to quickly reduce the uncertainty of the search map.In this thesis,the objective function of directly optimizing the uncertainty of search map is designed to establish the synergy relationship between drones.The distributed rolling time domain optimization model is used to solve the decision information of each drone,and the Nash equilibrium model is used to iterate the current Nash equilibrium solution at the moment.This thesis designs a time stamp map and time stamp synchronization mechanism for drones.The feasibility of the strategy and the rationality of the model are verified by simulation.The simulation show the strategies presented in this chapter are better than others.The task goal of the clear scene is to determine all the targets within the search map in the shortest amount of time.This thesis adopts the multi-level planning method to model,firstly use the balanced multi-traveler model to carry out multiUAV search path pre-planning,and after the pre-planned search path is completed.UAV will take the multi-level Distributed single traveling salesman model path planning.At the same time,the drone in the communication range can perform the search map information fusion under the time stamp synchronization mechanism.This thesis introduces three probabilistic map fusion strategies and compares their search aging and error rates.Finally,it is verified by simulation that this multi-level distributed search strategy based on information fusion searches all targets in a shorter time than other search strategies.