Research On Distributed Multi-UAV Area Search Algorithms

With the development of the UAV(Unmanned Aerial Vehicle)technology,UAV have attracted more and more attention because of their flexible and inexpensive feature.UAV have appeared in military field,disaster rescue field and intelligent transportation,etc.In unknow environment,taking a UAV to replace reconnaissance aircraft can not only avoid unnecessary casualties but also reduce military expenses.In the task of target searching and rescuing,UAV is facing a complex searching scenario which have wide area and uncertain target distribution characteristics.Due to the limitation of the endurance and detection accuracy,it is difficult to use a single UAV to cope with the task requirements in such scenario.It is an effective way to use a multi-UAV cooperative system to improve the adaptability of the complex task.Currently,centralized and distributed method are two main type of multi-UAV cooperative system.A common centralized model is the lead aircraft and wingman collaboration model.It is susceptible to uncertain situation in the actual task.Once the lead-aircraft crash,the entire team of UAVs will be destroyed.For the distributed Multi-UAV model,the system can work as usual even if several UAVs were destroyed.It is more robustness than the centralized model.This dissertation is concentrating on the unknow searching scenario.Probability map is built base on the sensor detection model.It is used to describe the distribution of the target.This dissertation designed a distributed multi-UAV cooperative system framework.This framework is including a model of UAV group environment relationship and a model of UAVs relationship.The model of UAV groups environment relationship is a distributed model.Every single UAV is searching the information of unknown environment independently and updating the local probability map by the information.The model of UAVs relationship is used to describe the way of information interaction between UAVs.Multi-UAV cooperating system is worked by interacting and fusing probability map.During the searching process,the UAV is continuously searching the target location information.UAV is updating the action strategy by its action and information reward.This process is highly compatible with the Markov decision process.This dissertation is taking the MDP to optimize the searching strategies and discussing some of their solve methods.