Modelling and Implementation of Unmanned Aircraft Collision Avoidance

With the increasing application of unmanned aircraft in civil airspace, collision avoidance systems for unmanned aircraft are becoming more and more important and valuable. An ideal collision avoidance system gives the aircraft an optimal strategy for choosing flight actions to avoid collision risks when it detects other aircraft nearby. Currently the general approach to generating collision avoidance logics is to model the problem as a partially observable Markov decision process (POMDP), and then synthesize an optimal policy. However, the existing systems require the precise position information of the intruder aircraft to generate the avoidance actions and ignore the effects of the flight path changes, which may result in its lower robustness or a wasting of flying resources.