Planning for Decentralized Formation Flight of UAV fleets in Uncertain Environments with Dec-POMDP

Abstract

The increasingly interest in cooperative flight of unmanned aerial vehicles (UAVs), alongside the development of sophisticated embedded systems, made possible for numerous algorithms to control multi-agent systems of flying vehicles. However, many of these algorithms still require high communications between vehicles or with a central unit, as well as reliable sensors. The Decentralized Partially Observable Markov Decision Process (Dec-POMDP) is a model for decentralized planning of multi-agents systems, considering uncertain observations and low communication reliability. Although this framework has been used for UAV systems, formation flight is still a problem not considered in this scenario. This paper proposes the use of the Dec-POMDP planning algorithm for high-level decision making in UAV formation flight, making use of existent flight controllers in an inner-loop and allowing computational feasibility. Simulation results show that the proposed method allows for decentralized formation tracking with communication restrictions as well as collision avoidance and time-varying formations.