Bayesian Modeling for Decentralized UAV Control and Task Allocation

The current research represents a first step towards developing a decentralized network of small autonomous, intelligent and inexpensive unmanned aerial vehicles (UAV), which could be used for a variety of scientific missions where measurements from a distributed network of nodes could significantly improve the assessment. The paper presents an adaptive airborne sensing platform (e.g., UAV) that could be used to fuse the information acquired through multiple sensing modalities (obtained from sensors located close to or inside the phenomena) to monitor environmental processes over space and time. The sensing platform is able to continuously reconfigure its trajectory according to the circumstances (e.g., continuously evolving scientific phenomena) to optimize the location of individual sensing waypoints. As proof of concept and validation, we will apply the proposed sensing approach to monitoring volcanic plumes. The paper focuses on (i) developing a Bayesian statistical method to create sensor fusion algorithms that is used to model the plume and guide the UAV in an optimal way (ii) implementing a distributed task allocation scheme which is able guide all the UAVs efficiently and optimize the UAV paths for various criteria.