Autonomous UAV control: Balancing target tracking and persistent surveillance
Lucas W. Krakow, Edwin K. P. Chong · 2017 IEEE Conference on Control Technology and Applications (CCTA) · 2017
Persistent surveillance and target tracking are often accomplished through the use of unmanned aerial vehicles (UAVs). In many scenarios both tasks are concurrently required, but are disparate in their area investigation techniques; surveillance requires global exploration, whereas tracking takes a local target-centric view. Thus, the two tend to require different actions when controlling UAVs tasked with their respective objectives. We propose formulating the UAV control problem for concurrent surveillance and target tracking as a partially observable Markov decision process (POMDP) applying a Q-value approximation technique called nominal belief-state optimization (NBO). Using the Probability Hypothesis Density (PHD) filter as the tracking algorithm, we are able to exploit the birth-intensity components combined with non-myopic action selection inspiring the desired accompanying surveillance behaviors.