Long-Term Search Through Energy Efficiency and Harvesting

Abstract

We study a search problem motivated by our ongoing work on finding radio-tagged invasive fish with an Autonomous Surface Vehicle (ASV). We focus on settings where the fish tend to move along the boundary of a lake. This setting allows us to formulate the problem as a one-dimensional search problem in which the searcher chooses between station keeping and moving so as to maximize the probability of finding the target in a given amount of time without violating its energy-budget. We model the movement of the target as a random-walk and present a closed-form solution for this search problem. Next, we investigate how long-term autonomy can be enabled by energy harvesting. In this case, the search strategy should incorporate the amount of solar energy available at a particular location and particular time. We show how this quantity can be predicted by estimating the geometry of the tree line along the shore. We then obtain the optimal strategy which maximizes the probability of finding the target by formulating the problem as finding the optimal strategy for a Markov Decision Process. Data collected from field experiments validate our approach.