Marine Trash Collection: A Multi-Agent, Multi-Target Search

Oceanic trash collection is a challenging problem. The garbage patches in the oceans are a growing concern for scientists, environmentalists and policy makers. In this paper, we formulate the trash collection problem as a search problem with multiple autonomous surface vehicles searching for multiple passively drifting targets. We propose a search strategy with theoretical guarantees on securing the targets under a predefined set of conditions. In addition, our search strategy accounts for uncertainties in target trajectories caused by environmental factors. Our algorithm is validated using a physics based simulator which considers the physical properties of the targets and the influence of environmental factors on them. Our experimental results suggest that our proposed search strategy can approximately capture 76% of the targets in an unbounded region within the first 48 hours of search, for the worst case scenario considered.