Diffuse algorithm for robotic multi-source localization
Kathleen McGill, Stephen William Taylor · 2011
This paper describes a new robotic multi-source localization algorithm based on diffusion of robots. Our DIFFUSE algorithm accelerates source localization by combining robot diffusion from purely local communication with random local search. We evaluate our approach on a set of validation benchmarks that provide ground-truth for comparative analysis of multi-source algorithms. Our DIFFUSE algorithm locates all sources in all benchmarks, unlike the BRW, GSO, and HYBRID algorithms tested previously. The algorithm excels when the initial robot deployment is non-uniform because the diffusion of robots facilitates broad exploration of the search space. The performance of the DIFFUSE algorithm improves as the communication range of the robots increases, up to a critical communication range. At this critical range, the robots diffuse across the entire search space, and search performance is independent of the initial robot distribution.