Coordinated Bayesian-Based Bioinspired Plume Source Term Estimation and Source Seeking for Mobile Robots

Abstract

A new nonparametric Bayesian-based motion planning algorithm for autonomous plume source term estimation (STE) and source seeking (SS) is presented in this paper. The algorithm is designed for mobile robots equipped with gas concentration sensors. Specifically, robots coordinate and utilize a Gaussian-plume likelihood model in a Bayesian-based STE process, then they simultaneously search for and navigate toward the source through model based, bioinspired SS methods such as biased-random-walk and surge-casting. Compared with the state-of-the-art Bayesian- and sensor-based STE/SS motion planners, the strategy described takes advantage of coordination between multiple robots and the estimated plume model for faster and more robust SS, rather than rely on direct or filtered sensor measurements. A set of Monte Carlo simulation studies are conducted to compare the performance between the uncoordinated and coordinated algorithms for different robot team sizes and starting conditions. Additionally, the algorithms are validated experimentally through a laboratory-safe, realistic humid-air plume that behaves similar to a gas plume, to test STE and SS using mobile ground robots equipped with humidity sensors. Simulation and experimental results show consistently that the algorithm involving coordination outperforms traditional bioinspired SS algorithms and it is approximately twice as fast as the uncoordinated case. Finally, the plume source is distorted to study the algorithm's limitations and impact on STE and SS, where results show that even for distorted plumes, useful source localization information can be obtained.