We consider the problem of simultaneous exploration and information collection in an initially unknown environment by multiple autonomous robots when the communication between robots is unreliable and intermittent. We propose a novel algorithm where decisions to select locations for exploration and information collection are guided by a utility function that combines Gaussian Process-based distributions for information entropy and communication signal strength, along with a distributed coordination protocol to avoid path conflicts between robots and repeated exploration and information collection from the same region by different robots. Our proposed algorithm was experimentally validated in simulation and on hardware Clearpath Jackal robots while using a realistic signal loss model from the literature. Our results show that our approach plans it samples such that it receives up to 25 dBm more signal strength throughout navigation compared to approaches that do not consider communications when selecting locations to sample from while accomplishing similar levels of model error.
- Communications constraints
- Information-driven path planning
- Multi-robot informed path planning
ASJC Scopus subject areas
- Artificial Intelligence