Planetary Environment Prediction Using Generative Modeling

Planetary rovers have a limited sensory horizon and operate in environments where limited information about the surrounding terrain is available. The rough and unknown nature of the terrain in planetary environments potentially leads to scenarios where the rover gets stuck and has to replan its path frequently to escape such situations. For avoiding such scenarios, we need to exploit spatial knowledge of the environment beyond the rover’s sensor horizon. The solutions presented by existing approaches are limited to indoor environments which are structured. Predicting spatial knowledge for outdoor environments, particularly planetary environments, has not be done before. We attempt to solve planetary environment prediction by exploiting generative learning to (1) learn the distribution of spatial landmarks like rocks and craters which the rover encounter on the planetary surface during exploration and (2) predict spatial landmarks beyond the sensor horizon. We aim to utilize the proposed approach of environment prediction to improve path planning and decision-making processes needed for safe planetary navigation.