Rover Self Localization in Planetary-like Environments

The ability for a rover to localize itself with respect to its environment is a crucial issue to tackle autonomous long range navigation. In this paper, we first present and classify the various kind of functionalities a rover should be endowed with to estimate its position during long traverses. We then present a technique that relies on stereovision and pixel tracking to estimate the 6 parameters of the rover displacements, and discuss experimental results obtained with the robot Lama. The paper ends by a brief presentation of a complementary localization function, with respect to an object-based environment model built by the rover as it navigates.