Vision based Modeling and Localization for Planetary Exploration Rovers
Exploration of large unknown planetary environments will rely on rovers that can
autonomously cover distances of kilometres and maintain precise information
about their location with respect to local features.
During such traversals, the rovers will create photo-realistic three dimensional (3D)
models of visited sites for autonomous operations on-site and
mission planning on Earth. Currently rover position is estimated
using wheel odometry, which is sufficient for short traversals but as error
accumulates quickly, it is unsuitable for long distances. At MD Robotics,
we are working on imaging technologies for future planetary rover missions.
Two complementary technologies are currently investigated: a stereo based vision
system and a scanning time-of-flight LIDAR system. Both imaging systems have been installed
on board of two experimental rovers and tested in laboratory and outdoor environments.
With stereo cameras, the rover can create photo-realistic 3D model as well as provide
visual odometry that is more accurate than the rover dead reckoning. With the
LIDAR, the rover can match 3D scans to estimate the relative location to improve
the wheel and visual odometry.