Along Similar Lines: Local Obstacle Avoidance for Long-term Autonomous Path Following

Robot navigation in unstructured outdoor environments With the advent of LiDAR implementations of teach and presents a challenging-yet-critical task for many mobile repeat [4]-[6] whose localization is less sensitive to viewpoint robotics applications including transportation, mining, and changes when repeating a path than stereo, we explore forestry. In particular, robust localization in the presence of the possibility of temporarily deviating from the teach path to both short-and long-term scene variations without reliance avoid new obstacles and increase the practicality of VT&R. on a global positioning system (GPS) becomes very difficult. Specifically, we tailor our perception, planning, and control Furthermore, the off-road terrain-assessment problem is nontrivial methods to exploit the teach-and-repeat problem structure to generalize as the variety of potential obstacles increases, while continuing to leverage the human terrain assessment all of which require careful identification, planning, prior whenever possible.