Merging Path Planners and Controllers through Local Context

Proceedings of the AAAI Conference on Artificial Intelligence, 12

Volume

Robotics

Robot Control, Locomotion and Manipulation

This paper presents an implemented approach to robotic tasks involving intermittent contact and changing dynamics in uncertain environments. The approach is to use global planning to find paths in a tesselated representation of the environment, and a set of local controllers to take into account possibly time varying dynamics. The important difference from conventional pathplanning in robotic tasks is how this approach uses local sensory information, and the important difference from reactive or behavior-based approaches is that the local controllers are learnt from simulation models or actual trials and are not programmed in a-priori.