Harnessing Reinforcement Learning for Neural Motion Planning
Marc Hauck, Baris Tura
Technical University of Munich
This is the repository of the project "Harnessing Reinforcement Learning for Neural Motion Planning" for the course Advanced Deep Learning in Robotics. In this work, RL is employed to help a mobile or planar joint robot navigate through an unseen environment around the obstacles within. We make use of different environment representation options and runtime options, the list of which is accessible from the parser script.
Models
models.py/MobileFinder -- Mobile robot
models.py/PlanarFinder -- Planar joint robot with arbitrary DoF
models.py/PlanarFinderImage -- 3 DoF planar joint robot with raw images as environment representation
models.py/PlanarFinderImage2 -- 2 DoF planar joint robot with raw images as environment representation