Sim-to-real Transfer of Visuo-motor Policies for Reaching in Clutter: Domain Randomization and Adaptation with Modular Networks

A modular method is proposed to learn and transfer visuo-motor policies from simulation to the real world in an efficient manner by combining domain randomization and adaptation. The feasibility of the approach is demonstrated in a table-top object reaching task where a 7 DoF arm is controlled in velocity mode to reach a blue cuboid in clutter through visual observations. The learned visuo-motor policies are robust to novel (not seen in training) objects in clutter and even a moving target, achieving a 93.3% success rate and 2.2 cm control accuracy.