Header logo is

Learning to Explore in Motion and Interaction Tasks

2019

Conference Paper

mg


Model free reinforcement learning suffers from the high sampling complexity inherent to robotic manipulation or locomotion tasks. Most successful approaches typically use random sampling strategies which leads to slow policy convergence. In this paper we present a novel approach for efficient exploration that leverages previously learned tasks. We exploit the fact that the same system is used across many tasks and build a generative model for exploration based on data from previously solved tasks to improve learning new tasks. The approach also enables continuous learning of improved exploration strategies as novel tasks are learned. Extensive simulations on a robot manipulator performing a variety of motion and contact interaction tasks demonstrate the capabilities of the approach. In particular, our experiments suggest that the exploration strategy can more than double learning speed, especially when rewards are sparse. Moreover, the algorithm is robust to task variations and parameter tuning, making it beneficial for complex robotic problems.

Author(s): Miroslav Bogdanovic and Ludovic Righetti
Book Title: IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)
Year: 2019
Month: November
Publisher: IEEE

Department(s): Movement Generation and Control
Bibtex Type: Conference Paper (conference)

Event Place: Macau

Links: arXiv

BibTex

@conference{bogdanovic2019learning,
  title = {Learning to Explore in Motion and Interaction Tasks},
  author = {Bogdanovic, Miroslav and Righetti, Ludovic},
  booktitle = {IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
  publisher = {IEEE},
  month = nov,
  year = {2019},
  month_numeric = {11}
}