Versatile Skill Control via Self-supervised Adversarial Imitation of Unlabeled Mixed Motions
Chenhao Li, Sebastian Blaes, Pavel Kolev, Marin Vlastelica, Jonas Frey, Georg Martius
- Year
- 2022
- Access
- Open access
Abstract
Learning diverse skills is one of the main challenges in robotics. To this end, imitation learning approaches have achieved impressive results. These methods require explicitly labeled datasets or assume consistent skill execution to enable learning and active control of individual behaviors, which limits their applicability. In this work, we propose a cooperative adversarial method for obtaining single versatile policies with controllable skill sets from unlabeled datasets containing diverse state transition patterns by maximizing their discriminability. Moreover, we show that by utilizing unsupervised skill discovery in the generative adversarial imitation learning framework, novel and useful skills emerge with successful task fulfillment. Finally, the obtained versatile policies are tested on an agile quadruped robot called Solo 8 and present faithful replications of diverse skills encoded in the demonstrations.
Keywords
Related papers
Trajectory tracking control for 6WID/4WIS UGV via nonlinear sliding mode-model predictive control with adaptive following steering and dynamic-static constraints
Shengyang Lu, Guanpeng Chen, Lijing Zhao +2 more
Robotics and Autonomous Systems · 2026
Bioinspired underwater robotics: Advances across the materials, design, control, and applications
Dilip Muchhala, Pramod Kumar Maurya, Adarsh Raut +3 more
Robotics and Autonomous Systems · 2026
Modeling and control of a rigid–soft hybrid-link humanoid robot
Zewen He, Taiki Ishigaki, Ko Yamamoto
Robotics and Autonomous Systems · 2026
Artificial pushing adaptive coordinated control for the human-exoskeleton-walker system
Xinhao Zhang, Chen Yang, Chaobin Zou +4 more
Robotics and Autonomous Systems · 2026