Jin Cheng, Marin Vlastelica, Pavel Kolev, Chenhao Li, Georg Martius
Corresponding: jin.cheng@inf.ethz.ch
Autonomous Learning Group, Max Planck Institute for Intelligent Systems, Tübingen, Germany
ETH Zürich, Zürich, Switzerland
University of Tübingen, Tübingen, Germany
Despite many successful applications of data-driven control in robotics, extracting meaningful diverse behaviors remains a challenge. Typically, task performance needs to be compromised in order to achieve diversity. In many scenarios, task requirements are specified as a multitude of reward terms, each requiring a different trade-off. In this work, we take a constrained optimization viewpoint on the quality-diversity trade-off and show that we can obtain diverse policies while imposing constraints on their value functions which are defined through distinct rewards. In line with previous work, further control of the diversity level can be achieved through an attract-repel reward term motivated by the Van der Waals force. We demonstrate the effectiveness of our method on a local navigation task where a quadruped robot needs to reach the target within a finite horizon. Finally, our trained policies transfer well to the real 12-DoF quadruped robot, Solo12, and exhibit diverse agile behaviors with successful obstacle traversal.
Behind the box (low l0)
Behind the box (high l0)
On the box (low l0)
On the box (high l0)
Flat ground (low l0)
Flat ground (high l0)
Long wall (low l0)
Long wall (high l0)
Strong style constraint -> move directly towards the target
High diversity -> diverse joint configurations
strong style constraint
low diversity
strong style constraint
high diversity
weak style constraint
low diversity
weak style constraint
high diversity
Detour (left)
Jump onto the box
Detour (right)