Affine Transport for Sim-to-Real Domain Adaptation
This addresses the problem of adapting dynamics models between simulation and reality for robotics, though it appears incremental as it builds on optimal transport with specific extensions.
The paper tackles sample-efficient domain adaptation in robotics by introducing affine transport, a variant of optimal transport that models mappings between state transition distributions with affine transformations, and shows it significantly reduces model adaptation error in sim-to-sim and sim-to-real experiments.
Sample-efficient domain adaptation is an open problem in robotics. In this paper, we present affine transport -- a variant of optimal transport, which models the mapping between state transition distributions between the source and target domains with an affine transformation. First, we derive the affine transport framework; then, we extend the basic framework with Procrustes alignment to model arbitrary affine transformations. We evaluate the method in a number of OpenAI Gym sim-to-sim experiments with simulation environments, as well as on a sim-to-real domain adaptation task of a robot hitting a hockeypuck such that it slides and stops at a target position. In each experiment, we evaluate the results when transferring between each pair of dynamics domains. The results show that affine transport can significantly reduce the model adaptation error in comparison to using the original, non-adapted dynamics model.