We investigate the use of transformer sequence models as dynamics models (TDMs) for control. We find that TDMs exhibit strong generalization capabilities to unseen environments, both in a few-shot setting, where a generalist TDM is fine-tuned with small amounts of data from the target environment, and in a zero-shot setting, where a generalist TDM is applied to an unseen environment without any further training. Here, we demonstrate that generalizing system dynamics can work much better than generalizing optimal behavior directly as a policy. Additional results show that TDMs also perform well in a single-environment learning setting when compared to a number of baseline models. These properties make TDMs a promising ingredient for a foundation model of control.
@article{arxiv.2305.10912,
title = {A Generalist Dynamics Model for Control},
author = {Ingmar Schubert and Jingwei Zhang and Jake Bruce and Sarah Bechtle and Emilio Parisotto and Martin Riedmiller and Jost Tobias Springenberg and Arunkumar Byravan and Leonard Hasenclever and Nicolas Heess},
journal= {arXiv preprint arXiv:2305.10912},
year = {2023}
}