Distributionally Robust Control via Stein Variational Inference for Contact-Rich Manipulation
For robotic manipulation, this provides a more reliable model-based control approach that handles uncertainty without requiring large training data, addressing a key limitation of current methods.
This work improves model-based manipulation control by using distributionally robust optimization with Stein variational inference to model parameter uncertainty, achieving up to 3x improved robustness in contact-rich tasks under broad parametric uncertainty, outperforming existing model-based methods.
Reliable robotic manipulation requires control policies that can accurately represent and adapt to uncertainty arising from contact-rich interactions. Modern data-driven methods mitigate uncertainty through large-scale training and computation, and degrade significantly in performance with limited number of training samples. By contrast, classical model-based controllers are computationally efficient and reliable, but their limited ability to represent task-relevant uncertainty can hinder performance in contact-rich interactions. In this work, we propose to expand the capabilities of model-based manipulation control through more flexible uncertainty modeling that retains performance while exactly adapting to uncertainty. Our approach casts the manipulation problem as a distributionally robust control optimization and proposes a novel deterministic formulation based on Stein variational inference that preserves performance while explicitly modeling task-sensitive parameter uncertainty. As a result, the derived controllers are more aware of task sensitivities to uncertainty, yielding high reliability without compromising performance. Experimental results demonstrate up to 3$\times$ improved robustness across a range of contact-rich manipulation tasks under broad parametric uncertainty, outperforming existing model-based control methods.