Distributionally Robust Control via Stein Variational Inference for Contact-Rich Manipulation

2026-05-18

AI Summary

This work addresses the challenge of insufficient control reliability in contact-rich robotic manipulation due to parametric uncertainty by proposing a distributionally robust optimization approach that integrates Stein variational inference with model predictive control (MPC). The method uniquely incorporates Stein variational inference into the MPC framework to explicitly model task-sensitive parameter uncertainties, enabling flexible and accurate uncertainty characterization while maintaining computational efficiency. Experimental evaluations across diverse contact-intensive tasks demonstrate that, under substantial parametric perturbations, the proposed approach achieves up to a threefold improvement in control robustness compared to existing model-based control strategies, significantly enhancing performance and reliability in uncertain environments.

Abstract

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.

Problem

• distributionally robust control
• contact-rich manipulation
• uncertainty modeling
• model-based control
• task-sensitive uncertainty

Innovation

• Distributionally Robust Control
• Stein Variational Inference
• Contact-Rich Manipulation
• Uncertainty Modeling
• Model-Based Control