Global Planning for Contact-Rich Manipulation via Local Smoothing of Quasi-dynamic Contact Models

TL;DR

This paper introduces a novel smoothing approach for contact dynamics that enhances model-based planning in contact-rich manipulation, achieving RL-like results with less computation by combining theoretical insights and a convex formulation.

Contribution

It establishes the theoretical equivalence of RL and smoothing methods, introduces a convex quasi-dynamic contact model, and demonstrates effective global planning for complex manipulation tasks.

Findings

Smoothing contact dynamics aligns model-based methods with RL performance.

The convex formulation improves planning efficiency and accuracy.

Global planning with smoothing achieves RL-level results with less computation.

Abstract

The empirical success of Reinforcement Learning (RL) in the setting of contact-rich manipulation leaves much to be understood from a model-based perspective, where the key difficulties are often attributed to (i) the explosion of contact modes, (ii) stiff, non-smooth contact dynamics and the resulting exploding / discontinuous gradients, and (iii) the non-convexity of the planning problem. The stochastic nature of RL addresses (i) and (ii) by effectively sampling and averaging the contact modes. On the other hand, model-based methods have tackled the same challenges by smoothing contact dynamics analytically. Our first contribution is to establish the theoretical equivalence of the two methods for simple systems, and provide qualitative and empirical equivalence on a number of complex examples. In order to further alleviate (ii), our second contribution is a convex, differentiable and quasi-dynamic formulation of contact dynamics, which is amenable to both smoothing schemes, and has proven through experiments to be highly effective for contact-rich planning. Our final contribution resolves (iii), where we show that classical sampling-based motion planning algorithms can be effective in global planning when contact modes are abstracted via smoothing. Applying our method on a collection of challenging contact-rich manipulation tasks, we demonstrate that efficient model-based motion planning can achieve results comparable to RL with dramatically less computation. Video: https://youtu.be/12Ew4xC-VwA