Agile Catching with Whole-Body MPC and Blackbox Policy Learning

TL;DR

This paper compares model predictive control and reinforcement learning approaches for high-speed object catching in agile robotics, analyzing their performance, robustness, and transferability through extensive hardware experiments.

Contribution

It provides a comparative analysis of classical and learning-based methods for agile robot catching, highlighting their respective advantages and trade-offs.

Findings

Model Predictive Control offers high precision but lower sample efficiency.

Reinforcement Learning demonstrates better robustness and transferability.

Fusing classical and learning methods can enhance overall performance.

Abstract

We address a benchmark task in agile robotics: catching objects thrown at high-speed. This is a challenging task that involves tracking, intercepting, and cradling a thrown object with access only to visual observations of the object and the proprioceptive state of the robot, all within a fraction of a second. We present the relative merits of two fundamentally different solution strategies: (i) Model Predictive Control using accelerated constrained trajectory optimization, and (ii) Reinforcement Learning using zeroth-order optimization. We provide insights into various performance trade-offs including sample efficiency, sim-to-real transfer, robustness to distribution shifts, and whole-body multimodality via extensive on-hardware experiments. We conclude with proposals on fusing "classical" and "learning-based" techniques for agile robot control. Videos of our experiments may be found at https://sites.google.com/view/agile-catching