Learning from Suboptimal Demonstration via Self-Supervised Reward Regression

TL;DR

This paper introduces a novel self-supervised reward regression method that effectively learns from suboptimal demonstrations, significantly improving reward estimation and policy performance in robotic tasks.

Contribution

The authors develop a new approach that bootstraps suboptimal demonstrations to synthesize optimality-parameterized data for training reward functions, overcoming limitations of previous ranking-based methods.

Findings

Achieves ~0.95 correlation with ground-truth reward

Policy improvements of ~200% over suboptimal demos

Physical robot demonstration with faster, more topspin shots

Abstract

Learning from Demonstration (LfD) seeks to democratize robotics by enabling non-roboticist end-users to teach robots to perform a task by providing a human demonstration. However, modern LfD techniques, e.g. inverse reinforcement learning (IRL), assume users provide at least stochastically optimal demonstrations. This assumption fails to hold in most real-world scenarios. Recent attempts to learn from sub-optimal demonstration leverage pairwise rankings and following the Luce-Shepard rule. However, we show these approaches make incorrect assumptions and thus suffer from brittle, degraded performance. We overcome these limitations in developing a novel approach that bootstraps off suboptimal demonstrations to synthesize optimality-parameterized data to train an idealized reward function. We empirically validate we learn an idealized reward function with ~0.95 correlation with ground-truth reward versus ~0.75 for prior work. We can then train policies achieving ~200% improvement over the suboptimal demonstration and ~90% improvement over prior work. We present a physical demonstration of teaching a robot a topspin strike in table tennis that achieves 32% faster returns and 40% more topspin than user demonstration.