Programmatic Imitation Learning from Unlabeled and Noisy Demonstrations

TL;DR

PLUNDER is a novel programmatic imitation learning algorithm that effectively learns from noisy demonstrations by integrating probabilistic program synthesis within an EM framework, outperforming existing methods.

Contribution

It introduces a probabilistic program synthesizer within an EM framework for imitation learning from noisy data, enhancing interpretability and robustness.

Findings

PLUNDER achieves 95% demonstration matching accuracy.

It outperforms baselines by 19% in accuracy.

Policies complete tasks 17% more often than nearest baselines.

Abstract

Imitation Learning (IL) is a promising paradigm for teaching robots to perform novel tasks using demonstrations. Most existing approaches for IL utilize neural networks (NN), however, these methods suffer from several well-known limitations: they 1) require large amounts of training data, 2) are hard to interpret, and 3) are hard to repair and adapt. There is an emerging interest in programmatic imitation learning (PIL), which offers significant promise in addressing the above limitations. In PIL, the learned policy is represented in a programming language, making it amenable to interpretation and repair. However, state-of-the-art PIL algorithms assume access to action labels and struggle to learn from noisy real-world demonstrations. In this paper, we propose PLUNDER, a novel PIL algorithm that integrates a probabilistic program synthesizer in an iterative Expectation-Maximization (EM) framework to address these shortcomings. Unlike existing PIL approaches, PLUNDER synthesizes probabilistic programmatic policies that are particularly well-suited for modeling the uncertainties inherent in real-world demonstrations. Our approach leverages an EM loop to simultaneously infer the missing action labels and the most likely probabilistic policy. We benchmark PLUNDER against several established IL techniques, and demonstrate its superiority across five challenging imitation learning tasks under noise. PLUNDER policies achieve 95% accuracy in matching the given demonstrations, outperforming the next best baseline by 19%. Additionally, policies generated by PLUNDER successfully complete the tasks 17% more frequently than the nearest baseline.