Stage Conscious Attention Network (SCAN) : A Demonstration-Conditioned Policy for Few-Shot Imitation

TL;DR

This paper introduces SCAN, a novel attention-based policy for few-shot imitation learning that can handle multi-stage, variable-length, and cross-expert demonstrations, significantly improving robotic task performance.

Contribution

The paper proposes a stage conscious attention network (SCAN) that simultaneously retrieves knowledge from diverse demonstrations and learns from different experts without fine-tuning.

Findings

SCAN outperforms baselines in complex compound tasks

It can learn from different experts without fine-tuning

Provides explainable visualization of learned stages

Abstract

In few-shot imitation learning (FSIL), using behavioral cloning (BC) to solve unseen tasks with few expert demonstrations becomes a popular research direction. The following capabilities are essential in robotics applications: (1) Behaving in compound tasks that contain multiple stages. (2) Retrieving knowledge from few length-variant and misalignment demonstrations. (3) Learning from a different expert. No previous work can achieve these abilities at the same time. In this work, we conduct FSIL problem under the union of above settings and introduce a novel stage conscious attention network (SCAN) to retrieve knowledge from few demonstrations simultaneously. SCAN uses an attention module to identify each stage in length-variant demonstrations. Moreover, it is designed under demonstration-conditioned policy that learns the relationship between experts and agents. Experiment results show that SCAN can learn from different experts without fine-tuning and outperform baselines in complicated compound tasks with explainable visualization.