Sample Efficient Robot Learning in Supervised Effect Prediction Tasks

TL;DR

This paper introduces MUSEL, a novel active learning framework that enhances sample efficiency in robotic effect prediction tasks by combining model uncertainty, learning progress, and input diversity, validated through robotic experiments.

Contribution

MUSEL is the first active learning framework specifically designed for continuous, high-dimensional regression tasks in robotics, integrating multiple uncertainty measures.

Findings

MUSEL improves learning accuracy in robotic effect prediction.

MUSEL reduces the number of samples needed for effective learning.

Experimental validation shows MUSEL's effectiveness in real robotic tasks.

Abstract

In self-supervised robotic learning, agents acquire data through active interaction with their environment, incurring costs such as energy use, human oversight, and experimental time. To mitigate these, sample-efficient exploration is essential. While intrinsic motivation (IM) methods like learning progress (LP) are widely used in robotics, and active learning (AL) is well established for classification in machine learning, few frameworks address continuous, high-dimensional regression tasks typical of world model learning. We propose MUSEL (Model Uncertainty for Sample-Efficient Learning), a novel AL framework tailored for regression tasks in robotics, such as action-effect prediction. MUSEL introduces a model uncertainty metric that combines total predictive uncertainty, learning progress, and input diversity to guide data acquisition. We validate our approach using a Stochastic Variational Deep Kernel Learning (SVDKL) model in two robotic tabletop tasks. Experimental results demonstrate that MUSEL improves both learning accuracy and sample efficiency, validating its effectiveness in learning action effects and selecting informative samples.