Active Constraint Learning in High Dimensions from Demonstrations

TL;DR

This paper introduces an active learning algorithm within the learning from demonstrations framework that efficiently infers unknown environmental constraints in high-dimensional systems by iteratively selecting informative demonstrations.

Contribution

It proposes a novel iterative active constraint learning method using Gaussian processes to improve constraint inference from sparse demonstrations in high-dimensional settings.

Findings

Outperforms baseline random sampling in simulation and hardware tests

Accurately infers unknown nonlinear constraints from sparse demonstrations

Effective in high-dimensional nonlinear dynamic environments

Abstract

We present an iterative active constraint learning (ACL) algorithm, within the learning from demonstrations (LfD) paradigm, which intelligently solicits informative demonstration trajectories for inferring an unknown constraint in the demonstrator's environment. Our approach iteratively trains a Gaussian process (GP) on the available demonstration dataset to represent the unknown constraints, uses the resulting GP posterior to query start/goal states, and generates informative demonstrations which are added to the dataset. Across simulation and hardware experiments using high-dimensional nonlinear dynamics and unknown nonlinear constraints, our method outperforms a baseline, random-sampling based method at accurately performing constraint inference from an iteratively generated set of sparse but informative demonstrations.