2511.14988

Total: 1

#1 An Alignment-Based Approach to Learning Motions from Demonstrations [PDF1] [Copy] [Kimi] [REL]

Authors: Alex Cuellar, Christopher K Fourie, Julie A Shah

Learning from Demonstration (LfD) has shown to provide robots with fundamental motion skills for a variety of domains. Various branches of LfD research (e.g., learned dynamical systems and movement primitives) can generally be classified into ''time-dependent'' or ''time-independent'' systems. Each provides fundamental benefits and drawbacks -- time-independent methods cannot learn overlapping trajectories, while time-dependence can result in undesirable behavior under perturbation. This paper introduces Cluster Alignment for Learned Motions (CALM), an LfD framework dependent upon an alignment with a representative ''mean" trajectory of demonstrated motions rather than pure time- or state-dependence. We discuss the convergence properties of CALM, introduce an alignment technique able to handle the shifts in alignment possible under perturbation, and utilize demonstration clustering to generate multi-modal behavior. We show how CALM mitigates the drawbacks of time-dependent and time-independent techniques on 2D datasets and implement our system on a 7-DoF robot learning tasks in three domains.

Subject: Robotics

Publish: 2025-11-19 00:13:17 UTC