Learning the Contact Manifold for Accurate Pose Estimation During Peg-in-Hole Insertion of Complex Geometries

TL;DR

This paper presents a hybrid contact-based pose estimation method for complex peg-in-hole assembly, achieving high accuracy and speed by constructing and aligning contact manifolds, significantly improving success rates over primitive strategies.

Contribution

It introduces a novel hybrid framework that uses contact-state information and a learned projection network to accurately estimate pose during complex assembly tasks.

Findings

Achieves sub-mm and sub-degree pose accuracy.

Attains a 93.3% success rate in industrial geometries.

Runs in under 10 seconds with 95x faster projection.

Abstract

Contact-rich assembly of complex, non-convex parts with tight tolerances remains a formidable challenge. Purely model-based methods struggle with discontinuous contact dynamics, while model-free methods require vast data and often lack precision. In this work, we introduce a hybrid framework that uses only contact-state information between a complex peg and its mating hole to recover the full SE(3) pose during assembly. In under 10 seconds of online execution, a sequence of primitive probing motions constructs a local contact submanifold, which is then aligned to a precomputed offline contact manifold to yield sub-mm and sub-degree pose estimates. To eliminate costly k-NN searches, we train a lightweight network that projects sparse contact observations onto the contact manifold and is 95x faster and 18% more accurate. Our method, evaluated on three industrially relevant geometries with clearances of 0.1-1.0 mm, achieves a success rate of 93.3%, a 4.1x improvement compared to primitive-only strategies without state estimation.