Learning A Simulation-based Visual Policy for Real-world Peg In Unseen Holes

TL;DR

This paper introduces a learning-based visual policy for peg-in-hole tasks that generalizes to unseen shapes by decoupling perception and control, enabling rapid adaptation with minimal real-world data.

Contribution

It presents a novel framework combining a segmentation network, virtual sensor, and controller that adapts to unseen shapes with minimal fine-tuning and auto-labeled data collection.

Findings

Achieves 10/10 success rate in electric vehicle charging task

Requires only hundreds of auto-labeled samples for transfer

Demonstrates rapid adaptation in real-world scenarios

Abstract

This paper proposes a learning-based visual peg-in-hole that enables training with several shapes in simulation, and adapting to arbitrary unseen shapes in real world with minimal sim-to-real cost. The core idea is to decouple the generalization of the sensory-motor policy to the design of a fast-adaptable perception module and a simulated generic policy module. The framework consists of a segmentation network (SN), a virtual sensor network (VSN), and a controller network (CN). Concretely, the VSN is trained to measure the pose of the unseen shape from a segmented image. After that, given the shape-agnostic pose measurement, the CN is trained to achieve generic peg-in-hole. Finally, when applying to real unseen holes, we only have to fine-tune the SN required by the simulated VSN+CN. To further minimize the transfer cost, we propose to automatically collect and annotate the data for the SN after one-minute human teaching. Simulated and real-world results are presented under the configurations of eye-to/in-hand. An electric vehicle charging system with the proposed policy inside achieves a 10/10 success rate in 2-3s, using only hundreds of auto-labeled samples for the SN transfer.