A Learning-Based Estimation and Control Framework for Contact-Intensive Tight-Tolerance Tasks

TL;DR

This paper introduces a two-stage learning-based framework combining a Bayesian particle filter with MDN and a transformer-enhanced RL controller for contact-intensive, tight-tolerance tasks, demonstrated on bolting tasks in simulation and real-world settings.

Contribution

It proposes a novel integrated estimator and controller framework that effectively handles contact uncertainties and improves generalization in contact-rich manipulation tasks.

Findings

Successful transfer from simulation to real-world bolting tasks

Enhanced estimation accuracy with the MDN-based Bayesian particle filter

Improved control performance using transformer-augmented self-supervised learning

Abstract

We present a two-stage framework that integrates a learning-based estimator and a controller, designed to address contact-intensive tasks. The estimator leverages a Bayesian particle filter with a mixture density network (MDN) structure, effectively handling multi-modal issues arising from contact information. The controller combines a self-supervised and reinforcement learning (RL) approach, strategically dividing the low-level admittance controller's parameters into labelable and non-labelable categories, which are then trained accordingly. To further enhance accuracy and generalization performance, a transformer model is incorporated into the self-supervised learning component. The proposed framework is evaluated on the bolting task using an accurate real-time simulator and successfully transferred to an experimental environment. More visualization results are available on our project website: https://sites.google.com/view/2stagecitt