UniTeam: Open Vocabulary Mobile Manipulation Challenge

TL;DR

This paper presents UniTeam, an improved baseline agent for the HomeRobot challenge, focusing on navigation, manipulation, and recognition of open-vocabulary objects in unfamiliar environments, advancing embodied AI research.

Contribution

The work introduces enhancements to perception, navigation, and manipulation modules of the baseline agent, addressing key deficiencies and improving overall performance in open-vocabulary mobile manipulation tasks.

Findings

Reduced misclassification errors in perception

Prevented infinite loop navigation failures

Improved object placement accuracy

Abstract

This report introduces our UniTeam agent - an improved baseline for the "HomeRobot: Open Vocabulary Mobile Manipulation" challenge. The challenge poses problems of navigation in unfamiliar environments, manipulation of novel objects, and recognition of open-vocabulary object classes. This challenge aims to facilitate cross-cutting research in embodied AI using recent advances in machine learning, computer vision, natural language, and robotics. In this work, we conducted an exhaustive evaluation of the provided baseline agent; identified deficiencies in perception, navigation, and manipulation skills; and improved the baseline agent's performance. Notably, enhancements were made in perception - minimizing misclassifications; navigation - preventing infinite loop commitments; picking - addressing failures due to changing object visibility; and placing - ensuring accurate positioning for successful object placement.