GraspMAS: Zero-Shot Language-driven Grasp Detection with Multi-Agent System

TL;DR

GraspMAS introduces a multi-agent system that leverages natural language to improve robotic grasp detection, especially in complex and cluttered environments, without requiring domain-specific training or fine-tuning.

Contribution

This paper presents a novel multi-agent framework for language-driven grasp detection that enhances reasoning and decision-making in real-world scenarios, outperforming existing methods.

Findings

Significant performance improvements over baseline methods.

Effective in both simulation and real-world robot experiments.

Handles complex language instructions and cluttered environments.

Abstract

Language-driven grasp detection has the potential to revolutionize human-robot interaction by allowing robots to understand and execute grasping tasks based on natural language commands. However, existing approaches face two key challenges. First, they often struggle to interpret complex text instructions or operate ineffectively in densely cluttered environments. Second, most methods require a training or finetuning step to adapt to new domains, limiting their generation in real-world applications. In this paper, we introduce GraspMAS, a new multi-agent system framework for language-driven grasp detection. GraspMAS is designed to reason through ambiguities and improve decision-making in real-world scenarios. Our framework consists of three specialized agents: Planner, responsible for strategizing complex queries; Coder, which generates and executes source code; and Observer, which evaluates the outcomes and provides feedback. Intensive experiments on two large-scale datasets demonstrate that our GraspMAS significantly outperforms existing baselines. Additionally, robot experiments conducted in both simulation and real-world settings further validate the effectiveness of our approach. Our project page is available at https://zquang2202.github.io/GraspMAS