RoboChat: A Unified LLM-Based Interactive Framework for Robotic Systems

Abstract

In the rapidly evolving landscape of ro botics, Large Language Models (LLMs) have emerged as a pivotal tool in enhancing the capabilities of robotic systems. This paper introduces a unified LLM-based framework tailored for the embodied AI of robotic systems. Drawing inspiration from the recent advancements in autonomous navigation, interaction, and real-world planning using LLMs, our framework seeks to bridge the gap between high-level linguistic instructions and low-level robotic actions. By integrating open-vocabulary, scene representations, our approach enables robots to navigate and interact with their environment in a more human-like manner. Furthermore, the framework emphasizes safety and adaptability, ensuring that robots can operate seamlessly in dynamic and uncertain environments. Through a combination of real-world planning, the proposed framework offers a holistic solution for the next generation of embodied robotic systems, pushing the boundaries of what robots can perceive, understand, and execute.