[AML]InteractGen: Enhancing Human-Involved Embodied Task Reasoning through LLM-Based Multi-Agent Collaboration
Keywords: Multi-agent System, LLM-based Agent, Autonomous Robot, Human-robot Interaction, Embodied AI
TL;DR: A framework for leveraging LLM-based multi-agent collaboration to enhance embodied robots’ ability to execute text-based tasks in human-populated environments.
Abstract: The growing demand for intelligent assistants capable of functioning in human-centered environments has spurred extensive research into autonomous robotic systems. Traditional service robots and virtual assistants often fall short in real-world task execution due to their limited ability to perform dynamic reasoning, adapt to changing contexts, and collaborate effectively with humans. These limitations become particularly evident in scenarios that require complex decision-making or seamless integration with human collaborators. Recent advancements in Large Language Models have introduced transformative possibilities for enhancing these systems, equipping them with advanced reasoning and natural language interaction capabilities.
In this paper, we present InteractGen, an LLM-powered, proactive autonomous framework designed to seamlessly operate within a physical office environment. Unlike conventional service robots constrained by rigid reasoning frameworks, InteractGen is built upon an innovative multi-agent collaboration paradime. This paradime empowers InteractGen with advanced inferential capabilities, enabling it to understand nuanced contexts, retrieve and utilize relevant information from memory, and collaborate effectively with humans and other agents in dynamic, real-world settings.
InteractGen bridges the gap between virtual intelligence and physical task execution through its ability to engage in proactive behaviors. It not only reacts promptly to explicit instructions but also anticipates potential challenges, seeks supplementary information autonomously, and collaborates with human team members when necessary to ensure the successful completion of complex tasks. This level of reasoning and interaction marks a significant step forward in the development of autonomous systems for human-populated environments.
Our comprehensive evaluation of InteractGen demonstrates its robust performance across a variety of real-world scenarios. The results highlight the effectiveness of the multi-agent collaboration in enabling InteractGen to dynamically adapt to evolving situations, retrieve critical contextual information, and engage in collaborative problem-solving. These capabilities showcase the potential of InteractGen to redefine the role of intelligent assistants, offering a practical, human-centric solution to complex task management in modern office environments.
Submission Number: 15
Loading