This is a page that curates AI-related papers published worldwide. All content here is summarized using Google Gemini and operated on a non-profit basis. Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.
The open API of the social robot Misty allows users to customize open-domain interactions, but it remains inaccessible to users without programming experience. In this paper, we introduce AutoMisty, the first multi-agent collaboration framework based on a large-scale language model (LLM). AutoMisty generates executable Misty robot code from natural language commands. It integrates four specialized agent modules that manage task decomposition, assignment, problem solving, and result synthesis, each of which incorporates a two-stage optimization mechanism that includes self-reflection for iterative improvement and human-in-the-loop optimization to align with user preferences. Natural language feedback allows users to iteratively improve tasks, ensuring accurate execution. We designed a set of benchmark tasks spanning four levels of complexity and conducted experiments in a real-world Misty robot environment to evaluate the effectiveness of AutoMisty. Extensive evaluation results demonstrate that AutoMisty consistently generates high-quality code while enabling precise code control, significantly outperforming direct inference using ChatGPT-4o and ChatGPT-o1. All code, optimized APIs, and experimental videos will be publicly distributed via the webpage ( https://wangxiaoshawn.github.io/AutoMisty.html) .
Takeaways, Limitations
•
Takeaways:
◦
Improving Social Robot Accessibility through Natural Language-Based Robot Control
◦
Demonstrating the feasibility of performing complex robotic tasks through a multi-agent collaboration framework.
◦
Improving the accuracy and efficiency of LLM-based code generation
◦
Presenting the possibility of iterative work improvement based on transparent reasoning process and user feedback.
◦
Expanding research and increasing applicability through open source disclosure
•
Limitations:
◦
Currently, the system is specialized for the Misty robot, and its scalability to other robot platforms needs to be verified.
◦
There is a possibility of unpredictable results due to the limitations of LLM.
◦
Safety and reliability verification is needed for robot operation in real environments.
◦
There are limitations in processing power for complex tasks and the possibility of performance degradation.
◦
The possibility that biases in large-scale language models could influence robot behavior
