Model Context Protocol를 활용한 효율적인 로봇 제어 프레임워크

초록

This study proposes an efficient and generalizable robot control framework that leverages the structural advantages of the Model Context Protocol (MCP) to transform natural language commands, issued by a user through a Large Language Model (LLM), into executable robot actions. The proposed framework structures high-level user intent into standardized MCP messages and delivers them to the robot in a form clearly separated from low-level control mechanisms, enabling clear, consistent, and interpretable command execution. Additionally, the MCP server interprets incoming messages in real time and converts them into ROS instructions, allowing the robot to perform responsive, adaptive, and context-aware behaviors in dynamic environments. To verify the applicability and effectiveness of the proposed framework, we implemented and tested MCP-based control using the MOCA mobile manipulator within the NVIDIA Isaac Sim simulation environment. The results demonstrate the system's capability to handle natural language inputs and translate them into meaningful robot actions. This approach is easily extendable to a wide range of robot platforms and task scenarios, contributing to the development of scalable, natural language-driven robot control systems.

키워드