Robot-agent frameworks for industrial inspection
My current work develops a model-agnostic framework that connects natural-language instructions with robot perception, navigation, and inspection skills. The architecture uses the Robot Operating System (ROS 2) and Model Context Protocol (MCP) to expose capabilities to AI agents, with Boston Dynamics Spot as the primary platform. The design separates AI task reasoning from deterministic robot control and safety constraints, with the aim of making inspection workflows reusable across models and platforms. The paper is under review at the Journal of Field Robotics.