Development of autonomous intelligent systems
I am working on a unified framework for natural-language control of robotic manipulators using the Model Context Protocol (MCP). The approach connects large language models with real-time perception, spatial reasoning, and robot control, allowing users to command robots through flexible natural-language instructions. High-level requests are interpreted by an LLM and mapped to modular tools that handle perception, world modelling, and manipulation, while platform-agnostic controllers support deployment on different robot arms and interaction via text, speech, and graphical interfaces.
The system was evaluated on a range of manipulation tasks with a robotic arm, showing reliable performance in scenarios involving object recognition, spatial reasoning, and multi-step actions, even under ambiguity and visual similarity.
All tools, controllers, and environment components are publicly available at github.com/dgaida/robot_mcp.
