I've developed an AI agent for the TurtleBot3 mobile robot using ROS2. This agent can interpret commands such as “move to (X, Y)” or “check surroundings,” and dynamically selects from a toolkit of actions, including movement, sensor data retrieval (from odometry, LiDAR, or camera), and basic navigation decisions.
So far, I've tested the agent in the Gazebo simulator. If you're interested, please check out the demo video in my repository linked below. My next step is to deploy the agent on a real TurtleBot3 to evaluate how it handles the complexities of real-world environments.https://github.com/Yutarop/turtlebot3_agent
I'd love to hear feedback or ideas from the community, especially regarding the challenges of transferring a robot from simulation to real-world deployment.
Also, I'm wondering whether MCP might be a better fit for robotics applications. Any insights or experiences would be greatly appreciated!