Skip to main content

TurtleBot Maze

Repo: aegean-ai/turtlebot-maze Environment: ros.dev.gpu Status: Active development

Overview​

The canonical user-plane reference application. ROS 2 navigation + SLAM in a simulated maze world, demonstrating Claude Code real-time robot control via MCP.

Key capabilities​

  • Claude Code /navigate skill: Real-time robot control via ros-mcp-server (MCP over rosbridge :9090)
  • Behavior trees: Autonomous navigation + search sequences (py_trees / BehaviorTree.CPP)
  • YOLOv8: Object detection via PyTorch, decoupled from ROS via Zenoh transport
  • stella_vslam: Visual SLAM for mapping and localization
  • Nav2: Path planning and collision avoidance
  • Cosmos stack (v1.5): Predict2 + Transfer2.5 + Reason2 loop for world model + sim2real + physical reasoning

Platform services​

  • User plane: All ROS 2 nodes run as Ray workers on ros.dev.gpu
  • Data plane: ROS bag recordings stored in lakehouse; SDG pipeline outputs to warehouse/
  • Control plane: ClusterAgent monitors ros.dev.gpu health