AI/ML platform for the Stretch robot — 234 Python files across perception, planning, mapping, and motion modules.
Vision models, object detection, segmentation, and image-text encoders for understanding the robot's environment
Language model clients and prompt builders for task planning and natural language interaction
Robot behavior orchestration — LLM-driven task decomposition, operation execution, and state management
3D scene understanding — voxel maps, instance tracking, and scene graphs for spatial reasoning
Motion planning, inverse kinematics, and trajectory control for robot movement
Speech recognition and synthesis for voice interaction with the robot
Abstract robot interfaces, communication, and configuration — the integration layer to stretch_body_ii
End-user demonstration applications combining perception, planning, and execution