Welcome to Care Robot’s Documentation

Contents

This is the documentation for the care robot project.

Contents:

• Architecture
  • Overview
  • Three operating modes
  • System overview
  • Cognition: the closed-loop brain
  • Embodiment: skills, vision, hardware
  • Walkthrough: one command, end to end
  • Where to go next
• Cognition: planning & verification
  • The closed loop
  • WorldState — the robot’s belief
  • Mapping symbolic actions to skills
  • Layered verification
  • SAGE planner via the Planner Protocol
  • The LLM backend
  • The event stream
• Embodiment: skills, vision & hardware
  • The skill contract
  • Skill groups
  • Perception & grasp pipeline
  • VisionServe — GPU inference server
  • Robot hardware (ROS2)
• Get started
  • Repositories
  • Prerequisites
  • Install the Python packages
  • Run the stack
  • Your first command
  • Deployment scenarios
  • Appendix: host setup
• Operator dashboard
  • Choosing a robot and a site
  • Sending a task
  • Watching the run
  • Safety
  • Build & deploy
• Manual
  • How to use
  • Init Arm
  • Arm Control
  • Gripper / Litf/ Head / Mobile Control
  • Recognition Skills
  • Approach Skills
  • Pick Skills
  • Place Skills
  • Wipe Skills
  • Open/Close Drawer Skills
• Configuration
  • Location (site) profiles
  • Device connections
  • The name map
  • World state
  • LLM provider & keys
• ROS2-based Framework
  • Challenges
  • Key Advantages of Custom ROS2 Node
  • Overview of the Custom ROS2 Node Implementation
  • Detailed Breakdown of Components
  • Advantages in Flexibility and Handling Complex Workflows
  • Quick Start
  • Use the Agents (Super Simple API)
  • Run the Node
  • Useful Extras
  • Full Minimal Example (Publisher + Subscriber)
• Examples
  • Topic Pub/Sub
  • Image Sub
  • Service Client/Server
  • Get plan from user’s prompt
  • Detect an object using structured command
  • Detect grasp pose of an object using structured command
  • Detect place pose of an object using structured commnad
• Troubleshooting & FAQ
  • The backend won’t boot
  • No devices / ROS not connected
  • VisionServe / detection fails
  • Planning / LLM problems
  • Grasping issues
  • Replanning never finishes
  • Voice / TTS
  • Two operators conflict
• API & code map
  • HTTP / WebSocket endpoints
  • Where the code lives
  • Enabling autodoc