Agipix (Autonomy) — Quick Guide

Use this page to jump into the autonomy stack. It stays brief and links to the full guides.

Perception

• ROS 2 Humble packages for sensing and environment understanding.
• Typical inputs: LiDAR and/or RGB‑D; relies on odometry/pose from SLAM/VIO.
• Outputs feed mapping and planning (detections, depth/point clouds, occupancy/map layers).
• Runs on Linux x86_64 and ARM (Jetson) per the module badges.

Planning & Control

• Modular ROS 2 stack: mapping, global/trajectory planning, tracking controller, and mission wrappers.
• Supports tasks like autonomous navigation, exploration, and inspection.
• Built with colcon; typically used with Docker + Isaac ROS runtime.
• Demos mirror between simulation and PX4/real flight for behavior parity.

Data‑Driven Autonomy

• End‑to‑end policy that navigates via sub‑goals using learned visual representations.
• Leverages pretrained features; reports up to ~50% higher navigation success.
• Generalizes across ground and aerial robots.
• See the page for the illustrative figure and demo video.

Next steps

• If you haven’t set up the runtime yet, see real robot Software Setup.
• To run in simulator first: go to simulation Quick Guide.
• Coming from hardware bring-up? Start at real robot Quick Guide.