noisebridge-nav

Creating an E2E AI for navigating an indoor environment.

The goal is to create a robot that can navigate an indoor space. Although the original idea is something that can fetch an object or look for things in a space, we will focus on image or location-based navigation for the time being. This (rather large) goal can be broken down into several smaller goals:

• Train a PPO to do the IndoorNav project
  • Visual Cortex 1 can help train a better PPO model by creating a visual representation
• Optionally, create a bare-bones navigation and obstacle avoidance system. I have Roomba RL, but a more successful version can be found here
• Develop a map of Noisebridge compatible with the IndoorNav project
  • Needs to convert a 5 DoF position/viewing to an image
  • Can be done through NeRFs
  • We (i.e. Noisebridge) also have a GitLab with a 3D model of Noisebridge , no idea how to convert this to the format above

How to run VC-1

conda activate habitat
python example.py

Currently set up to work on renderbox

Install for RPi code

Follow the instructions for CppGPIO

This also depends on ROS 2 Humble and rclcpp (which should come with ROS2).

TODO

• Create an environment file for Noisebridge environment