ML-ready simulator
WHAT
We want a 'headless' or Reinforcement-Learning-ready version of the simulator. API where we can call a step() function from outside - ideally without ROS, but ROS is fine too - where we would give command velocities for the robot and get lidar, image, position data in return.
HOW
Based on this paper https://arxiv.org/pdf/1809.02627.pdf
Branch: https://gitlab.inria.fr/CrowdBot/CrowdBotUnity/-/tree/headless-RL-version
Edited by Fabien Grzeskowiak