openai-rosbot-env

Trying to create customized gym environments based on ros/gazebo platform

Setup:

Ros Distribution - ROS Noetic Ninjemys
Operating System - Ubuntu 20.04 (Focal)
Python Distribution - Python 3.X
Gym - doc

Steps:

• create gazebo environment
  1. convert layout xxx.png to occupancy map xxx.pgm using layout_to_occpmap.py
     
  2. create corresponding xxx.yaml for occupancy map as sample
  3. read the map and publish using command === rosrun map_server map_server xxx.yaml ===
  4. run map2gazebo using command === roslaunch map2gazebo map2gazebo.launch export_dir:=/path/to/export_dir === to read published map and create stl/dae based on map layout

  Note: above occupany map is used only to generate dae/stl gazebo environment, for localization task we still use occupancy map generated from APIs like gmapping

References:

codebase

• openai-ros
• custom gym env creation guide
• HouseExpo indoor layout dataset
• Gazebo env from 2D maps

rosbots

• TurtleBot3 turtle-bot3 , turtle-bot3-msgs , turtle-bot3-simulations , turtle-bot3-driver

Notes:

• configure #export TURTLEBOT3_MODEL=waffle in .bashrc file
• assumption is all sensor data, poses, etc are in same frame ie. 'map'
• fix for map2gazebo utf encoding error is to change lines open(export_dir + "/map.stl", 'w') => open(export_dir + "/map.stl", 'wb')
• https://answers.gazebosim.org//question/16397/roslaunch-gazebo-2-process-has-died-pid-7605-exit-code-139/ => this issue occuring sometime ** need to fix **
• deleting turtlebot3 model from gazebo not corresponding services & topics sor currently turtlebot3 is never removed after gazebo is launched