ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question
2

Could i use my own Gazebo world as RL Environment for a turtlebot3? how?

asked 2022-05-31 04:13:41 -0500

Belghiti gravatar image

updated 2022-05-31 16:05:32 -0500

Hi everyone! I've just created my own Maze with Gazebo Simulator, and im wandering if i could use it as a Simulation Environment for my RL-project while using OpenAI and ROS. I think that OpenAI Pakage can provide different Environment, but is it possible to use my own ? if so, then i'd be so grateful to know how.

edit retag flag offensive close merge delete

Comments

1

Can you add links to these packages? I think they sound interesting. Thanks!

ljaniec gravatar image ljaniec  ( 2022-05-31 04:27:20 -0500 )edit

@ljaniec, yes with pleasure! https://github.com/edowson/openai_ros. and i hope im not wrong...

Belghiti gravatar image Belghiti  ( 2022-05-31 16:05:59 -0500 )edit

Hi, how did you manage to speed up the simulation ? I triedadding the following lines to my .world but it didn't changes anything

<max_step_size>0.001</max_step_size> <real_time_update_rate>1000</real_time_update_rate>

Hedwin gravatar image Hedwin  ( 2023-06-04 12:12:41 -0500 )edit

Can you speed up the simulation in the Gazebo window? Or is it not enough?

ljaniec gravatar image ljaniec  ( 2023-06-05 04:15:14 -0500 )edit

Thanks ! Indeed I was able to modify those two parameters in my simulation and it changes the simulation speed. However, the problem came from the values I choose. According to https://classic.gazebosim.org/tutoria... : "real_time_update_rate: This is the frequency at which the simulation time steps are advanced. The default value in Gazebo is 1000 Hz. Multiplying with the default max_step_size of 0.001 seconds gives a real_time_factor of 1"

But setting max_step_size to 0.01 instead of 0.001 broke the robot so I guess I have to find the right parameters now !

Hedwin gravatar image Hedwin  ( 2023-06-06 08:53:48 -0500 )edit

1 Answer

Sort by ยป oldest newest most voted
1

answered 2022-06-01 17:58:03 -0500

ljaniec gravatar image

I think this tutorial "Using Gazebo for Reinforcement Learning" can be helpful:

and this:

edit flag offensive delete link more

Comments

1

@ljaniec, I am indeed deeply grateful for this great help. the Hero we all need but don't deserve :D

Belghiti gravatar image Belghiti  ( 2022-06-02 02:03:00 -0500 )edit

Question Tools

2 followers

Stats

Asked: 2022-05-31 04:13:41 -0500

Seen: 461 times

Last updated: Jun 01 '22