ROS Resources: Documentation | Support | Discussion Forum | Service Status | Q&A answers.ros.org

[TB3] Reinforcement Learning with TB3!


#1

Hello everyone! :slight_smile:

We introduce a teaser video about the Machine Learning with TurtleBot3.

We’ve started deploying Machine Learning onto TurtleBot3 to make progress in navigation using Deep Q Network (DQN).

We’ll keep posting how-to videos and source code later on.

Thanks!


#2

@Gilbert It’s cool and interesting. I’d like to check with you:

  • What does your current solution base on ? pure vision or Lidar invovled ?
  • I’m really curious about whether you have a try in the actual nature environment after tuning the model well within the simulation ?
  • Where to find your details or will you open it the others ?

Thanks for your sharing !


#3

Hi @Roser :slight_smile:

Thanks for your interesting!


#4

Good work! Liked the moving obstacles scenario, specially.

@Roser, if I’m not mistaken, the simulation has been (at least partially) inspired by https://github.com/erlerobot/gym-gazebo environments. Is that right @Gilbert?

A very cool project would be to extend this and try a few policy gradient methods. You could even go ahead an compare it with the value iteration one you just tried (DQN). I did a while ago a tutorial comparing different methods for a simple environment but yours is indeed much cooler.


#5

Hello, @vmayoral :slight_smile:

Actually, It was helpful to me to refer https://github.com/erlerobot/gym-gazebo environments.
But, I used just gazebo for the simulation.

I absolutely agree with your advice and will find better policy.
Thanks for your interesting!


#6

I just released https://github.com/TensorSwarm/TensorSwarm which allows you to over 100 robots at the same time. At the moment Proximal Policy Optimization is used as it seems to provide the best results.

The backend is a ROS service so you should be able to adopt your robots pretty easily. I’d be also glad to provide you with some support.