86. Reproducing Robotic Simulations for Reinforcement Learning with Gym-Ignition

Gym-ignition-reinforcement-learning-Diego-Ferigo

22/12/2020

I would like to dedicate this episode to all the ROS Developers that are doing research on reinforcement learning applied to robotics. That is a difficult subject for which there is not much about it, specially with ROS. There are some works using other frameworks and simulators but not many that use ROS as their framework. So if you are in that field ROS + Reinforcement Learning + Robots, then this episode is dedicated to you.

Today we are going to learn how we can speed our development of such algorithms for ROS robots by using Gym-Ignition.

But before going into that, let me remind you about our ROS online academy. Yes, at the Construct we have created an online academy named The Robot Ignite Academy which contains a ROS learning path for beginners. The path is composed by:

Linux for Robotics
Python for Robotics
ROS Basic Concepts
Understanding TF
URDF for modeling robots
ROS Control
Build your own robot and ROSify it… and become a ROS Developer!

Some other robotics theory courses included are:

Of course all our courses are based on practice with simulated robots! (theory+practice)

Now let me introduce you Diego Ferigo. Diego is a PhD fellow at Italian Institute of Technology, where he is doing research about using reinforcement learning for humanoid robot locomotion. Today he is here to talk about reinforcement learning with ROS, and about the framework he has developed for that, the Gym-Ignition.

Welcome to the podcast Diego!