How to make the VecEnv of Gazebo


I found that there are a lot of openai environment support VecEnv, which generates many environment instances at the same time. Using this way, the learning algorithm can interactive with environments at the same time and collect more data, which can speed up training.

Is there some method that allows us to make the Gazebo works like that?



Well I havent tried to use VecEnv with OpenAI-ROS, but in theory you can launch multiple instances of gazebo in multiple computers with ROSDevelopementStudio with GymComputers. What I havent tried is the moethod of unifiing everything. If you do please share the ROSject here in teh forum.


I still didn’t make that work. But I found that the project “gym-gazebo” in github seems meets the need.