reinforcement learning algorithm

[yani-rl-ai]

Hi @lubiluk i found this interesting project from your repo. May i know what kind of continuous control reinforcement learning algorithm that work with this environment?

[lubiluk]

Hi, this project is actually still work in progress, I intend to get back to it in the near future. I can't guarantee that it works at the moment. When it comes to RL algorithms to use with it, in theory you can use any policy gradient method like DDPG since the action space is continuous. But in practice I haven't got any good results yet.

[yani-rl-ai]

hi thanks for reply, good luck for your project

[yani-rl-ai]

hey @lubiluk how was this project? did you get the optimal policy from this environment?

[lubiluk]

It doesn't go well so far, I abandoned the project in frustration :D I have this problem that after some training time there is some kind of communication problem with ROS or Gazebo and the robot stops moving (an error says that there was a problem sending a message). I've run out of ideas for now on what could be the cause. Contributions are welcome ;)

[yani-rl-ai]

hi this repo using HSR for reinforcement learning and the results has been published. https://github.com/robot-learning-freiburg/kinematic-feasibility-rl

[lubiluk]

Thanks, I'll have a look on their implementation