search

ac-93 / tactile_gym

Suite of PyBullet reinforcement learning environments targeted towards using tactile data as the main form of observation.
GNU General Public License v3.0
127 stars 20 forks source link

tactile_gym + DI-engine #8

Open PaParaZz1 opened 2 years ago

PaParaZz1 commented 2 years ago

It's really a very nice project, tactile robotics is a novel topic.

I am the developer of DI-engine, looking for some interesting environments to apply advanced DRL algorithms and collect meaningful demonstrations. Are you willing to develop more examples and benchmark results together, or any other ideas?

ac-93 commented 2 years ago

Hi PaparaZz1, thanks for the kind words!

Did you have anything in particular in mind or was this more of an open ended collaboration?

Feel free to contact me via email if you would prefer: alex.church@bristol.ac.uk

PaParaZz1 commented 2 years ago

To begin with, we can add a training and evaluation demo for this env suite, like something in examples or ding_helpers which is similar to sb3_helpers.

In addition, we can discuss how to design more reasonable representation learning for multi-modal sensing/data like RGBT

ac-93 commented 2 years ago

Could you point me to a PPO example with DI-engine on some simple env such as cartpole? I'll try and find some time to setup something like dieng_helpers for learning on these environments.

PaParaZz1 commented 2 years ago

You can start from this cartpole demo. If you meet any problems, you can raise a issue or discuss in our channel.