Add instructions on how to run the code

hsp-iit / pybullet-robot-envs

A Python package that collects robotic environments based on the PyBullet simulator, suitable to develop and test Reinforcement Learning algorithms on simulated grasping and manipulation applications.

GNU Lesser General Public License v2.1

187 stars 24 forks source link

Add instructions on how to run the code #2

Closed giuliavezzani closed 5 years ago

giuliavezzani commented 5 years ago

Please @eleramp, could you update the README.md with detailed instructions on how to install the dependencies and how to run some examples?

I also suggest to include here just simple trainings and evaluations of the policies with one/two standard baselines (like the openai baselines) to show the correctness of the implemented environments. Then, the real implementation of new algorithms will be done in another repository.

giuliavezzani commented 5 years ago

@eleramp I tested your installation instructions. They are very well written, even if I encountered some problems.

Here are some suggestions to improve the instructions:

In the requirements.txt the requiredmatplotlib version is 3.1.0 However, this version requires Python >=3.6, while you say here that pybullet-robot-envs requires Python>=3.5.
At the end of the installation instructions, it'd be nice if you link or list the id_envs we designed, so that the user can immediately test them.

I'm not able to run:

$ python pybullet_robot_envs/examples/algos_baselines/train/icub_envs/train_deep_reaching.py

I got the error:

TypeError: learn() got an unexpected keyword argument 'network'

I think it's a matter of baselines version. Following your instructions I got version 0.1.5.

eleramp commented 5 years ago

@giuliavezzani thanks for your feedback! I updated the instruction trying to fix the issues you encountered. Please check if they are correct now and if you can finally run the script for training.

giuliavezzani commented 5 years ago

The new installation instructions go smoothly! Good job @eleramp :rocket: Let's leave this issue open, because @H2SO4T will need to make his code compatible with @eleramp's installation process and/or add other instructions to play with the Panda.

giuliavezzani commented 5 years ago

@eleramp I think the structure of this repo can be considered stable. Here are some suggestions:

Please remember to add in the README.md the description of the newly implemented environments (I was you added also the icub_push_env)
Don't add other training example in this folder, the example with one openai baseline algorithm is enough.
In the section RL Examples please add a link and a short description to you robot-agents project (which we need to move to robotology-playground too).

@H2SO4T, asap yu should align panda branch to master.

giuliavezzani commented 5 years ago

@H2SO4T please, once everything will be fixed, please add the instructions to launch your code in the README.md following @eleramp's template (in every section named panda).

giuliavezzani commented 5 years ago

Well done guys @eleramp @H2SO4T