pybullet-workbook

PyBullet installation

1. Create a virtual environment:

virtualenv pybullet

2. Activate the virtual environment:

source pybullet/bin/activate

3. Install all the dependencies necessary for doing RL training with py bullet:

pip install -r requirements.txt

4. Test installation running:

python helloworld.py

5. Test RL framework running:

python -m pybullet_envs.examples.enjoy_TF_HumanoidBulletEnv_v0_2017may

or

python -m pybullet_envs.examples.kukaGymEnvTest

6. Get OpenAI baselines:

git clone https://github.com/openai/baselines.git

7. Test everything working with:

cd baselines/
python -m pybullet_envs.agents.train_ppo --config=pybullet_pendulum --logdir=pendulum

Pybullet summary

Pybullet tutorial is available here. Main features:

• It is possible to load urdf/sdf models;
• Interactive OpenGl gui;
• It is possible to control robots (inverse dynamics and kinematics);
• Options in camera rendering;
• RL framework:
  • a suit for gym environments is installed together with PyBullet;
  • It includes pybullet version of openai gym envs;
  • To train we can use
    • openai baselines;
    • agents: a python package for tensorflow implementation of RL algorithms;
• Also VR is available.

Examples

• helloworld.py: a basic script for loading one of pybullet models
• helloworld_icub.py: a basic script loading an iCub sdf model.

This model has the base fixed to the ground, suitable for manipulation tasks. However, no accurate models for the iCub hand are available so far in a format suitable for pybullet.

Envs

Gym-like environments for the iCub and the Franka Panda are under development in /envs.

The iCub model has been obtained from the urdf including hands manually fixed the joint reference frames.

Tha Panda model has been obtained by converting the .urdf.xarco model provided by franka_ros using Ros.

The structureof the environments takes inspiration from pybullet kuka env.