This repository contains simulation environments for the TIAGo robot based on pybullet. They follow OpenAI's gym environment structure in order to by used in combination with reinforcement learning algorithms that follow the same convention (e.g. baselines).
- Clone repository and change into the directory:
git clone https://github.com/llach/tiago_rl && cd tiago_rl
- Install dependencies:
pip install -r requirements.txt
- Install the package itself:
pip install -e .
Using the pip flag -e leads to an editable installation of the project. This means, that changes to the source code are available directly without re-installing the project.
Currently, this repository contains three different environments. All of them are based on OpenAI's robotics environments and thus similar to them in many regards. As opposed to the standard robotics environments, this package offers tactile sensors also for mobile manipulators.
TIAGo with PAL Gripper using standard (non-sensorized) fingers.
TIAGo with PAL Gripper using TA11 load cell sensors as fingers.
Only PAL Gripper using TA11 load cell sensors as fingers.
Uses not as many complex collision bodies and thus may speed up policy learning for gripper-only policies.
