XWorld: a simulator package for Reinforcement Learning
This repository currently contains a collection of simulators for Reinforcement Learning research.
| Name | Description | Thread-compatible?* | Optional? | Build-on-the-fly? |
|---|---|---|---|---|
| SimpleGame | A simple 1D array-walking game. | Yes | No | Yes |
| SimpleRace | A simple synthetic car racing game. | Yes | No | Yes |
| XWorld2D | A 2D environment for an agent to learn vision and language abilities. | Yes | No | Yes |
| Atari | Wrappers for the Arcade Learning Environment (ALE) environment. For stability, we use a fork version. | Yes | No | Yes |
| Malmo | Wrappers for Microsoft Malmo Project. | TBD | Yes | No,has to be installed manually first. |
| DeepMind Lab | Wrappers for Deepmind Lab. We use a fork version of the original repository to address a compiling issue. | Yes | Yes | Yes |
*If each thread has a different environment instance, then they don't interfere with each other. But if all threads share the same instance, then there is interference.
Dependencies
The following softwares must be installed before building XWorld.
In Ubuntu 14.04 and 16.04, you can do
sudo apt-get install libboost-all-dev libgflags-dev libgoogle-glog-dev libgtest-dev
Build
First get this git repository
git clone https://github.com/PaddlePaddle/XWorld
Suppose the directory is xworld_path, then do
cd <xworld_path>
mkdir -p build
cd build
cmake [<optional parameters>] ..
For example,
cd ~/XWorld; mkdir build; cd build
cmake -DWITH_DEEPMIND_LAB=ON ..
Finally, in the build directory do
make
make test
Use
You can compile a C++ project with XWorld. The C++ simulator APIs are located in simulator.h. Alternatively, we provide a set of simple Python APIs for interacting with the simulators. After building XWorld, you need to export the path of the python module:
export PYTHONPATH=<xworld_path>/python:$PYTHONPATH
You can add the above line to ~/.bashrc to avoid doing the export in the future.
To get started, several examples of the simulator Python APIs can be found in
<xworld_path>/python/examples
Citations
If you use the XWorld2D environment for your research, please consider citing
- Haonan Yu, Haichao Zhang, Wei Xu, A Deep Compositional Framework for Human-like Language Acquisition in Virtual Environment, arXiv 1703.09831, 2017.
- Haichao Zhang, Haonan Yu, Wei Xu, Listen, Interact and Talk: Learning to Speak via Interaction, arXiv 1705.09906, 2017.
If you use our wrappers for the third-party simulators, please follow their original guide for citation.
License
This repository has the Apache2.0 license, except that all the third-party simulators (i.e., ALE, Malmo, DeepMind Lab) have their own licenses.