This is the original source code of the ASE 2019 paper (distinguished paper award) for game testing, which combines multi-objective evolutionary algorithm (MOEA) and deep reinforcement learning (DRL) to explore game state and discover bugs. Detailed information can be found in:
@InProceedings{,
author = {Yan Zheng and Xiaofei Xie and Ting Su and Lei Ma and Jianye Hao and Zhaopeng Meng and Yang Liu and Ruimin Shen and Yinfeng Chen and Changjie Fan},
title = {Wuji: Automatic Online Combat Game Testing Using Evolutionary Deep Reinforcement Learning},
booktitle = {IEEE/ACM International Conference on Automated Software Engineering},
year = {2019},
}
This code is tested in Ubuntu 18.04.
First the dependent Python libraries should be installed by typing pip3 install -r requirements.txt
in a terminal.
Then start a clean (delete all model data via the -d
argument if you want) experiment by typing python3 ea.py -d
.
After a while, the number of founded bugs can be seen via TensorBoard (tensorboard --logdir ~/model/wuji
).