| Shihong Song# | Jiayi Weng# | Hang Su | Dong Yan | Haosheng Zou | Jun Zhu |
| Tsinghua University #The authors contributed equally to this work. |
Learning rational behaviors in First-person-shooter (FPS) games is a challenging task for Reinforcement Learning (RL) with the primary difficulties of huge action space and insufficient exploration. To address this, we propose a hierarchical agent based on combined options with intrinsic rewards to drive exploration. Specifically, we present a hierarchical model that works in a manager-worker fashion over two levels of hierarchy. The high-level manager learns a policy over options, and the low-level workers, motivated by intrinsic reward, learn to execute the options. Performance is further improved with environmental signals appropriately harnessed. Extensive experiments demonstrate that our trained bot significantly outperforms the alternative RL-based models on FPS games requiring maze solving and combat skills, etc. Notably, we achieved first place in VDAIC 2018 Track(1).
You can download full paper through IJCAI site.
@inproceedings{ijcai/SongWSYZZ19,
author = {Shihong Song and
Jiayi Weng and
Hang Su and
Dong Yan and
Haosheng Zou and
Jun Zhu},
title = {Playing {FPS} Games With Environment-Aware Hierarchical Reinforcement
Learning},
booktitle = {Proceedings of the Twenty-Eighth International Joint Conference on
Artificial Intelligence, {IJCAI} 2019, Macao, China, August 10-16,
2019},
pages = {3475--3482},
year = {2019},
url = {https://doi.org/10.24963/ijcai.2019/482},
doi = {10.24963/ijcai.2019/482},
timestamp = {Thu, 19 Sep 2019 14:00:20 +0200},
biburl = {https://dblp.org/rec/bib/conf/ijcai/SongWSYZZ19},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
