Deep Reinforcement Learning & Rogue

Andrea Asperti, Daniele Cortesi, Carlo De Pieri, Gianmaria Pedrini, Francesco Sovrano



Watch us on youtube.

Rogue is a famous dungeon-crawling video-game of the 80ies, the ancestor of its gender. Rogue-like games are known for the necessity to explore partially observable and always different randomly-generated labyrinths, preventing any form of level replay. As such, they serve as a very natural and challenging task for reinforcement learning, requiring the acquisition of complex, non-reactive behaviors involving memory and planning. We both developed an applicative interface called Rogueinabox allowing a simple interaction with the game, and several different agents based on deep reinforcement techniques. At the moment, our best agent exploits a version of A3C partitioned on different situations: this version of the agent is able to reach the stairs and descend to the next level in 98% of cases.

Code

The code is available on GitHub at this url.

References

Cite

@article{crawling,
  author    = {Andrea Asperti and
               Daniele Cortesi and
               Francesco Sovrano},
  title     = {Crawling in Rogue's dungeons with (partitioned) {A3C}},
  journal   = {CoRR},
  volume    = {abs/1804.08685},
  year      = {2018},
  url       = {http://arxiv.org/abs/1804.08685},
  archivePrefix = {arXiv},
  eprint    = {1804.08685},
  timestamp = {Wed, 02 May 2018 15:55:01 +0200},
  biburl    = {https://dblp.org/rec/bib/journals/corr/abs-1804-08685}
}
 
    
@article{RogueinaboxA,
  author    = {Andrea Asperti and 
               Carlo De Pieri and 
               Mattia Maldini and 
               Gianmaria Pedrini and
               Francesco Sovrano},
  title     = {A Modular Deep-learning Environment for Rogue},
  journal   = {WSEAS Transactions on Systems and Control},
  volume    = {12},
  year      = {2017},
  url       = {http://www.wseas.org/multimedia/journals/control/2017/a785903-070.php}
}
@article{RogueinaboxB,
  author    = {Andrea Asperti and 
               Carlo De Pieri and  
               Gianmaria Pedrini},
  title     = {Rogueinabox: an Environment for Roguelike Learning},
  journal   = {International Journal of Computers},
  volume    = {2},
  pages     = {146-154},
  year      = {2017},
  url       = {http://www.iaras.org/iaras/filedownloads/ijc/2017/006-0022(2017).pdf},
}