2022/01/01 by Dominik Kowalczyk, Dominik A. Kowalczyk, Jakub Kowalski +4
Computer Science · Economics, Econometrics and Finance · Mathematics · Social Sciences · #Adversary #Algorithm #Artificial Intelligence in Games #Artificial intelligence #Beam search #Best-first search #Computer science #Computer security #Digital Games and Media #Encoding (memory) #Evaluation function #Function (biology) #Game theory #Game tree #Information retrieval #Iterative deepening depth-first search #Machine learning #Mathematics #Monte Carlo method #Monte Carlo tree search #Position (finance) #Pruning #Reinforcement learning #Search algorithm #Search engine #Search tree #Sequential game #Sports Analytics and Performance #State (computer science) #Tree (set theory) #cs.AI #cs.NE
paper · pdf · doi:10.5220/0010840200003116
International Conference on Agents and Artificial Intelligence 2022
openalex publication_date 2022/01/01 · arxiv created 2022/03/17 · arxiv updated 2022/03/21 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/05
In this paper, we study AI approaches to successfully play a 2-4 players, full information, Bomberman variant published on the CodinGame platform. We compare the behavior of three search algorithms: Monte Carlo Tree Search, Rolling Horizon Evolution, and Beam Search. We present various enhancements leading to improve the agents' strength that concern search, opponent prediction, game state evaluation, and game engine encoding. Our top agent variant is based on a Beam Search with low-level bit-based state representation and evaluation function heavy relying on pruning unpromising states based on simulation-based estimation of survival. It reached the top one position among the 2,300 AI agents submitted on the CodinGame arena.