BaziGooshi: A Hybrid Model of Reinforcement Learning for Generalization in Gameplay
Sara Karimi, Sahar Asadi, Amir H. Payberah · IEEE Transactions on Games · 2024
While Reinforcement Learning (RL) is gaining popularity in gameplay, creating a generalized RL model is still challenging. This study presentsBaziGooshi, a generalized RL solution for games, focusing on two different types of games: (i) a puzzle game Candy Crush Friends Saga and (ii) a platform game Sonic the Hedgehog Genesis.BaziGooshirewards RL agents for mastering a set of intrinsic basic skills as well as achieving the game objectives. The solution includes a hybrid model that takes advantage of a combination of several agents pre-trained using intrinsic or extrinsic rewards to determine the actions. We propose an RL-based method for assigning weights to the pre-trained agents. Through experiments, we show that the RL-based approach improves generalization to unseen levels, andBaziGooshisurpasses the performance of most of the defined baselines in both games. Also, we perform additional experiments to investigate further the impacts of using intrinsic rewards and the effects of using different combinations in the proposed hybrid models.