AI based serious game design - Kleptomancy

Abstract

In this paper, we explored the use of Artificial Intelligence to create an adversary that demonstrates reasonable intelligence through the extensive use of Machine Learning techniques, Deep Reinforcement Learning and Imitation Learning techniques. In particular, we used Proximal Policy Optimization (PPO) algorithm, a branch of Model-Free RL Policy Optimization model, as well as Generative Adversarial Imitation Learning (GAIL) to train our intelligent agent. This project aims to evaluate and demonstrate the Intelligent Agent’s adaptive responses and strategies when faced with player-generated challenges in an edutainment game that was developed as part of this project, ‘Kleptomancy’.