Ms. Pac-Man is a challenging, classic arcade game that provides an interesting platform for Artificial Intelligence (AI) research. This paper reports the first Monte-Carlo approach to develop a ghost avoidance module of an intelligent agent that plays the game. Our experimental results show that the look-ahead ability of Monte-Carlo simulation often prevents Ms. Pac-Man being trapped by ghosts and reduces the chance of losing Ms. Pac-Man's life significantly. Our intelligent agent has achieved a high score of around 21,000. It is sometimes capable of clearing the first three stages and playing at the level of a novice human player.
Ms. Pac-Man is a challenging video game which provides an interesting platform for artificial intelligence and computational intelligence research. This paper introduces the novel concept of path testing and reports an effective Monte-Carlo approach to develop an endgame module of an intelligent agent that plays the game. Our experimental results show that the proposed method often helps Ms. Pac-Man to eat pills effectively in the endgame. It enables the agent to advance to higher stages and earn more scores. Our agent with the endgame module has achieved a 20% increase in average score over the same agent without the module.
In this paper, the behavior and performance of Pure Random Orthogonal Search (PROS), a parameter-free evolutionary algorithm (EA) that outperforms many existing EAs on the well-known benchmark functions with finite-time budget, are analyzed. The sufficient conditions to converge to the global optimum are also determined. In addition, we propose two modifications to PROS, namely Triangular-Distributed Random Orthogonal Search (TROS) and Quadratic-Distributed Random Orthogonal Search (QROS). With our local search mechanism, both modified algorithms improve the convergence rates and the errors of the obtained solutions significantly on the benchmark functions while preserving the advantages of PROS: parameterless, excellent computational efficiency, ease of applying to all kinds of applications, and high performance with finite-time search budget. The experimental results show that both TROS and QROS are competitive in comparison to several classic metaheuristic optimization algorithms.