An evaluation of the benefits of look-ahead in Pac-Man

Thompson, Tommy; McMillan, Lauren; Levine, John; Andrew, Alastair

doi:10.1109/cig.2008.5035655

Cited by 10 publications

(5 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Rule-based & Finite State Machines [71], [16], [15], [72], [18], [23], [24], [9], [52], [10], [65] 67…”

Section: Ai / CImentioning

confidence: 99%

“…Thompson et al [72] analyse the impact of looking ahead in Pac-Man and compare their approach to simpler controllers based on greedy and random decisions. The game used by the authors is based on Pac-Man but has some important differences, most notably non-deterministic ghosts (other differences relate to the lack of bonus fruits and speed of the ghosts, amongst others).…”

Section: A Rule-basedmentioning

confidence: 99%

See 1 more Smart Citation

Pac-Man Conquers Academia: Two Decades of Research Using a Classic Arcade Game

Rohlfshagen¹,

Liu

Peŕez-Liebana

et al. 2018

IEEE Trans. Games

View full text Add to dashboard Cite

Pac-Man and its equally popular successor Ms Pac-Man are often attributed to being the frontrunners of the golden age of arcade video games. Their impact goes well beyond the commercial world of video games and both games have featured in numerous academic research projects over the last two decades. In fact, scientific interest is on the rise and many avenues of research have been pursued, including studies in robotics, biology, sociology and psychology. The most active field of research is computational intelligence, not least because of popular academic gaming competitions that feature Ms Pac-Man. This paper summarises peer-reviewed research that focuses on either game (or close variants thereof) with particular emphasis on the field of computational intelligence. The potential usefulness of games like Pac-Man for higher education is also discussed and the paper concludes with a discussion of prospects for future work.

show abstract

“…Rule-based & Finite State Machines [71], [16], [15], [72], [18], [23], [24], [9], [52], [10], [65] 67…”

Section: Ai / CImentioning

confidence: 99%

Section: A Rule-basedmentioning

confidence: 99%

Pac-Man Conquers Academia: Two Decades of Research Using a Classic Arcade Game

Rohlfshagen¹,

Liu

Peŕez-Liebana

et al. 2018

IEEE Trans. Games

View full text Add to dashboard Cite

show abstract

“…Note that look-ahead is typical in games, including Ms. Pac-Man (see e.g. [33]) and it is also considered in the MDP literature (see, e.g., [9] and the references therein). Here, we consider look-ahead from the point of view of a general RL architecture that inherits knowledge, learns, and faces new challenges.…”

Section: Factored Finite State Model Approximation For Planningmentioning

confidence: 99%

“…Finally, since branching ratio is low, such models are easy to run and can be used for look-ahead (LA). LA is commonly used in RL, including games (see, e.g., [13,1,33]).…”

Section: World May Change: Considerable Part (If Not All) Of the Know...mentioning

confidence: 99%

Decision Making Agent Searching for Markov Models in Near-Deterministic World

Matuz¹,

Lörincz²

2011

Preprint

View full text Add to dashboard Cite

Reinforcement learning has solid foundations, but becomes inefficient in partially observed (non-Markovian) environments. Thus, a learning agentborn with a representation and a policy -might wish to investigate to what extent the Markov property holds. We propose a learning architecture that utilizes combinatorial policy optimization to overcome non-Markovity and to develop efficient behaviors, which are easy to inherit, tests the Markov property of the behavioral states, and corrects against non-Markovity by running a deterministic factored Finite State Model, which can be learned. We illustrate the properties of architecture in the near deterministic Ms. Pac-Man game. We analyze the architecture from the point of view of evolutionary, individual, and social learning.

show abstract

“…Another work with a simplified simulator was conducted by De Bonet and Stauffer [6] , who used reinforcement learning to evolve an agent. Thompson, McMillan, Levin, and Andrew [7] compared a greedy-look-ahead agent to a greedyrandom agent, a random agent, and a human player. The human player scored the highest score, followed by the greedy-look-ahead agent.…”

Section: Previous Workmentioning

confidence: 99%

Evolving diverse Ms. Pac-Man playing agents using genetic programming

Alhejali

Lucas

2010

2010 UK Workshop on Computational Intelligence (UKCI)

View full text Add to dashboard Cite

This paper uses genetic programming (GP) to evolve a variety of reactive agents for a simulated version of the classic arcade game Ms. Pac-Man. A diverse set of behaviours were evolved using the same GP setup in three different versions of the game. The results show that GP is able to evolve controllers that are well-matched to the game used for evolution and, in some cases, also generalise well to previously unseen mazes. For comparison purposes, we also designed a controller manually using the same function set as GP. GP was able to significantly outperform this hand-designed controller. The best evolved controllers are competitive with the best reactive controllers reported for this problem.

show abstract

An evaluation of the benefits of look-ahead in Pac-Man

Cited by 10 publications

References 3 publications

Pac-Man Conquers Academia: Two Decades of Research Using a Classic Arcade Game

Pac-Man Conquers Academia: Two Decades of Research Using a Classic Arcade Game

Decision Making Agent Searching for Markov Models in Near-Deterministic World

Evolving diverse Ms. Pac-Man playing agents using genetic programming

Contact Info

Product

Resources

About