Ms Pac-Man Competition
(screen capture mode)
IEEE CIG 2010 Results
Simon M. Lucas

Entries

There were seven functioning entries as listed below.

Congratulations go to Martin Emilio, Martinez Moises, Recio Gustavo, Saez Yago from Universidad Carlos III in Madrid with their entry Pac-mAnt - an Ant-colony-based agent which achieved a high score on ten runs of 21,250.  Clearly Spain were not content with just winning the football world cup in 2010!

However, none of the entries on the day were able to beat the ICE Pambush 3's CIG 2009 high score of 30,010.

There is definitely room for improvement in time for the planned CIG 2011 competition (to be confirmed).

Entry / Author(s) (click link for authors) Affiliation High Score
Emilio - Pac-mAnt Universidad Carlos III de Madrid 21,250
ICE
ICE Pambush 4 (also see IEEE CIG 2010 Paper)
Ritsumeikan University, Kyoto, Japan 20,580
Jave University of Nottingham, UK 14,660
Bruce City University of Hong Kong 10,820
Sojoodi

Shiraz University, Tehran

9990
CoboPac University of Würzburg, Germany 5790
Kim

Sejong University, Seoul

18690*
* I was unable to run this entry on my machine; the authors provided an updated version after the live competition which did run and gave these results; Jave still finishes in third position


Results

The full results are shown below (see note above about the Kim entry).

IEEE CIG 2010 Ms Pac-Man Results Table

The winning entry is the one that achieved the highest score given ten runs each.  The averages are also shown for interest, and this time the winning entry had only the third-highest average.

Some obvious weaknesses of all the entries:

  • they all have a fairly greedy short term strategy, and in many cases single food pills are left behind in the pursuit of more direct rewards, which makes life difficult towards the end of the level when these must be consumed in the absence of the relative security provided by the energiser pills.

  • they are not fully competent: they sometimes make unforced errors, and die by failing to make an obvious turn for example.  In some cases this appears to be due to screen parsing and/or control software failing to send the correct key response in sufficient time

There are now some diverse and interesting ideas behind these entries, which need some refinement and to be coupled to an improved screen-capture and game control system.  We look forward to improved entries for IEEE CIG 2011, and there is a good chance that we'll see some of these breaking the 50k barrier. 

Videos

May be coming soon...

 

Follow-Up

There are two planned competitions for 2011 (see links below for confirmation).

Links

When you have a technique that works well, you may want to write it up as a paper for the IEEE Transactions on Computational Intelligence and AI in Games.

References

We encourage entrants to try evolutionary and machine learning methods.  The papers below might provide a useful starting point.

Also see the Pac Man papers in IEEE CIG 2008 and in IEEE CIG 2009 and 2010.

[1] Simon M. Lucas, Evolving a Neural Network Location Evaluator to Play Ms. Pac-Man, IEEE Symposium on Computational Intelligence and Games (2005), pages: 203 -- 210 [pdf]

[2] Szita and A. Lorincz, Learning to Play Using Low-Complexity Rule-Based Policies: Illustrations through Ms. Pac-Man, Journal of Artificial Intelligence Research (2007), Volume 30, pages 659-684 [pdf].

 

end of page