Academic Staff

Dr Diego Perez

Emaildperez@essex.ac.uk
Room3A.527
Biography

Born in Madrid (Spain) in 1983 and living in London (United Kingdom). I am graduated in 2007 in Computer Science and I own a master degree in the same field (both in Carlos III University, Madrid). My research is centered in the application of Artificial Intelligence to games, Reinforcement Learning and Evolutionary Computation. I have participated in several AI Game competitions held in IEEE International Conferences. I have also organized the Physical Travelling Salesman Competition, and I am organizing the General Video Game AI Competition, to be held at IEEE WCCI and CIG International Conferences. I have experience in the videogames industry as a programmer and as a Chief Technical Officer at Game Brains (Dublin, Ireland), developing Artificial Intelligence tools that can be applied to the latest industry videogames. I am currently a PhD student in the University of Essex (United Kingdom), under the supervision of Simon M. Lucas, applying Monte Carlo Tree Search to real-time games. I am the lecturer of the 'Games Console Programming' module at the University of Essex.

Websitehttp://www.diego-perez.net
Current research

General Video Game Playing

Research interests
  • Computational Intelligence for Games,
  • Reinforcement Learning,
  • Evolutionary Computation.
Teaching responsibilities
 Lecturer of the module High Level Games Development and Game Artificial Intelligence.

Publications

2015

  • Spyridon Samothrakis, Diego Perez, Simon M. Lucas and Maria Fasli. Neuroevolution for General Video Game PlayingProceedings of the IEEE Conference on Computational intelligence and Games (CIG) (2015). [pdf]
  • Diego Perez, Jens Dieskau, Martin Hünermund, Sanaz Mostaghim and Simon M. Lucas. Open Loop Search for General Video Game PlayingProceedings of the Genetic and Evolutionary Computation Conference (GECCO) (2015). [pdf]
  • Diego Perez, Spyridon Samothrakis, Julian Togelius, Tom Schaul, Simon Lucas, Adrien Couetoux, Jerry Lee, Chong-U Lim, Tommy Thompson. The 2014 General Game Playing Competition,IEEE Transactions on Computational Intelligence and AI in Games, DOI: 10.1109/TCIAIG.2015.2402393 (2015). [pdf | ieeexplore]
  • Diego Perez, Adaptive Controllers for Real-Time GamesPhD Thesis, School of Computer Science and Electronic Engineering, University of Essex, UK (January 2015). [pdf]

2014

  • Diego Perez, Sanaz Mostaghim, Spyridon Samothrakis and Simon M. Lucas, Multi-Objective Monte Carlo Tree Search for Real-Time GamesIEEE Transactions on Computational Intelligence and AI in Games, DOI: 10.1109/TCIAIG.2014.2345842 (2014). [pdf | ieeexplore]
  • Spyridon Samothrakis, Diego Perez, Philipp Rohlfshagen and Simon M. Lucas, Predicting Dominance Rankings for Score-based GamesIEEE Transactions on Computational Intelligence and AI in Games, DOI: 10.1109/TCIAIG.2014.2346242 (2014). [pdf | ieeexplore]
  • Diego Perez, Spyridon Samothrakis and Simon M. Lucas, Knowledge-based Fast Evolutionary MCTS for General Video Game PlayingProceedings of the IEEE Conference on Computational Intelligence and Games (2014), pp. 68-75. [pdf]
  • Spyridon Samothrakis, Samuel Roberts, Diego Perez and Simon M. Lucas, Rolling Horizon methods for Games with Continuous States and ActionsProceedings of the IEEE Conference on Computational Intelligence and Games (2014), pp. 224-231 [pdf]
  • Diego Perez, Edward Powley, Daniel Whitehouse, Spyridon Samothrakis, Simon M. Lucas, Peter Cowling, The 2013 Multi-Objective Physical Travelling Salesman Problem Competition,Proceedings of the IEEE Congress on Evolutionary Computation, 2014, pages: to appear. [pdf]
  • Simon M. Lucas, Spyridon Samothrakis and Diego Perez. Fast Evolutionary Adaptation for Monte Carlo Tree SearchEvo Games 2014. [pdf]

2013

  • Diego Perez, Julian Togelius, Spyridon Samothrakis, Philipp Rolhfshagen and Simon M. Lucas. Automated Map Generation for the Physical Travelling Salesman ProblemIEEE Transactions on Evolutionary Computation (2013), DOI: 10.1109/TEVC.2013.2281508, pages: to appear. [pdf | ieeexplore]
  • Diego Perez, Edward Powley, Philipp Rohlfshagen, Daniel Whitehouse, Spyridon Samothrakis, Peter Cowling and Simon Lucas, Solving the Physical Travelling Salesman Problem: Tree Search and Macro ActionsIEEE Transactions on Computational Intelligence and AI in Games, 6:1, pp. 31-45, 2013. [pdf | ieeexplore]
  • Diego Perez, Spyridon Samothrakis and Simon M. Lucas. Online and Offline Learning in Multi-Objective Monte Carlo Tree SearchProceedings of the IEEE Conference on Computational Intelligence and Games (CIG) (2013), pp. 121-128. [pdf]
  • Diego Perez, Spyridon Samothrakis, Simon M. Lucas and Philipp Rolfshagen. Rolling Horizon Evolution versus Tree Search for Navigation in Single-Player Real-Time GamesProceedings of the Genetic and Evolutionary Computation Conference (GECCO) (2013), pages: 351-358. [pdf]

2012

  • Diego Perez, Philipp Rolfshagen and Simon M. Lucas. Monte Carlo Tree Search: Long-term versus Short-term PlanningIEEE Conference on Computational Intelligence and Games (2012), pp. 219 -- 226 (Best paper award). [pdf]
  • Diego Perez, Philipp Rohlfshagen and Simon Lucas. The Physical Travelling Salesman Problem: WCCI 2012 CompetitionProceedings of the IEEE Congress on Evolutionary Computation, pp. to appear, 2012. [pdf]
  • Diego Perez, Philipp Rohlfshagen and Simon Lucas. Monte-Carlo Tree Search for the Physical Travelling Salesman ProblemEvoApplications 2012, Vol. 7248, pp. 255--264, Springer, Heidelberg (2012). [pdf]
  • Simon M. Lucas, Philipp Rohlfshagen and Diego Perez. Towards More Intelligent Adaptive Video Game Agents: A Computational Intelligence PerspectiveProceedings of ACM International Conference on Computing Frontiers (2012), pages: 293-298. [pdf]
  • Cameron Browne, Edward Powley, Daniel Whitehouse, Simon Lucas, Peter Cowling, Philipp Rohlfshagen, Stephen Tavener, Diego Perez, Spyridon Samothrakis and Simon Colton. A Survey of Monte Carlo Tree Search MethodsIEEE Transactions on Computational Intelligence and AI in Games, Vol. 4:1, March, pp. 143, 2012. [pdf]

2011

  • Diego Perez, Miguel Nicolau, Michael O'Neill and Anthony Brabazon. Reactiveness and Navigation in Computer Games: Different Needs, Different ApproachesProceedings of the 2011 IEEE Conference on Computational Intelligence and Games, pp 273-280. [pdf]
  • Diego Perez, Miguel Nicolau, Michael O'Neill and Anthony Brabazon. Evolving Behaviour Trees for the Mario Bros Game Using Grammatical Evolution, LNCS 6624, Proceedings of EvoGAMES 2011 the 3rd European Event on Bio-inspired Algorithms in Games, Springer, pp.121-130. (Best paper award). [pdf]

2010

  • Daniele Loiacono, Pier Luca Lanzi, Julian Togelius, Enrique Onieva, David A. Pelta, Martin V. Butz, Thies D. Lnneker, Luigi Cardamone, Diego Perez, Yago Saez, Mike Preuss and Jan Quadieg.The 2009 Simulated Car Racing ChampionshipIEEE Transactions on Computational Intelligence and AI in Games, Vol. 2, pp. 131 - 147, June 2010. [pdf]

2009

  • Diego Perez, Gustavo Recio, Yago Saez and Pedro Isasi. Evolving a Fuzzy Controller for a Car Racing Competition, In Proc. IEEE Symposium on Computational Intelligence and Games, pp. 263-270 (2009). [pdf]

2008

  • Yago Saez, Diego Perez, Oscar Sanjuan and Pedro Isasi. Driving Cars by Means of Genetic AlgorithmsProceedings of the 10th international conference on Parallel Problem Solving from Nature, SPRINGER-VERLAG Berlin, Lecture notes in Computer Science, Vol. 5199, pp. 1101-1110, 2008. [pdf]
  • Diego Perez, Yago Saez, Gustavo Recio and Pedro Isasi. Evolving a Rule System Controller for Automatic Driving in a Car RacingProc. IEEE Symposium on Computational Intelligence and Games, pp. 336 - 342, 2008 Perth, Australia. [pdf]
  • Daniele Loiacono, Julian Togelius, Pier Luca Lanzi, L. Kinnaird-Heether, Simon Lucas, Matt Simmerson, Diego Perez, R. G. Reynolds and Yago Saez. The WCCI 2008 Simulated Car Racing CompetitionProc. IEEE Symposium on Computational Intelligence and Games, pp. 119-126, 2008 Perth, Australia. [pdf]

Study areas

 Computational Intelligence for Games, Reinforcement Learning, Evolutionary Computation.

Thesis titleAdaptive Controllers for Real Time Games
Abstract

This thesis analyzes the performance of Artificial Intelligence techniques to create adaptive controllers that play real-time games. Research in games has been a prolific field of study in the last decade, and multiple approaches have been suggested in the literature. Real-time games are a subset of the games used as testbeds in the field, and they establish a constraint that make them interesting: a small time budget, mere milliseconds, for the algorithm to decide the next action to take. With this limited time, the algorithm is unable to explore a significant part of the search space, leaving terminal states of the game far from the simulation horizon.

Adaptive controllers for real-time games are proposed in this thesis that mainly employ two algorithms: Monte Carlo Tree Search and Evolutionary Algorithms. These approaches incorporate solutions to the problems posed by real-time scenarios, and they are thoroughly tested in several games. This thesis also proposes the use of adaptive controllers as a tool to automatically generate content for games, and introduces an initial study into employing adaptive controllers for General Video Game Playing, an area of research where game specific heuristics are not available.

The experimental work conducted in this research examines and compares different approaches, showing the benefits of the proposed algorithms and modifications. Among the main contributions of this thesis, it is worth highlighting the coarsening of the action space via macro-actions, the use of evolutionary algorithms with a rolling horizon function for real-time control, the definition of a novel multi-objective algorithm for real-time games as well as knowledge discovery and re-use of past experiences without domain specific information.

<< Back