About me
I am a PhD student in the group, with interests in AI-bots for video games and the dynamic alteration of video games for improved gameplay.
The aim of my PhD project is to show that player behaviour in games can be meaningfully classified into subgroups, and that rule sets of the games in consideration can be adapted to improve the enjoyability for players in each considered subgroup. Predictive machine learning can then be used to classify new players during the first minutes of play and adapt the game rules to apply the rule changes learned earlier to increase enjoyability. This process can be divided into several subprojects that act as milestones during the work:
Data collection in a video game: A suitable platform for behaviour analysis has to be found and adapted. The chosen video game will be implemented / adapted for easy and efficient data collection and user polls.
Sub-grouping and Clustering of game data: The collected game data will be used to create groups of players. The measures which are most distinguishing can be used to assign attributes to the groups in order for game developers to understand the separation.
Predictive machine learning: Given the behaviour groups, incomplete data from a player can be used to predict the group in which the current player fits best. The earlier this distinction is possible, the sooner the software can be adapted to the specific group.
Iterative adaptation learning: Adapting software to a user is not straight forward. A priori it is not clear which modifications are satisfying for the user.
Generalisation of framework: Several games will be used to show that the user groups are game-independent, i.e., if two players are classified into a group for one game, it is likely that they are classified in a group for another game.
Publications
Browne, Cameron; Colton, Simon; Cook, Michael; Gow, Jeremy; Baumgarten, Robin Toward the Adaptive Generation of Bespoke Game Content Book Chapter In: IEEE Handbook of Digital Games , pp. 15–61, John Wiley & Sons, Inc., 2014. @inbook{browne2014toward, title = {Toward the Adaptive Generation of Bespoke Game Content}, author = { Cameron Browne and Simon Colton and Michael Cook and Jeremy Gow and Robin Baumgarten}, url = {http://ccg.doc.gold.ac.uk/wp-content/uploads/2016/10/browne_ieeechapter14-2.pdf}, year = {2014}, date = {2014-01-01}, journal = {Handbook of Digital Games}, volume = {IEEE Handbook of Digital Games}, pages = {15--61}, publisher = {John Wiley & Sons, Inc.}, abstract = {In this chapter, we explore methods for automatically generating game content — and games themselves — adapted to individual players, in order to improve their playing experience or achieve a desired effect. This goes beyond notions of mere replayability, and involves modelling player needs to maximise their enjoyment, in- volvement and interest in the game being played. We identify three main aspects of this process: Generation of new content and rule sets; Measurement of this content and the player; Adaptation of the game to change player experience. This process forms a feedback loop of constant refinement, as games are continually improved while being played. Framed within this methodology, we present an overview of our recent and ongoing research in this area. This is illustrated by a number of case studies that demonstrate these ideas in action over a variety of game types, includ- ing: 3D action games, arcade games, platformers, board games, puzzles and open world games. We draw together some of the lessons learned from these projects to comment on the difficulties, the benefits and the potential for personalised gaming via adaptive game design.}, keywords = {}, pubstate = {published}, tppubtype = {inbook} } In this chapter, we explore methods for automatically generating game content — and games themselves — adapted to individual players, in order to improve their playing experience or achieve a desired effect. This goes beyond notions of mere replayability, and involves modelling player needs to maximise their enjoyment, in- volvement and interest in the game being played. We identify three main aspects of this process: Generation of new content and rule sets; Measurement of this content and the player; Adaptation of the game to change player experience. This process forms a feedback loop of constant refinement, as games are continually improved while being played. Framed within this methodology, we present an overview of our recent and ongoing research in this area. This is illustrated by a number of case studies that demonstrate these ideas in action over a variety of game types, includ- ing: 3D action games, arcade games, platformers, board games, puzzles and open world games. We draw together some of the lessons learned from these projects to comment on the difficulties, the benefits and the potential for personalised gaming via adaptive game design. |
Gow, Jeremy; Baumgarten, Robin; Cairns, Paul A; Colton, Simon; Miller, Paul Unsupervised Modeling of Player Style With LDA Journal Article In: IEEE Trans. Comput. Intellig. and AI in Games, 4 (3), pp. 152–166, 2012. @article{DBLP:journals/tciaig/GowBCCM12, title = {Unsupervised Modeling of Player Style With LDA}, author = {Jeremy Gow and Robin Baumgarten and Paul A. Cairns and Simon Colton and Paul Miller}, url = {http://ccg.doc.gold.ac.uk/wp-content/uploads/2016/03/gow_tciaig12.pdf}, doi = {10.1109/TCIAIG.2012.2213600}, year = {2012}, date = {2012-01-01}, journal = {IEEE Trans. Comput. Intellig. and AI in Games}, volume = {4}, number = {3}, pages = {152--166}, abstract = {Computational analysis of player style has signifi- cant potential for video game design: it can provide insights into player behaviour, as well as the means to dynamically adapt a game to each individual’s style of play. To realise this potential, computational methods need to go beyond considerations of challenge and ability and account for aesthetic aspects of player style. We describe here a semi-automatic unsupervised learning approach to modelling player style using multi-class Linear Discriminant Analysis (LDA). We argue that this approach is widely applicable for modelling player style in a wide range of games, including commercial applications, and illustrate it with two case studies: the first for a novel arcade game called Snakeotron, the second for Rogue Trooper, a modern commercial third-person shooter video game.}, keywords = {}, pubstate = {published}, tppubtype = {article} } Computational analysis of player style has signifi- cant potential for video game design: it can provide insights into player behaviour, as well as the means to dynamically adapt a game to each individual’s style of play. To realise this potential, computational methods need to go beyond considerations of challenge and ability and account for aesthetic aspects of player style. We describe here a semi-automatic unsupervised learning approach to modelling player style using multi-class Linear Discriminant Analysis (LDA). We argue that this approach is widely applicable for modelling player style in a wide range of games, including commercial applications, and illustrate it with two case studies: the first for a novel arcade game called Snakeotron, the second for Rogue Trooper, a modern commercial third-person shooter video game. |
Ramirez-Cano, Daniel; Colton, Simon; Baumgarten, Robin Player Classification Using a Meta-Clustering Approach Inproceedings In: Proceedings of the 3rd Annual International Conference Computer Games, Multimedia & Allied Technology, 2010. @inproceedings{RamirezCano2010, title = {Player Classification Using a Meta-Clustering Approach}, author = {Daniel Ramirez-Cano and Simon Colton and Robin Baumgarten}, url = {http://ccg.doc.gold.ac.uk/wp-content/uploads/2016/10/ramirez_cgat10.pdf}, year = {2010}, date = {2010-10-01}, booktitle = {Proceedings of the 3rd Annual International Conference Computer Games, Multimedia & Allied Technology}, abstract = {Player classification has recently become a key aspect of game design in areas such as adaptive game systems, player behaviour prediction, player tutoring and non-player character design. Past research has focused on the design of hierarchical, preference- based and probabilistic models aimed at modelling players' behaviour. We propose a meta-classification approach that breaks the clustering of gameplay mixed data into three levels of analysis. The first level uses dimensionality reduction and partitional clustering of aggregate game data in an action/skill- based classification. The second level applies similarity-based clustering of action sequences to group players according to their preferences. For this we propose a new approach which uses Rubner’s Earth Mover’s Distance (EMD) as a similarity metric to compare histograms of players’ game world explorations. The third level applies a combination of social network analysis metrics, such as shortest path length, to social data to find clusters in the players' social network. We test our approach in a gameplay dataset from a freely available first-person social hunting game}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Player classification has recently become a key aspect of game design in areas such as adaptive game systems, player behaviour prediction, player tutoring and non-player character design. Past research has focused on the design of hierarchical, preference- based and probabilistic models aimed at modelling players' behaviour. We propose a meta-classification approach that breaks the clustering of gameplay mixed data into three levels of analysis. The first level uses dimensionality reduction and partitional clustering of aggregate game data in an action/skill- based classification. The second level applies similarity-based clustering of action sequences to group players according to their preferences. For this we propose a new approach which uses Rubner’s Earth Mover’s Distance (EMD) as a similarity metric to compare histograms of players’ game world explorations. The third level applies a combination of social network analysis metrics, such as shortest path length, to social data to find clusters in the players' social network. We test our approach in a gameplay dataset from a freely available first-person social hunting game |
Gow, Jeremy; Cairns, Paul; Colton, Simon; Miller, Paul; Baumgarten, Robin Capturing Player Experience With Post-Game Commentaries Inproceedings In: Proceedings 3rd International Conference on Computer Games, Multimedia & Allied Technology (CGAT 2010), 2010. @inproceedings{Gow2010PlayerExperience, title = {Capturing Player Experience With Post-Game Commentaries}, author = {Jeremy Gow and Paul Cairns and Simon Colton and Paul Miller and Robin Baumgarten}, url = {http://ccg.doc.gold.ac.uk/wp-content/uploads/2016/10/gow_cgat10.pdf}, year = {2010}, date = {2010-10-01}, booktitle = {Proceedings 3rd International Conference on Computer Games, Multimedia & Allied Technology (CGAT 2010)}, abstract = {Player experience is at the heart of good game design, but designers typically have limited experience data to work with. Detailed and fine-grained accounts of gaming experience would be of great value to designers and researchers alike, but recording such data is a significant challenge. We describe an approach based on post-game player commentaries, retrospective verbal reports cued by video of the gaming session and a word list. A pilot study was carried out to capture player experience of a tutorial level for a third person shooter game. We show how the technique can be used to provide useful game design feedback.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Player experience is at the heart of good game design, but designers typically have limited experience data to work with. Detailed and fine-grained accounts of gaming experience would be of great value to designers and researchers alike, but recording such data is a significant challenge. We describe an approach based on post-game player commentaries, retrospective verbal reports cued by video of the gaming session and a word list. A pilot study was carried out to capture player experience of a tutorial level for a third person shooter game. We show how the technique can be used to provide useful game design feedback. |
Martin, Andrew; Lim, Andrew; Colton, Simon; Baumgarten, Robin Evolving 3D Buildings for the Prototype Video Game Subversion Inproceedings In: Proceedings of the 2nd EuropeanEvent on Bio-inspired Algorithms in Games (EVOGAMES) , 2010. @inproceedings{Martin2010EvoGames, title = {Evolving 3D Buildings for the Prototype Video Game Subversion}, author = {Andrew Martin and Andrew Lim and Simon Colton and Robin Baumgarten}, url = {http://ccg.doc.gold.ac.uk/wp-content/uploads/2016/10/martin_evogames10.pdf}, year = {2010}, date = {2010-10-01}, booktitle = {Proceedings of the 2nd EuropeanEvent on Bio-inspired Algorithms in Games (EVOGAMES) }, abstract = {We investigate user-guided evolution for the development of virtual 3D building structures for the prototype (commercial) game Subversion, which is being developed by Introversion Software Ltd. Build- ings are described in a custom plain-text markup language that can be parsed by Subversion's procedural generation engine, which renders the 3D models on-screen. The building descriptions are amenable to ran- dom generation, crossover and mutation, which enabled us to implement and test a user-driven evolutionary approach to building generation. We performed some fundamental experimentation with ten participants to determine how visually similar child buildings are to their parents, when generated in differing ways. We hope to demonstrate the potential of user-guided evolution for content generation in games in general, as such tools require very little training, time or effort to be employed effectively.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } We investigate user-guided evolution for the development of virtual 3D building structures for the prototype (commercial) game Subversion, which is being developed by Introversion Software Ltd. Build- ings are described in a custom plain-text markup language that can be parsed by Subversion's procedural generation engine, which renders the 3D models on-screen. The building descriptions are amenable to ran- dom generation, crossover and mutation, which enabled us to implement and test a user-driven evolutionary approach to building generation. We performed some fundamental experimentation with ten participants to determine how visually similar child buildings are to their parents, when generated in differing ways. We hope to demonstrate the potential of user-guided evolution for content generation in games in general, as such tools require very little training, time or effort to be employed effectively. |
Lim, Chong-U.; Baumgarten, Robin; Colton, Simon Evolving behaviour trees for the commercial game DEFCON Inproceedings In: Proceedings of the 2nd European eEvent on Bio-inspired Algorithms in Games (EVOGAMES) , 2010. @inproceedings{Lim2010, title = {Evolving behaviour trees for the commercial game DEFCON}, author = {Chong-U. Lim and Robin Baumgarten and Simon Colton}, url = {http://ccg.doc.gold.ac.uk/wp-content/uploads/2016/10/lim_evogames10.pdf}, year = {2010}, date = {2010-10-01}, booktitle = {Proceedings of the 2nd European eEvent on Bio-inspired Algorithms in Games (EVOGAMES) }, abstract = {Behaviour trees provide the possibility of improving on existing Artificial Intelligence techniques in games by being simple to implement, scalable, able to handle the complexity of games, and modular to improve reusability. This ultimately improves the development process for designing automated game players. We cover here the use of behaviour trees to design and develop an AI-controlled player for the commercial real-time strategy game DEFCON. In particular, we evolved behaviour trees to develop a competitive player which was able to outperform the game’s original AI-bot more than 50% of the time. We aim to highlight the potential for evolving behaviour trees as a practical approach to developing AI-bots in games.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Behaviour trees provide the possibility of improving on existing Artificial Intelligence techniques in games by being simple to implement, scalable, able to handle the complexity of games, and modular to improve reusability. This ultimately improves the development process for designing automated game players. We cover here the use of behaviour trees to design and develop an AI-controlled player for the commercial real-time strategy game DEFCON. In particular, we evolved behaviour trees to develop a competitive player which was able to outperform the game’s original AI-bot more than 50% of the time. We aim to highlight the potential for evolving behaviour trees as a practical approach to developing AI-bots in games. |
Baumgarten, Robin; Colton, Simon; Morris, Mark Combining AI Methods for Learning Bots in a Real-Time Strategy Game Journal Article In: International Journal of Computer Games Technology, 2009. @article{Baumgarten2009Combining, title = {Combining AI Methods for Learning Bots in a Real-Time Strategy Game}, author = {Robin Baumgarten and Simon Colton and Mark Morris}, url = {http://ccg.doc.gold.ac.uk/wp-content/uploads/2016/10/baumgarten_ijcgt09.pdf}, year = {2009}, date = {2009-10-01}, journal = {International Journal of Computer Games Technology}, abstract = {We describe an approach for simulating human game-play in strategy games using a variety of AI techniques, including simulated annealing, decision tree learning, and case-based reasoning. We have implemented an AI-bot that uses these techniques to form a novel approach for planning fleet movements and attacks in DEFCON, a nuclear war simulation strategy game released in 2006 by Introversion Software Ltd. The AI-bot retrieves plans from a case-base of recorded games, then uses these to generate a new plan using a method based on decision tree learning. In addition, we have implemented more sophisticated control over low-level actions that enable the AI-bot to synchronize bombing runs, and used a simulated annealing approach for assigning bombing targets to planes and opponent cities to missiles.We describe how our AI-bot operates, and the experimentation we have performed in order to determine an optimal configuration for it. With this configuration, our AI-bot beats Introversion’s finite state machine automated player in 76.7% of 150 matches played.We briefly introduce the notion of ability versus enjoyability and discuss initial results of a survey we conducted with human players.}, keywords = {}, pubstate = {published}, tppubtype = {article} } We describe an approach for simulating human game-play in strategy games using a variety of AI techniques, including simulated annealing, decision tree learning, and case-based reasoning. We have implemented an AI-bot that uses these techniques to form a novel approach for planning fleet movements and attacks in DEFCON, a nuclear war simulation strategy game released in 2006 by Introversion Software Ltd. The AI-bot retrieves plans from a case-base of recorded games, then uses these to generate a new plan using a method based on decision tree learning. In addition, we have implemented more sophisticated control over low-level actions that enable the AI-bot to synchronize bombing runs, and used a simulated annealing approach for assigning bombing targets to planes and opponent cities to missiles.We describe how our AI-bot operates, and the experimentation we have performed in order to determine an optimal configuration for it. With this configuration, our AI-bot beats Introversion’s finite state machine automated player in 76.7% of 150 matches played.We briefly introduce the notion of ability versus enjoyability and discuss initial results of a survey we conducted with human players. |
Baumgarten, Robin; Nika, Maria; Gow, Jeremy; Colton, Simon Towards the Automatic Invention of Simple Mixed Reality Games Inproceedings In: Proc. of the AISB’09 Symp. on AI and Games, 2009. @inproceedings{baumgarten2009towards, title = {Towards the Automatic Invention of Simple Mixed Reality Games}, author = { Robin Baumgarten and Maria Nika and Jeremy Gow and Simon Colton}, url = {http://ccg.doc.gold.ac.uk/wp-content/uploads/2016/08/colton_evomusart08.pdf}, year = {2009}, date = {2009-01-01}, booktitle = {Proc. of the AISB’09 Symp. on AI and Games}, abstract = {We investigate the automatic construction of visual scenes via a hybrid evolutionary/hill-climbing approach using a correlation- based fitness function. This forms part of The Painting Fool system, an automated artist which is able to render the scenes using simulated art materials. We further describe a novel method for inventing fitness functions using the HR descriptive machine learning system, and we com- bine this with The Painting Fool to generate and artistically render novel scenes. We demonstrate the potential of this approach with applications to cityscape and flower arrangement scene generation.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } We investigate the automatic construction of visual scenes via a hybrid evolutionary/hill-climbing approach using a correlation- based fitness function. This forms part of The Painting Fool system, an automated artist which is able to render the scenes using simulated art materials. We further describe a novel method for inventing fitness functions using the HR descriptive machine learning system, and we com- bine this with The Painting Fool to generate and artistically render novel scenes. We demonstrate the potential of this approach with applications to cityscape and flower arrangement scene generation. |
Baumgarten, Robin; Colton, Simon Case-based Player Simulation for the Commercial Strategy Game DEFCON Inproceedings In: Proceedings of CGames, 2007. @inproceedings{Baumgarten2007, title = {Case-based Player Simulation for the Commercial Strategy Game DEFCON}, author = {Robin Baumgarten and Simon Colton}, url = {http://ccg.doc.gold.ac.uk/wp-content/uploads/2016/11/baumgarten_cgames07.pdf}, year = {2007}, date = {2007-11-01}, booktitle = {Proceedings of CGames}, abstract = {DEFCON is a nuclear war simulation strategy game released in 2006 by Introversion Software Ltd. We describe an approach to simulating human game-play using a variety of AI techniques, including simulated annealing, decision tree learning and case-based reasoning. We have implemented an AI-bot that uses a novel approach to planning fleet movements and attacks. This retrieves plans from a case base of previous games, then merges these using a method based on decision tree learning. In addition, we have implemented more sophisticated control over low-level actions such as firing missiles and guiding planes. In particular, we have written routines to enable the AI-bot to synchronise bombing runs, and enabled a simulated annealing approach to assigning bombing targets to planes and opponent cities to missiles. We describe how our AI-bot operates, and the experimentation we have performed in order to determine an optimal configuration for it. With this configuration, our AI- bot beat Introversion’s finite state machine automated player in 76.7% of 150 matches played.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } DEFCON is a nuclear war simulation strategy game released in 2006 by Introversion Software Ltd. We describe an approach to simulating human game-play using a variety of AI techniques, including simulated annealing, decision tree learning and case-based reasoning. We have implemented an AI-bot that uses a novel approach to planning fleet movements and attacks. This retrieves plans from a case base of previous games, then merges these using a method based on decision tree learning. In addition, we have implemented more sophisticated control over low-level actions such as firing missiles and guiding planes. In particular, we have written routines to enable the AI-bot to synchronise bombing runs, and enabled a simulated annealing approach to assigning bombing targets to planes and opponent cities to missiles. We describe how our AI-bot operates, and the experimentation we have performed in order to determine an optimal configuration for it. With this configuration, our AI- bot beat Introversion’s finite state machine automated player in 76.7% of 150 matches played. |