My Research

On social learning and synthetic agents in game playing multi-agent systems

The objective of this research is to study agent socialization in large scale social events, with such events being in the form of game-playing Multi Agent Systems (MAS).

For our initial work on the subject of MAS socialization we implemented a gaming tournament, with players using reinforcement learning for playing and learning.

There was initial evidence that a socially trained agent performs better against self-trained agents and there was, also, evidence of interesting performance by minimax-based players. We also observed that socially-trained agents performed better than self-trained ones, supporting our initial belief that richer social interaction creates a stronger playing behavior.

On one hand, the performance of synthetic agents in playing and learning scenarios in a turn-based zero-sum game highlights the ability of opponent-based learning models to demonstrate competitive playing performances in social environments.

On the other hand, by observing clusters of performance of synthetic agents, allows one to investigate how to best select opponents from a group; initial results suggest that good progress and stable evolution arise when an agent faces opponents of increasing capacity, and that an agent with a good learning mechanism progresses better when it faces less favorably set-up agents.

• Kiourt, C. and Kalles, D.: Synthetic learning agents in game playing social environments, Adaptive Behaviors, Vol. 24, No. 6, p 411-427, 2016.
• Kiourt, C. and Kalles, D.: Social Reinforcement Learning in Game Playing. IEEE International Conference on Tools with Artificial Intelligence (ICTAI 2012), Athens, pp 322-326, Nov. 7-9, Greece, 2012.

On opponent based learning/training

The agents' playing behavior is a key factor in social organizations for the simulation of more realistic environments. However, to achieve this, one must first investigate behavior versatility in the same environment, especially in the way that differences in learning style can produce a variety of playing behaviors. This helps set the ground for subsequent studies on opponent selection.

• C. Kiourt and D. Kalles: Using Opponent Models to Train Inexperienced Synthetic Agents in Social Environments, IEEE Conference on Computational Intelligence and Games (IEEE CIG), p 174-177, Santorini, Greece, 20-23 September, 2016.
• C. Kiourt and D. Kalles: Learning in Multi Agent Social Environments with Opponent Models, 13th European Conference on Multi-Agent Systems, pp 137-144, Athens, Greece, 17-18 December, 2015.

On gamification and virtual environments for cultural and educational purposes

Gaming for educational purposes is a significant and active research domain. This has taken either the form of game-based learning or serious gaming. The importance of playing has been emphasized in many studies from various domains.

- Playing is an archetypical activity that arises from primordial biological structures existing before the conscience or the capacity for speech; it is not something a person decides to do.

- Gamification is nothing more than the use of specific game design approaches and techniques in various environments, in order to attract people in problem solving and to enhance their contribution

We study new generation methods to bring closer the students in education and culture with a pleasant ways, which seems to be the Serius Games, very well known from student. In a sense, SGs can be considered as an efficient approach for blending domain specific activities, like in cultural heritage and education, with gaming. By utilising contemporary visualisation and simulation technologies SGs enhance the user’s experience through photorealistic interactive environments

We developed several application in this area, such as the Synthesis virtual museum (link) or the DynaMus (link), where we try to combine knowledge, culture and art with a pleasant way for the student to build its critical, thinking and development skill based an innovative cunstrative training methods.

• Kiourt, C.,Koutsoudis, A., and Pavlidis, G.: DynaMus: A fully dynamic 3D virtual museum framework, Journal of Cultural Heritage, Vol. 22, p 984-991 , 2016.
• Petsa, G., Kiourt, C. , Koutsoudis, A., Arnaoutoglou, F., Markantonatou, S., Pavlidis, G., Towards a unified cultural and educational portal prototype for museums and exhibitions, Poster display, International Workshop on Virtual Archaeology: Museums & Cultural Tourism, 23-26 September 2015, Delphi, Greece.
• Kiourt, C.,Koutsoudis, A., Arnaoutoglou, F., Markantonatou, S. and Pavlidis, G., The 'Synthesis' Virtual Museum – an open virtual exhibition creation tool, International Workshop on Virtual Archaeology: Museums & Cultural Tourism, 23-26 September 2015, Delphi, Greece.
• Kiourt, C. , Koutsoudis, A., Arnaoutoglou, F., Markantonatou, S. and Pavlidis, G., A dynamic web-based 3D virtual museum framework based on open data, Digital Heritage 2015 , 28 September – 2 October 2015, Granada, Spain.