Gamifying Reinforcement-based Learning
This project is already completed.
Motivation:
Reinforcement Learning [1] is an adaptive process in which an agent uses its past experiences to improve the outcomes of its future decisions. Several aspects of motivation and cognition, i.e. goal-directed control, approach-avoidance behavior, can be accounted by formal reinforcement learning algorithms. These methods constitute a formal framework but need to be endorsed by several empirical experiments, such as reversal learning, sequential decision-making or go-no-go tasks. The suitability of the testing environment, however, is usually not part of the research question.
From a Games Engineering perspective, we can already notice the similarities in terminology between Game Elements (Agent, Game World, etc.) and Reinforcement Learning Elements (Agent, Environment, etc.). Indeed, a game is composed of elements such as a Game World, Goals, Agents and Mechanics. Agents, in games, perform specific Mechanics or Actions in order to achieve specific Goals and, thus, change the Game State. In Reinforcement Learning [2], the Agent interacts with the Environment by performing a specific set of Actions resulting in getting a Reward or a Punishment and, thus, reaching a new State.
Using the power and versatility of games, we want to associate each game element [3] to its corresponding element from an RL perspective. This mapping would give us the ability to design and develop games for testing and simulating the experiments developed within the RL framework. To test the effectiveness of this approach, we need to design and develop a game for each RL method
Tasks:
Literature research
- Definition of Reinforcement Learning and its methods
- Definition of a set of RL Elements that suit the mapping approach
- Analysis and categorization of the use of RL Elements/Methods in Games
Conceptualization
- Choice of suitable RL Elements
- Association of each RL Element to its corresponding Game Element (according to the Unifying Game Ontology framework)
Implementation
- Design and Development of, at least, one game based on the RL Elements/Game Elements mapping
- Revisions and enhancement by adding visual effects and animations
Testing
- Organize play test sessions
- Data collection and analysis
Prerequisites:
- Research methods
- Principles of real-time interactive systems
- Game engines (Unity or Unreal Engine)
- Programming (C# or C++)
References:
- [1] Sutton, Richard S., and Andrew G. Barto. Introduction to reinforcement learning. Vol. 135. Cambridge: MIT press, 1998.
- [2] Szepesvári, Csaba. “Algorithms for reinforcement learning.” Synthesis lectures on artificial intelligence and machine learning 4.1 (2010): 1-103.
- [3] Debus, Michael S. Unifying game ontology: a faceted classification of game elements. IT University of Copenhagen, Pervasive Interaction Technology Lab, 2019.
Contact Persons at the University Würzburg
Prof. Dr. Sebastian von MammenGames Engineering, Universität Würzburg
sebastian.von.mammen@uni-wuerzburg.de
Mounsif Chetitah (Primary Contact Person)
Games Engineering, Universität Würzburg
mounsif.chetitah@uni-wuerzburg.de
Prof. Dr. med. Lorenz Deserno
Cognitive and Computational Neuroscience in Developmental Psychiatry, Universitätsklinikum Würzburg
Deserno_L@ukw.de
Maria Waltmann
Cognitive and Computational Neuroscience in Developmental Psychiatry, Universitätsklinikum Würzburg
Waltmann_M@ukw.de