REINFORCEMENT LEARNING WITH MODEL DRIVEN APPROACH IN GAMIFICATION MODELS OF GIN RUMMY
Anil Kumar S, K. Muralidharan · 2015
Adaptive Learning Systems, as in Robotics, require taking most optimized decisions dynamically, based on the feedback from their environment. Also they have to enhance their knowledge by experience and represent the knowledge in a reusable form. Developing the actual models of such self learning systems is quite complex, expensive and time consuming. Concepts from certain gaming and gamification models exhibit resemblance to solution models of several real world problems. A model driven approach is helpful for formulating reusable meta- models from simpler and affordable gamification models which exhibit almost similar characteristics as in the actual solution models. Experimental learning with such meta-models will help to formulate the strategy for optimizing the actual solution models as well. This study explores the scope and challenges of applying model driven approach while architecting Reinforcement Learning Solutions using meta-models from Gaming and Gamification. Simple analogies are illustrated with Gamification Models derived from gaming meta-models based on Gin Rummy Game.