Adapting Playgrounds using Multi-Agent Systems

This paper introduces an approach on how versatile, dynamic and adaptive playgrounds can be developed using multi-agent systems (MAS), artificial intelligence (AI) and Playware technology. By modelling the children and the Playware playground into a MAS, knowledge about the children’s favourite playing behaviour can be acquired. Experiments with children were conduced in order to record their playing behaviours on the Playware playground, which consists of tiles capable of sensing and actuation, and having abilities to communicate with neighbouring tiles. An ANN capable of classifying the children’s behaviour within eleven categories (i.e. favourite playing behaviour) was trained using a subgroup of the children. Validating the ANN against the remaining children’s behaviours, 93% were correctly classified. An adaptive playground was implemented utilizing the ANN in classifying the children’s behaviour real-time and thus allowing for the children’s interest in the playground to be maintained and/or increased by using appropriate adaptation strategies.