Modelling Affordances for the Control and Evaluation of Intrinsically Motivated Robots
Kathryn Elizabeth Merrick · 2009
In psychological theory, affordances provide a way to describe an environment in terms of the opportunities it provides an organism to act. Affordance-based models have been applied to robotics in areas such as tool-use, interaction and vision, as an alternative to hybrid control architectures. This paper introduces a model of affordances for controlling and evaluating intrinsically motivated robots. We present an implementation and experimental evaluation of the model on four critter-bots using the Lego Mindstorms NXT platform. Results show that the critter-bots can evolve measurable, structured behaviour cycles adapted to their individual physical structures. 1