Comparative cognitive robotics: Using autonomous robots as empirical models of animal learning

In this article, we explore a methodology to use autonomous mobile robots as empirical models in the comparative psychology of learning and adaptation. Starting from the successful example of biorobotics which uses robots as models in biology, we propose an extended methodology called “comparative cognitive robotics”. Instead of biology and ethology, we cooperate with comparative psychology in the process of constructing empirical robot models. Like in biorobotics, the robot models and the animals to be modeled are tested in the same experiments within the same experimental environments, applying the same means of behavioral analysis. We use the experimentally controlled environments of comparative psychology to test both animal subjects and robot models. This way, we do not need to resort to the abstracted description of a human observer to provide input to our model or to interpret its output, which might otherwise bias our results. As an application of our new methodology, we describe the construction of an empirical autonomous robot model of visual discrimination learning in chickens and report the results of the first comparisons between the robot and the animal performance in the same behavioral expenments. Due to emergent properties of its interaction with the environment, our robot model does not have to use abstracted representations of features or stimulus categories, or associative strengths altered by learning rules. It can thus be considered to be more parsimonious than previous, non-robotic approaches to model discrimination learning.