Cognitive Robotics

Robotics #1
Cognitive architectures (e.g., as ACT-R; Anderson, 2007) integrate psychological assumptions about cognitive processes and knowledge representations. These architectures allow for modeling algorithmic solutions (cognitive models) with respect to precise cognitive  assumptions. ACT-R is developed by hundreds of psychologists and makes precise assumptions about the modular structure of the human mind, its knowledge representation and learning abilities. It can simulate and predict cognitive effects, response times, error rates, and produce fMRI data predictions. Cognitive architectures provide an excellent way to understand cognitive processes and to incorporate them into artificial systems.

Such an implementation of ACT-R on a cognitive robotic system has been recently pursued. We developed a Lego MindStorm robot (Bennati & Ragni, 2012) and a simple navigation module. This allows to demonstrate some of the learning mechanisms provided by ACT-R in such a simple navigation task. Other aspects are implementing qualitative calculi on a robotic system  (Moratz & Ragni, 2008) and so forth.


A tutorial for Cognitive Robotics by Marco Ragni, Rebecca Albrecht and Stefano Bennati.