Attention-based Cognitive Architecture

December 17,  2015, 10 am

138-511-2

Abstract:

 

Cognitive robots are desired to help humans to achieve tasks in open and complex environments. They do not only require robust perception, manipulation and navigation skills to achieve the tasks, but they also require flexible general control architectures to make and learn decisions in order to interact with the environments.

Cognitive architecture is an emerging research that investigates how various skills, capabilities and knowledge can be integrated to solve open and complex problems. It has been extensively studied to model human cognition. Recently, research in cognitive robotics has been developing cognitive architectures to provide robots with robust and flexible interactions.

In this talk, we will present the Attentive Self-Modifying (ASMO) Cognitive Architecture designed based on theories in cognitive science. ASMO has been used in several robotic platforms, including Willow Garage PR2, Aldebaran Nao, Sony AIBO and custom made bear robot. We will describe how ASMO models theories of attention and how it can be integrated with emotion and learning. We will demonstrate how ASMO is used in robot soccer and companion robot applications.