Implementation of cognitive control in a humanoid robot

This dissertation presents design and implementation of a human-inspired cognitive control ability within a humanoid robot ISAC. In cognitive psychology, cognitive control is defined as the ability for humans to focus on the task at hand while being able to control and adapt behaviors to respond to uncommon situations. This cognitive control functions were implemented through the Self Agent, a software agent responsible for task execution and control within the ISAC multi-agent architecture. Two atomic agents called the First-order Response Agent (FRA) and the Central Executive Agent (CEA) work together during task execution using working memory and long-term memory. CEA provides cognitive control functions to the rest of the system by suppressing routine responses generated by FRA, and deliberately comes up with more appropriate responses based on past experiences. Additionally, CEA has the ability to select and switch tasks based on their utility values. The performance of the system was evaluated through a set of experiments which showed that implementation of cognitive control allowed ISAC to decide to switch tasks at appropriate time and execute tasks with improved robustness over both routine and deliberative task execution.