An Introspection Approach to Self-Awareness in Robotics

This PhD project is about investigating computational introspection for self-awareness in robotics. The main goal is to devise an approach to introspection for interpretable artificial self-awareness. Computational introspection will allow robots to understand their limitations, adapt accordingly and reason about themselves or even how to gain new knowledge. In combination with sensory information, self-aware robots will be able see and understand their body and adjust it to better suite a given task. Therefore, this PhD builds on the notion that through understanding themselves, robots can reconfigure their abilities to better suit the purpose they serve in the environment in which they act, after considering their current performance and their mechanical limitations. Self-aware robots will therefore require less direct expert human intervention during task execution.