Genetic Algorithms for human-like expressions in an autonomous vehicle

Abstract

Advancing AI technologies have bridged the gap between science fiction and reality, and a future where advanced human-robotic interaction is common doesn’t seem far away. While AI models are getting smarter in conversation and reasoning they are still missing a crucial piece in human communication, namely facial expressions. In this project, we investigate the prospect of using a Genetic Algorithm with Behavior Trees to create the illusion of having emotions for an autonomous vehicle. The vehicle, called Will-e, determines on its own an appropriate emotional state and then expresses those emotions using LED-matrices in the form of eyes and a mouth. We implement these algorithms, train them, and test them in a populated setting and explore how people interact with the robot.