Model-Free Algorithms to Produce Assistance for Human Operators Multi-tasking with Machines
Lead Presenter: Keivan Sadeghzadeh
Faculty Advisor/Principal Investigator: Rifat Sipahi
Method of Presentation: Poster
Mathematical modeling of human dynamic behavior is extremely difficult, if not impossible, since many aspects of humans are still unknown. Under this condition, our question is that can we still develop engineering tools that can help us control the interactions arising between human operators and machines? In this project, under the funding of DARPA (Defense Advanced Research Projects Agency), we develop novel “model-free” control algorithms that will provide assistance and guidance to human operators engaged with multiple tasks with machines, and thereby to improve performance harvested from such human-machine interactions. These algorithms are envisioned to not only provide assistance but also to reduce human cognitive overload, and ultimately preventing catastrophic events that may otherwise occur under increased cognitive overload. In this presentation, we primarily discuss the state of the art, our experimental setup, instrumentations, the designed Java based air-traffic-control game, how we plan to “close the loop”, and an approach to assess cognitive load via its inverse correlation to performance metrics of the game we designed.