Investigation of Human First-Person Guidance Strategy from Gaze Tracking Data

This paper investigates human guidance strategies in first-person guidance tasks. A first-person guidance simulation setup is used, which records human gaze, vehicle trajectory and control inputs. For the evaluation of human guidance strategy, an optimal control framework that is formulated using the interaction patterns embedded in the corresponding optimal solution is presented. A metric called ISE (instantaneous sub goal error) is proposed to evaluate human guidance strategies. This paper also investigates the relationship among vehicle dynamic behavior, human gaze and environment.