Supervisor: Ilaria Tiddi (i.tiddi@vu.nl)
Gaze detection plays a pivotal role in Human-Robot Interaction (HRI), especially in tasks where robots are required to infer human intentions based on their gaze direction. For safety and convenience, robotic applications are often developed and tested in simulation environments. In this project, we aim to investigate how gaze detection methods, originally designed for real humans, perform when applied to simulated humans in robotic simulation software. The project will involve identifying and preparing some of the most prominent gaze detection algorithms, alongside selecting widely-used simulation environments such as Gazebo and Webots. We will design an experimental setup where different gaze directions are systematically tested in both simulated human models and real-life humans. By comparing the performance of gaze detection algorithms in these two settings, we aim to evaluate the degree to which simulations accurately replicate human gaze behaviours.
RQ: To what extent do gaze detection algorithms vary in their performance between robotic simulations and real-life scenarios?