Simulated gaze lets social robots be tested without people at first

The University of Surrey and the University of Hamburg built a model that predicts where a person would look in a social scene and used it to drive a humanoid robot's eye movements. Tested on two public datasets, the robot's attention matched recorded human gaze patterns and held up in noisy settings. The researchers propose using such simulations in place of early human trials, which are slow and costly to run. If adopted, the approach could speed up iteration on social robot behaviour before user studies begin.

Source: Tech Xplore (University of Surrey)