Modeling Human-Drone Proxemics in an Augmented Reality Environment

Abstract

As drones increasingly move into human-occupied spaces, they must be able to navigate in ways that are not only safe but also socially acceptable. This thesis investigates human-drone proxemics in an augmented reality (AR) environment, with the aim of modeling the interpersonal distances people prefer when approached by a social drone. Building on prior two-dimensional rubber-sheet models from human-robot proxemics, this work extends the approach into a three-dimensional “rubber-bubble” model suitable for drones. A user study was conducted in Sweden and Vietnam (N = 69), where participants interacted with a virtual drone approaching from 21 different angles in an area in front of them. Participants indicated when the drone reached their preferred stopping distance, producing a dataset used to evaluate approach-angle effects, demographic variables as predictors, cultural differences, and various degrees of personalization. The methodology combined statistical testing, polynomial rubber-bubble models, linear mixed models, and Bayesian modeling for an exhaustive and exploratory analysis. The results show that rubber-bubble models can represent individualized proxemic preferences within the observed frontal zone, suggesting usefulness for autonomous drone navigation. Gender and culture both showed potential in predicting stopping distances. In summary, this work contributes with an extension of 2D rubber-sheet to 3D rubber-bubble modeling, a thorough analysis of demographics and culture as potential predictors for human-drone proxemics, and an open-source AR environment for safe and repeatable human-drone research.