Designing Privacy-Preserving Visual Perception for Robot Navigation Based on User Privacy Preferences

A research team from the University of Bonn, led by Xuying Huang, has published a paper proposing a novel, user-centered framework for designing privacy-preserving visual perception systems for mobile service robots. The core problem is that the cameras robots need for navigation inherently capture sensitive information, raising significant privacy concerns. The researchers argue that existing technical solutions are not sufficiently grounded in what users actually want, so they conducted two user studies to directly inform the design.

The studies revealed that users strongly prefer privacy-preserving visual abstractions (like blurring or obfuscation) and mechanisms that apply low-resolution capture at the moment of image acquisition. Crucially, the team found that a user's preferred camera resolution depends on two key factors: their desired level of privacy *and* the robot's physical proximity during navigation. A person might accept higher detail when a robot is far away in a hallway but demand heavy pixelation as it approaches their desk.

Based on these findings, the researchers formalized a 'user-configurable distance-to-resolution privacy policy.' This policy allows individuals to set rules, such as 'switch to very low resolution when within 2 meters,' creating a dynamic system where visual fidelity adjusts in real-time based on spatial context and personal preference. This moves the field from static, one-size-fits-all technical filters toward adaptive, human-in-the-loop privacy management for robotic systems.