Web3D Navigation & User Research

A browser-based 3D experiment connecting navigation interfaces, behavioural telemetry and user evaluation.

My contribution

Platform development, interface implementation, telemetry integration, study execution and analysis.

Context

Master’s thesis · Technical University of Munich (TUM) · June 2026. Completed under academic supervision.

Tools & methods

JavaScript / Three.js / WebGL / WebGazer.js / Flask / Python / SciPy / JSON

See the prototype in action

48 seconds · AI-generated English narration

Minimap navigation, X-ray navigation and a fire-evacuation scenario. Master’s thesis at TUM, completed under academic supervision.

Read the transcript

00:00A browser-based prototype for three-dimensional navigation.

00:05The minimap shows the maze layout, current position, heading, and exit.

00:14The X-ray view makes walls transparent, revealing the corridors and the exit.

00:22This reveals the surrounding layout during first-person movement.

00:33During evacuation, the navigation aid is switched off.

00:38The user moves through the maze with fire and smoke present.

00:45Explore the full project.

First-person indoor environment with a minimap — Minimap during learning: layout, position, heading and exit context.

Indoor environment with transparent blue walls and a visible exit label — X-ray learning view reveals corridor geometry through walls.

The question

How well does a navigation aid help people learn an unfamiliar indoor layout and later find another exit route when the original route is blocked?

Interactive application

I implemented first-person controls, collision detection, a minimap and an X-ray interface, together with dynamic fire and smoke.

Data collection

WebGazer.js and a Flask backend connect webcam gaze with structured session data. Movement sampled at 4 Hz, minimap interactions and questionnaire responses support behavioural and gaze analysis in Python.

Study design

28 participants completed a counterbalanced within-subjects experiment; 23 were analysed after outlier removal. Each used both aids. The aid was on during learning and off during the subsequent fire-detour evacuation.

Findings and interpretation

Learning time and distance did not differ significantly. Evacuation trends favoured the minimap, but time and distance differences were not conclusive at the .05 level. 19 of 23 preferred the minimap for learning; 15 of 23 preferred it for evacuation confidence. The findings motivate further testing of combined overview and direct visual cues.

Build the interface.
Understand the user.

I bring software prototyping, data analysis and user research to technical and product-focused teams.

xuweichen09@outlook.com GitHub Request the full PDF portfolio

Full application materials available by email.