Multisensor Bicycle Overtaking Raw Dataset for Predictive Criticality Assessment

Abstract

This dataset contains 71 multimodal field recordings collected with a mobile bicycle-mounted sensing platform to support research on the early assessment of potentially critical motor-vehicle overtaking manoeuvres. Each recording includes synchronized data from a rear-facing TI AWR1843BOOST 77 GHz mmWave radar, TF-Luna time-of-flight and HC-SR04 ultrasonic lateral-distance sensors, a camera, GPS, and an inertial measurement unit, together with diagnostic logs. Seventeen sessions additionally contain live machine-learning inference and warning logs. The data support causal vehicle tracking, trajectory reconstruction, overtaking-event detection, reference-assisted labeling, full-ride replay, and the evaluation of edge-deployable warning models. The final thesis evaluation used a manually reviewed event dataset derived from 53 rides, comprising 491 usable events. Two rides contain header-only IMU files and are handled using explicit missing-data indicators. Supplementary focus-group documentation is included only as a presentation-pdf and statistical questionnaire.