🏆 Best Work-in-Progress (WIP) Paper Award — IEEE PerCom 2026
The Publication
I am thrilled to share that the paper “Oh, Be Safe: A Field Study for Crowdsensed Cycling Safety Maps”, co-authored with Debasree Das, Leonie Ackermann, Katharina Ebner, Oleksandr Huba and Prof. Dr. Daniela Nicklas, has been published at the 2026 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops) in Pisa, Italy — and received the Best Work-in-Progress Paper Award. You can access the published work here: https://doi.org/10.1109/PerComWorkshops68308.2026.11585384
What the Paper Is About
Cyclists frequently face unsafe overtaking situations, yet many cities lack precise data to identify dangerous road sections. In this work, we ran a field study in Bamberg, Germany with ten volunteers covering 415 km, combining overtaking-distance sensing with physiological, video, and survey data to study how riders experience stress and post-ride anxiety. Using a customized OpenBikeSensor device, we verified data quality, enriched rides with contextual environmental factors, and identified spatial hotspots of “unsafe zones” that were validated against official accident records (gefahrenstellen.de). We then examined which multimodal features shape cyclists’ post-ride anxiety, applying PCA to reveal the dominant contributing factors.
My Contribution
From July to December 2025, I worked as a Data and Privacy Engineer (Wissenschaftlicher Mitarbeiter) at the Chair of Mobile Systems at the University of Bamberg on the BMBF-funded Explanym project. My primary responsibility was turning the raw crowdsensed telemetry into a clean, trustworthy, analysis-ready dataset. Concretely, I contributed:
- GPS trajectory reconstruction: Built a custom map-matching engine on the open-source Valhalla routing project. Using a Hidden Markov Model over OpenStreetMap road networks, I aligned noisy consumer GPS points to their most probable road segments — recovering 82.3% of the 109,000+ raw trajectory points and removing outliers via water, speed, and gpx filters.
- Multimodal alignment: Used Dynamic Time Warping (DTW) to temporally and spatially align the unsynchronized OpenBikeSensor and smartwatch (Polar Ignite 2) data streams.
- Feature engineering: Computed physiological features (heart-rate spikes around overtaking events), traffic-volume inference via YOLO-v8 object detection on the video footage, and weather-based contextual variables.
- Analysis pipeline: Applied PCA and t-SNE for dimensionality reduction and studied probabilistic dependencies, contributing to the identification of the features that best explain post-ride anxiety.
This work highlights a persistent engineering trade-off I care deeply about: collecting high-resolution mobility data to improve urban planning without compromising individual privacy. Throughout the project I applied these principles directly inside the data pipelines — such as differential privacy and spatial cloaking — so that safer infrastructure planning never has to come at the cost of data protection.
Key Insights
We found that low-cost OpenBikeSensor units reliably measure overtaking distances, and that riders’ physiological responses vary with overtaking events, cycling experience, and anxiety levels. Critically, both unsafe and seemingly “safe” overtaking zones (wider than 1.5m) still triggered measurable stress spikes — suggesting that current safety metrics may under-estimate perceived risk. Our validated hotspots demonstrate a scalable, low-cost approach for auto-annotating urban maps with overtaking-prone zones, which can inform cycling route recommendations and safer micromobility infrastructure planning.
