Home › Our People › Doctoral Candidates › DC7 – Aristotelis Tsoutsanis
DC7
Data fusion of traffic, behaviour & infrastructure for holistic driver assistance
Aristotelis Tsoutsanis
Safer mobility
for a brighter
tomorrow
for a brighter
tomorrow
♧ Work Package 5
AI for Road User Assistance
AI for Road User Assistance
Research at a glance
My PhD investigates road-safety-relevant driver behaviour with telematics. The core line of work detects harsh cornering, an underexplored lateral counterpart to well-studied braking/acceleration events, through orientation-invariant, map-validated signal processing, progressing from supervised (hEART2025) to unsupervised (TRB2025) and self-supervised (TRA2026) detection pipelines validated against OpenStreetMap intersections. Guided by my own systematic review of the field, the research has since extended beyond single-modality telematics into multimodal fusion: vision-language fault attribution for harsh events, transformer-based multimodal driver-state recognition, and 3D lane perception.
Research objectives
- Advance multimodal fusion methods: combine telematics, vision, and language signals into unified models for richer, more robust driver-behaviour understanding.
- Apply vision-language models for contextual interpretation: use visual context to causally interpret sensor-detected driving events, distinguishing driver-caused behaviour from externally-forced situations.
- Develop transferable, self-supervised representations: build representation-learning approaches, including transformer-based encoders, that generalise driver-behaviour understanding across datasets and sensing modalities.
Conference contributions
| Title | Conference | Date | Link |
|---|---|---|---|
| Classifying Vehicle Cornering Behavior using Mobile Sensor Data | 13th Symposium of the European Association for Research in Transportation (hEART2025) | 2025 | – |
| Identifying Dangerous Street Segments and Traffic Behavior in Athens using Telematics, Crash and Traffic data | 12th International Congress on Transportation Research (ICTR) | – | – |
| Unsupervised Detection of Harsh Cornering Behavior using Smartphone-based Telematics and Infrastructure Data | 105th TRB Annual Meeting | 2025 | – |
| Self-Supervised Detection of Harsh Cornering Events at Scale using Smartphone Sensor Data | Transportation Research Arena (TRA) | 2026 | – |
| Transformer-Based Driver Behavior Recognition Using the UAH-DriveSet Dataset | 8th IRTAD International Conference | – | – |
Secondments & collaborations
| Host organisation | Period | Purpose | Status |
|---|---|---|---|
| Haskoning | – | – | – |