Jiang, Sida (WSP Sweden), Wu, Jiaming (Chalmers University of Technology), Diaconu, Bogdan (FellowBot)

Identification of Drunk Driving Using Deep Learning Networks on Virtual Reality Simulation

Scheduled for presentation during the Regular Session "S32a-AI-Driven Traffic Monitoring, Safety, and Anomaly Detection" (FR-LM-T32), Friday, November 21, 2025, 11:50−12:10, Southport 2

2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), November 18-21, 2025, Gold Coast, Australia

This information is tentative and subject to change. Compiled on October 18, 2025

Abstract

The paper addresses safety concerns associated with electric micromobility vehicles, particularly e-scooters, which accounted for 1% of trips but nearly 4% of traffic accidents in Sweden, 2019. To mitigate potential risks from increased micromobility usage, we developed robust AI algorithms to detect drunk driving. Based upon our own developed Virtual Reality simulations, involving 30 participants, we utilized hand-tracking technology to enhance realism and reduce nausea. Results indicated that LSTM (Long Short-Term Memory network) model successfully identified 86% of drunk riders. A Transformer method was tested with further aggregated data where we have achieved 100% precise identification of drunk drivers. LSTM methods that we tested in the project have shown significant improvement for detecting drunk drivers, compared with machine learning methods e.g. Random Forest and Recurrent Neural Networks (RNNs). This research contributes to Vision Zero, zero fatalities and severe injuries from road traffic in the Nordics by 2050, by providing robust sensing algorithm for identifying e-scooter misuse, with future work focusing on refining models, expanding data collection, and integrating advanced neural networks. Our findings offer valuable insights for stakeholders like shared e-scooter operators and urban transport administrations to enhance urban mobility safety.