ITSC 2025 Paper Abstract

Paper FR-LA-T44.1

Pennino, Federico (Alma Mater Studiorum - Università di Bologna), Sette, Davide (Ducati Motor Holding s.p.a.), Attisano, David (Ducati Motor Holding s.p.a.), Gabbrielli, Maurizio (University of Bologna)

Contextual Contrastive Learning for Rider Behavior Modeling Using Motorcycle Sensor Data

Scheduled for presentation during the Regular Session "S44c-Human Factors and Human Machine Interaction in Automated Driving" (FR-LA-T44), Friday, November 21, 2025, 16:00−16:20, Currumbin

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

Keywords Driver Behavior Monitoring and Feedback Systems for Semi-autonomous Vehicles, User-Centric HMI Design for Autonomous Vehicle Control Systems, Human-Machine Interaction Systems for Enhanced Driver Assistance and Safety

Abstract

Accurate modeling of rider behavior is essential for advancing motorcycle safety systems, developing personalized assistance interventions, and evaluating rider skill levels. Although motorcycle sensor data provides detailed insights into rider actions and environmental conditions, effectively automating the process of behavioral pattern extraction remains challenging due to inherent temporal dependencies, context variability, and the difficulty of analyzing large volumes of data. To address these challenges, we used a self-supervised approach inspired by SimCLR that leverages context-target augmentation strategies to motorcycle riding data. By splitting each riding data window into context and target segments, our method applies distinct augmentations to generate informative positive pairs. Using the InfoNCE loss, this approach learns robust embeddings that closely align contextually similar segments while effectively separating dissimilar patterns. Experimental validation on a real-world motorcycle riding dataset underscores the effectiveness of this contextual contrastive learning approach as a powerful technique for extracting meaningful rider behavior representations from motorcycle sensor data.