Spatial Deep Learning Approach to Older Driver Classification

Area of Science:

• Artificial Intelligence
• Transportation Safety
• Gerontology

Background:

• Older driver classification is crucial for road safety, insurance, and interventions for cognitive decline.
• Telematics data presents challenges due to volume and heterogeneity.
• Existing methods struggle with complex, temporally-detailed vehicle datasets.

Purpose of the Study:

• To develop a novel spatial deep-learning approach for accurate older driver classification.
• To enhance the detection of abnormal driving behaviors using Grid-Index based data augmentation.
• To address the challenges posed by large and heterogeneous telematics datasets.

Main Methods:

• Proposed a novel spatial deep-learning model.
• Implemented Grid-Index based data augmentation techniques.
• Conducted extensive experiments and a real-world case study.

Main Results:

• The proposed approach achieved high accuracy in identifying abnormal drivers.
• Grid-based methods significantly improved telematics-based driving behavior analysis.
• Demonstrated consistent performance across various tests.

Conclusions:

• The spatial deep-learning approach effectively classifies older drivers.
• Grid-index methods offer a promising way to enhance driving behavior analysis.
• Findings support improved road safety, insurance assessment, and targeted interventions.