SAFE-D: A Spatiotemporal Detection Framework for Abnormal Driving Among Parkinson's Disease-like Drivers

TL;DR

This work addresses the challenge of detecting abnormal driving driven by Parkinson's disease (PD) motor symptoms in dynamic driving contexts. It introduces SAFE-D, a spatiotemporal attention-based framework that fuses in-vehicle sensor data from three control channels through global and local feature paths, culminating in a binary PD-disease anomaly detector. The approach maps PD motor deficits to steering, acceleration, and braking patterns, achieving 96.8% average accuracy across multiple driving maps on a Logitech G29–CARLA setup, with strong ablation results showing benefits from multi-channel inputs and attention modules. The findings suggest SAFE-D's potential for continuous PD monitoring and driving safety, while acknowledging privacy and real-world data challenges and outlining pathways for incorporating additional pathological indicators and disease progression analysis.

Abstract

A driver's health state serves as a determinant factor in driving behavioral regulation. Subtle deviations from normalcy can lead to operational anomalies, posing risks to public transportation safety. While prior efforts have developed detection mechanisms for functionally-driven temporary anomalies such as drowsiness and distraction, limited research has addressed pathologically-triggered deviations, especially those stemming from chronic medical conditions. To bridge this gap, we investigate the driving behavior of Parkinson's disease patients and propose SAFE-D, a novel framework for detecting Parkinson-related behavioral anomalies to enhance driving safety. Our methodology starts by performing analysis of Parkinson's disease symptomatology, focusing on primary motor impairments, and establishes causal links to degraded driving performance. To represent the subclinical behavioral variations of early-stage Parkinson's disease, our framework integrates data from multiple vehicle control components to build a behavioral profile. We then design an attention-based network that adaptively prioritizes spatiotemporal features, enabling robust anomaly detection under physiological variability. Finally, we validate SAFE-D on the Logitech G29 platform and CARLA simulator, using data from three road maps to emulate real-world driving. Our results show SAFE-D achieves 96.8% average accuracy in distinguishing normal and Parkinson-affected driving patterns.

Figures (14)

• Figure 1: Application scenario of our detection framework: leveraging sensing data from in-vehicle control components to identify abnormal driving behaviors and develop a detection model.
• Figure 2: Illustration of driving behavior difference between health and diseased drivers in (a) straight-line and (b) non-straight-line driving cases.
• Figure 3: Design of SAFE-D, including three main modules: data preprocessing, information fusion, and detection model.
• Figure 4: Attention-based local and global spatiotemporal feature fusion network for driving behavior detection.
• Figure 5: Architecture of ResNet blocks.