Personalized Driver Workload Inference by Learning From Vehicle Related Measurements

Yi, Dewei; Su, Jinya; Liu, Cunjia; Chen, Wen-Hua · 2019 · Crossref

DOI: 10.1109/tsmc.2017.2764263

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of inferring driver workload to enable adaptive in-vehicle systems, such as advanced driver assistance systems (ADAS) and in-vehicle information systems (IVIS). While these technologies enhance safety and convenience, they can also increase driver distraction and workload, potentially leading to accidents. Existing driver workload inference (DWI) systems often rely on physiological measurements or computer vision, which face issues regarding reliability, privacy, and environmental dependencies. To overcome these limitations, the authors propose a Personalized Driver Workload Inference (PDWI) system that utilizes nonintrusive, easily accessible vehicle-related measurements (VRMs), specifically vehicle speed and three-axis accelerations. The primary motivation is to create a model that adapts to individual driving characteristics rather than applying a generic average model, thereby allowing for timely and appropriate adaptive aiding. The proposed PDWI framework operates in two stages: offline training and online implementation. During offline training, the system first employs unsupervised clustering algorithms to automatically categorize individual drivers' workload levels based on the inherent characteristics of their VRM data. The authors compare centroid-based Fuzzy C-means (FCM) clustering with density-based DBSCAN clustering, selecting FCM for its ability to handle overlapped data. The optimal number of workload clusters is determined using the silhouette criterion. Subsequently, a supervised classification algorithm establishes an implicit mapping between the VRM features and the identified workload clusters. The study compares Gaussian Mixture Model (GMM) and Support Vector Machine (SVM) classifiers for this mapping. In the online phase, new VRM samples are classified in real-time to infer the current workload category. The system was validated using a dataset from real-world naturalistic driving experiments conducted by the Human-Computer-Interaction Laboratory at the University of Stuttgart, involving ten participants with diverse backgrounds. Comparative experiments demonstrated that the integration of FCM clustering and SVM classification yielded superior performance in terms of accuracy, precision, recall, F1-score, and prediction time compared to other algorithmic combinations. The framework successfully identified interindividual differences in workload thresholds and categories, accommodating these variations automatically without manual parameter tuning. The significance of this work lies in its demonstration that personalized, nonintrusive workload inference is feasible using standard vehicle sensors. By leveraging machine learning to adapt to individual drivers, the PDWI system provides a robust foundation for next-generation adaptive aiding systems. This approach ensures that safety interventions are tailored to the specific cognitive state of the driver, enhancing both safety and user acceptance without the privacy and reliability concerns associated with physiological or vision-based monitoring.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-nvidia on 2026-08-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• drowsiness detection algorithms
• distraction detection algorithms
• workload measurement
• exposure measurement
• telematics crash prediction

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: physiological data
• Theoretical Contribution: computational model, theory or model