Analysis of adaptive systems based on Driver's workload

Abstract

This study examined workload classification models and their application in adaptive in-vehicle systems. A meta-analysis of 31 studies assessed how predictor types (e.g., physiological data), experimental settings (simulator vs. on-road), and device types (wearable vs. remote) influence model accuracy. Results indicated that incorporating physiological data improved model accuracy, although ensuring generalizability remains a challenge. Random Forest models demonstrated the highest average accuracy for binary classification, while Neural Networks showed promise for multi-class models. Adaptive systems leveraging multi-input models were found effective in dynamically adjusting to workload, enhancing safety and user experience. However, challenges such as system over-reliance and limited system implementation persist. Additionally, this study analyzed the existing adaptive systems in the automotive market and proposed design guidelines and a framework for workload-based adaptive systems. Future research should focus on developing robust, context-aware systems tailored to occupational and real-world driving demands, ensuring reliability and widespread applicability.

Keywords: Adaptive systems; Cognitive load; Driver; Mental workload.