Measuring drivers’ takeover performance in varying levels of automation: Considering the influence of cognitive secondary task

• Published in: Transportation research. Part F, Traffic psychology and behaviour Vol. 82; pp. 96 - 110
• Main Authors: Lu, Guangquan, Zhai, Junda, Li, Penghui, Chen, Facheng, Liang, Liming
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier Ltd 01.10.2021; Elsevier Science Ltd
• Subjects: Automation; Automation levels; Automobile drivers; Automobile driving; Cognitive ability; Cognitive tasks; Driving conditions; Factorial design; Lane changing; Reaction time; Simulated driving; Subjective workload; Takeover quality; Taskload; Traffic accidents & safety; Vehicle safety; Workload; Reaction time; Subjective workload; Automation levels; Takeover quality; Simulated driving
• ISSN: 1369-8478; 1873-5517
• DOI: 10.1016/j.trf.2021.08.005

•The effects of cognitive load (CL) and automation level (AL) on drivers’ takeover quality (TQ) was investigated by a driving simulator study.•Driver’s subjective workload reduced with increased AL.•Higher AL lead to a less high-quality takeover.•CL improves drivers’ ability to follow instructions and TQ in critical highly automated driving condition. Mixed control by driver and automated system will remain in use for decades until fully automated driving is perfected. Thus, drivers must be able to accurately regain control of vehicles in a timely manner when the automated system sends a takeover request (TOR) at its limitation. Therefore, determining the factors that affect drivers’ takeover quality at varying levels of automated driving is important. Previous studies have shown that visually distracting secondary tasks impair drivers’ takeover performance and increase the subjective workload. However, the influence of purely cognitive distracting secondary tasks on drivers’ takeover performance and how this influence varies at different levels of automation are still unknown. Hence, a 5 (driving modes) × 3 (cognitive secondary tasks) factorial design with the within-subject factors was adopted for this driving simulator experiment. The sample consisted of 21 participants. The participants’ subjective workloads were recorded by the NASA-Task Load Index (NASA-TLX). Results showed that compared to manual driving conditions, the drivers’ subjective workloads were significantly reduced in both partially and highly automated driving conditions, even with a TOR, confirming the benefit of the automated driving system in terms of reducing the driving workload. Moreover, the drivers exhibited a lower takeover behavior quality at high levels of automation than manual driving in terms of increased reaction time, abnormal performance, standard deviation of lane position, lane departure probability, and reduced minimum of time to collision. However, at the highly automated driving condition, the drivers’ longitudinal driving safety and ability to follow instructions improved when performing a highly cognitive secondary task. This phenomenon possibly occurred because automated driving conditions lead to an underload phenomenon, and the execution of highly cognitive tasks transfers drivers into moderate load, which helps with the drivers’ takeover performance.