Related Experiment Video
Updated: Feb 28, 2026
Tactile Vibrating Toolkit and Driving Simulation Platform for Driving-Related Research
Published on: December 18, 2020
The interactive effects of NDRT cognitive complexity and traffic density on takeover performance in level 3 automated
Qi Guo1, Mengting Tang1, Yitao Yang2
1Visual Cognition Laboratory, The School of Media and Design, East China University of Science and Technology, Shanghai, China.
Objective:
In Level 3 automated driving, driver takeover performance is influenced by cognitive state and external traffic conditions. This study investigates the interactive effects of Non-Driving Related Task (NDRT) cognitive complexity and traffic density on takeover performance.
Methods:
A 2 (NDRT type: low-complexity perception task vs. high-complexity decision task) × 2 (traffic density: low 10 veh/km vs. high 30 veh/km) within-subjects design was employed, with electroencephalography (EEG) and behavioral metrics analyzed.
Results:
(1) NDRT cognitive complexity significantly influenced EEG spectral power, with the decision task leading to lower theta and alpha power, suggesting a higher cognitive load; (2) While neither factor alone had a significant main effect on takeover performance, their interaction had a significant impact on takeover reaction time (TORT). The shortest TORT was observed under the perception task with high traffic density. (3) Increased cognitive load during decision tasks prolonged the time taken to regain control (TORT) and reduced safety (measured by time to collision in minutes, TTCmin), but it improved stability (measured by mechanical response acceleration, MRA). This combination ultimately enhanced the Takeover Performance Index (TOPI). (4) Takeover performance metrics were closely interrelated. TORT effectively predicted TTCmin (R2 = 0.841), and TORT, MRA, and TTCmin collectively predicted TOPI (R2 = 0.899).
Conclusions:
The interaction between NDRT cognitive complexity and traffic density is a critical factor influencing driver takeover performance in Level 3 automated driving. The findings provide theoretical support and a modeling basis for the design of cognitive state-aware intelligent takeover systems.
