Min Duan

Mrs. Min Duan


Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
23 July 2026 in Vehicles
Verified
Journal Article
Open access
PDF file
MDMin Duan
Min Duan
LXLian Xie
Lian Xie
CSChuan Sun
Chuan Sun
JYJunru Yang
Junru Yang
SXShucai cai Xu
Shucai cai Xu
HSHaiming Sun
Haiming Sun
Autonomous driving systems relieve drivers from continuous vehicle operation and constant monitoring, allowing them to engage in non-driving-related tasks (NDRTs). However, immersion in such tasks can impair drivers’ perception of both the takeover situation and the surrounding environment. To quantitatively assess drivers’ risk perception capability during takeover, a driving simulation platform was used to design autonomous takeover scenarios involving three types of NDRTs, three takeover request times (TOR), and two obstacle avoidance conditions. A total of forty participants were recruited to complete the driving experiment. Drivers’ eye movement data were collected, and visual metrics—including fixation, saccade, and pupil diameter—were extracted by defining areas of interest (AOIs). A subjective risk perception scale was developed and administered to measure drivers’ subjective evaluations. Together with takeover reaction time, the K-means clustering method was applied to classify drivers’ risk perception levels into three categories: high, medium, and low. The LightGBM algorithm was selected to construct a baseline classification model for assessing drivers’ risk perception levels. Subsequently, the Whale Optimization Algorithm (WOA) was employed to optimize the hyperparameters of LightGBM, resulting in the WOA-LightGBM model. This optimized model demonstrated improved recall, accuracy, precision, and F1-score, reaching 0.9210, 0.9253, 0.9261, and 0.9201, respectively. Furthermore, SHapley Additive exPlanations (SHAP) analysis was conducted to quantify the contribution of eye movement indicators to risk perception assessment. The results revealed that saccade duration in the NDRT areas significantly reduced drivers’ risk perception levels (SHAP value = −0.71), whereas increased saccade duration in the forward road area effectively restored drivers’ risk perception capability (SHAP value = 0.71). In addition, higher risk perception levels were found to enhance drivers’ takeover performance in terms of vehicle control. These findings provide valuable insights for the management of NDRTs and the optimization of autonomous vehicle takeover systems.

7 November 2025 in Machines
Verified
Journal Article
Open access
PDF file
MDMin Duan
Min Duan
LXLian Xie
Lian Xie
JCJianrong Cai
Jianrong Cai
JYJunru Yang
Junru Yang
HLHaoran Li
Haoran Li
Current autonomous vehicles require human drivers to take over control during emergencies or in environments the system cannot handle. During other periods, drivers are permitted to engage in non-driving-related tasks. It is essential to investigate how the immersion in non-driving-related tasks affects drivers’ takeover performance under different scenarios. To address this, a mixed-design simulated driving experiment was conducted with 40 participants, incorporating three non-driving-related tasks (no task, watch video, play game), three takeover request lead times (3 s, 5 s, 7 s), and two obstacle types (dynamic, static). The takeover process was divided into three phases: preparation, obstacle avoidance, and recovery. Analysis of the areas of interest showed that engaging in non-driving-related tasks substantially reduced drivers’ visual attention tothe road ahead during the preparation phase. The Generalized Estimating Equations method was employed to investigate the effects of various factors on takeover performance. Model results showed that scenarios with static obstacles and longer takeover request times led to a significant reduction in mean lane deviation but a significant increase in the standard deviation of lane deviation, suggesting improved lateral control performance. A significant interaction was observed between the watch video task and static obstacles, which corresponded to a notable decrease in the mean vehicle speed during obstacle avoidance. Performance in the recovery phase was strongly predicted by that in the obstacle avoidance phase, indicating that the stability of the avoidance maneuver is a critical determinant of the subsequent recovery. These findings offer valuable insights for managing non-driving-related tasks and setting appropriate takeover request timings in automated driving systems.

3 October 2026 in Vehicles
Verified
Journal Article
Open access
PDF file
MDMin Duan
Min Duan
JSJin Shu
Jin Shu
LXLian Xie
Lian Xie
JCJianrong Cai
Jianrong Cai
CSChuan Sun
Chuan Sun
HSHao Shi
Hao Shi
To address the challenges of recognition lag and limited representational capacity of single-modal physiological information in driver intention recognition under human–machine co-driving environments, this study constructed a highway driving takeover experimental scenario. Typical takeover tasks including obstacle avoidance takeover, adjacent vehicle cut-in takeover, and curved section takeover were designed, with simultaneous collection of drivers’ electroencephalography (EEG) and eye-tracking data. On this basis, a multi-modal physiological information fusion method for driving intention recognition was proposed. First, areas of interest (AOIs) were defined to extract visual behavioral features, including fixation duration, fixation count, and pupil diameter, while eye-tracking recurrence plots were generated based on phase space reconstruction. Meanwhile, event-related potential (ERP) analysis was performed on the EEG data to extract N100, N200, and P300 components, and continuous wavelet transform was applied to construct EEG time–frequency maps. Subsequently, the EEG time–frequency maps and eye-tracking recurrence plots served as the dual inputs to the dual-branch Squeeze-and-Excitation Convolutional Neural Network (SE-CNN) model, which was further enhanced by fusing ERP features and eye-tracking statistical features, ultimately achieving accurate recognition of four types of driving intentions. The model results demonstrated that the dual-branch SE-CNN model achieved a recognition accuracy of 88.29% on the four-class intention recognition task, with precision, recall, and F1-score reaching 88.78%, 88.26%, and 88.31%, respectively—significantly outperforming both single-modal models and the dual-branch Convolutional Neural Network (CNN) model without the SE attention mechanism. The proposed model can effectively integrate EEG neural activity signals and eye-tracking visual behavioral information, thereby substantially enhancing driving intention recognition performance. This provides theoretical foundations and technical support for applications such as early driver intention perception, intelligent takeover decision-making, and human–machine collaborative control in human–machine co-driving environments.