Driver cognitive distraction is a critical factor in road safety, and its evaluation, especially under real conditions, presents challenges to researchers and engineers. In this study, we considered mental workload from a secondary task as a potential source of cognitive distraction and aimed to estimate the increased cognitive load on the driver with a four-channel near-infrared spectroscopy (NIRS) device by introducing a machine-learning method for hemodynamic data. To produce added cognitive workload in a driver beyond just driving, two levels of an auditory presentation n-back task were used. A total of 60 experimental data sets from the NIRS device during two driving tasks were obtained and analyzed by machine-learning algorithms. We used two techniques to prevent overfitting of the classification models: (1) k-fold cross-validation and principal-component analysis, and (2) retaining 25% of the data (testing data) for testing of the model after classification. Six types of classifier were trained and tested: decision tree, discriminant analysis, logistic regression, the support vector machine, the nearest neighbor classifier, and the ensemble classifier. Cognitive workload levels were well classified from the NIRS data in the cases of subject-dependent classification (the accuracy of classification increased from 81.30 to 95.40%, and the accuracy of prediction of the testing data was 82.18 to 96.08%), subject 26 independent classification (the accuracy of classification increased from 84.90 to 89.50%, and the accuracy of prediction of the testing data increased from 84.08 to 89.91%), and channel-independent classification (classification 82.90%, prediction 82.74%). NIRS data in conjunction with an artificial intelligence method can therefore be used to classify mental workload as a source of potential cognitive distraction in real time under naturalistic conditions; this information may be utilized in driver assistance systems to prevent road accidents.

The main purpose of this study is to investigate the effect of visual stimulus on voluntary eye movement during driving. In this paper, we examine two cases: driving with/without visual stimulus by means of a driving simulator. We develop a model that consists of both vestibulo-ocular reflex (VOR) and optokinetic reflex (OKR) components. By comparing the observed eye movement and the simulation result of the model, the results of eye movement simulation was more accurate than VOR model only even in a naturalistic situation with optic flow of the visual scene and the driver’s voluntary eye movement.

We have developed a method to estimate driver distraction by mean of involuntary eye movement of the vestibulo-ocular reflex (VOR) and optokinetic response (OKR) by using a driving simulator. This paper presents a practical application of the method to the real vehicle situation. The main concern was whether we can measure the eye movement of VOR and OKR caused by the road surface vibration in the situation. The experiments were conducted with participants sitting on the passenger seat for safety reasons. The results have shown that our method can be applied for evaluating cognitive distraction not only in a driving simulator but also in actual vehicles.

Fast and accurate three-dimensional (3D) Multiple Object Detection and Tracking (3DMODT) is a critical task for autonomous vehicles to perceive their surroundings and make safe decisions. However, many existing works focus on academic research to address individual problems rather than overall real-world applications, which are more generalizable and applicable to daily life. To address this gap, this paper presents a lightweight and efficient 3D object understanding framework for real-life autonomous vehicles from scratch. We introduce PHE3D, a new large-scale, complex Light Detection and Ranging (LIDAR) dataset that captures Vietnamese streets, notable for its distinct characteristics and wide variety of object classes. Additionally, this study discusses the entire process for 3D object understanding, from data collection to data handling to 3DMODT. To address the computational demands, we propose a suitable lightweight and informative Convolutional Neural Network (CNN) architecture for 3D Multi-Object Detection (3DMOD) and a mathematical-based approach for 3D Multi-Object Tracking (3DMOT). Extensive experiments on the dataset show that the system can precisely detect and track attributes of multiple objects while running at up to 100, and 2000 Frames Per Second (FPS), respectively.