Personalized Driver Workload Inference by Learning From Vehicle Related Measurements

Abstract

Adapting in-vehicle systems (e.g., advanced driver assistance systems and in-vehicle information systems) to individual drivers' workload can enhance both safety and convenience. To make this possible, it is a prerequisite to infer driver workload so that adaptive aiding can be provided to the driver at the right time and in an appropriate manner. Rather than developing an average model for all drivers, a personalized driver workload inference (PDWI) system considering individual drivers driving characteristics is developed using machine learning techniques via easily accessed vehicle related measurements (VRMs). The proposed PDWI system comprises two stages. In offline training, individual drivers workload is first automatically splitted into different categories according to its inherent data characteristics using fuzzy C-means (FCM) clustering. Then an implicit mapping between VRMs and different levels of workload is constructed via classification algorithms. In online implementation, VRMs samples are classified into different clusters, consequently driver workload type can be successfully inferred. A recently collected dataset from real-world naturalistic driving experiments is drawn to validate the proposed PDWI system. Comparative experimental results indicate that the proposed framework integrating FCM clustering and support vector machine classifier provides a promising workload recognition performance in terms of accuracy, precision, recall, F 1 -score, and prediction time. The interindividual differences in term of workload are also identified and can be accommodated by the proposed framework due to its adaptiveness.Index Terms-Fuzzy C-means (FCM) clustering, personalized aiding, support vector machine (SVM), workload recognition.

Implicit Personalization in Driving Assistance: State-of-the-Art and Open Issues

“…According to [3,22,26,[30][31][32][33][34], driving assistance systems should be safe, effective, and comfortable. To meet these criteria, personalization is introduced to understand the status of a specific driver [35], and take individual driving styles [29], requirements, and preferences [36] into account.…”

“…Moreover, the privacy issue involved in vision also needs to be addressed. Compared to vision measurements, vehicle dynamic measurements are more robust against light condition [3]. Vehicle dynamic measurements include steering angle, lateral acceleration, longitudinal acceleration, vehicle velocity amongst others.…”

“…To avoid this unnecessary fuel consumption, a scenario tree based stochastic model is introduced to adapt to a specific driver so that vehicle acceleration and braking can be reduced [114]. -Not easy to select an appropriate kernel; [109] Real-world -Labelled images Neural Network -Flexible non-linear capability; data-driven method; [48,49] -Hard to design layers and neurons; large volume of iterations to converge; [49,50] -Cost of thermal camera is higher than an infrared camera or a RGB camera; [3] Real-world -Speed, three dimensional accelerations In [32,43], probability weighted autoregressive exogenous models are used to learn individual driving behaviors for a specific driver so that fuel consumption can be estimated more precisely. Driving style and state are also important in driving safety and vehicle security.…”

“…Safety, efficiency, and convenience are three key concerns raised in recent studies on intelligent vehicles [1][2][3][4][5][6][7][8]. According to a World Health Organization report, up to 50 million people are injured or disabled in road accidents worldwide every year with 90% of deaths occurred in developing nations [9].…”

“…It is noticeable that the generic approach trains or designs a model by using the driving data of all drivers indiscriminately, and, as a result, personalized driving characteristics and preferences of individual drivers may be neglected [27]. In practice, different drivers may have distinct driving characteristics and preferences even in a similar driving scenario [3]. Therefore, it is not surprising that a conventional generic approach may provide limited performance and satisfaction for individual drivers.…”