Driver Anomaly Quantification for Intelligent Vehicles: A Contrastive Learning Approach With Representation Clustering

Abstract

Driver anomaly quantification is a fundamental capability to support human-centric driving systems of intelligent vehicles. Existing studies usually treat it as a classification task and obtain discrete levels for abnormalities. Meanwhile, the existing data-driven approaches depend on the quality of dataset and provide limited recognition capability for unknown activities.To overcome these challenges, this paper proposes a contrastive learning approach with the aim of building a model that can quantify driver anomalies with a continuous variable. In addition, a novel clustering supervised contrastive loss is proposed to optimize the distribution of the extracted representation vectors to improve the model performance. Compared with the typical contrastive loss, the proposed loss can better cluster normal representations while separating abnormal ones. The abnormality of driver activity can be quantified by calculating the distance to a set of representations of normal activities rather than being produced as the direct output of the model. The experiment results with datasets under different modes demonstrate that the proposed approach is more accurate and robust than existing ones in terms of recognition and quantification of unknown abnormal activities.

Human Resource Demand Prediction and Configuration Model Based on Grey Wolf Optimization and Recurrent Neural Network

“…Our research shows that the suggested methodology takes less time overall to complete than existing methods of allocating resources. The researchers hypothesized, using a numerous constituency perspective of the HR function, that organizational financial investment in their HR functions will have an impact on labor productivity and that this relationship will be moderated by the presence of professional HR staff and the adoption of high performance work systems [ 22 , 23 ]. Selection, training, working conditions, and assessment were included as independent variables in this study's analysis of HR planning while job satisfaction as a proxy for organizational performance was used as the dependent variable.…”

Review and Perspectives on Driver Digital Twin and Its Enabling Technologies for Intelligent Vehicles

“…A variety of nonintrusive and intrusive sensors are utilized to detect the driver's distractions at different levels of granularity by leveraging ML algorithms [88,89]. The nonintrusive sensors, including the vision-based and vehicle-related sensors, can be used to tackle high-level behavior-related distractions such as abnormal activity [90][91][92][93], distracted pose [94][95][96][97], false operation [98,99], and inappropriate focus area [100][101][102][103]. In contrast, intrusive sensors can precisely assist in evaluating the driver's inner consciousness and cognitive state [53,[104][105][106].…”

MFA: A Smart Glove with Multimodal Intent Sensing Capability

“…Liu et al [ 12 ] proposed a deep learning-based multimodal fusion model that combines three modal data sets: voice commands, hand gestures, and body movements, using various deep neural networks. In the field of automatic driving, Hu et al [ 13 ] introduced contrastive learning approach to train a feature extractor with good representation ability in order to improve driving performance and avoid possible fatal accidents. In the field of speech recognition, Ondas et al [ 14 ] proposed a combination of modified LIMA framework and iterative spectral subtraction algorithm to improve the robustness of speech recognition in noisy environment.…”