With the rapid development of intelligent vehicles and some successes in Advanced Driving Assistance Systems (ADAS), a new trend is that mixed levels of human driver engagements will be involved in the transportation system. Therefore, necessary visual guidance for drivers is vitally important under this situation to prevent potential risks. To advance the development of visual guidance systems, we introduce a novel vision-cloud data fusion methodology, integrating camera image and Digital Twin information from the cloud to help intelligent vehicles make better decisions. Target vehicle bounding box is drawn and matched by combining results of object detector running on ego vehicle and position information from the cloud. The best matching result, a 79.2% accuracy under 0.7 intersection over union threshold, is obtained with depth images served as an additional feature source. A case study on lane change prediction is conducted to show the effectiveness of the proposed data fusion methodology, where a multi-layer perceptron algorithm is proposed with modified approaches. Human-in-the-loop simulation results conducted in Unity game engine reveal that the proposed model can improve highway driving performance significantly in terms of safety, comfort, and environmental sustainability.
31 August 2021 in IEEE Open Journal of Intelligent Transportation Systems
Intelligent vehicles and Advanced Driver Assistance Systems (ADAS) are being developed rapidly over the past few years. Many applications such as vehicle localization, environment perception, and path planning have shown promising potentialities. While there is great interest in migrating from complete human-controlled vehicles towards fully autonomous vehicles, it is natural that researchers spending more effort trying to understand the interaction between vehicles with various levels of automation in large-scale traffic scenarios. Next-generation vehicles are expected to have the capacity of evaluating driver conditions, vehicle capabilities, surrounding traffic contexts, and take advantage of such knowledge to ensure safe and efficient driving. Three general research questions are raised to achieve this goal, which are (i) how can we acquire sufficient data, (ii) how to evaluate and understand driving behavior, and (iii) how to deliver information effectively to drivers. In this article, we present a review of previous studies from the UTDrive project attempts to answer above questions.
Federated learning (FL) is widely recognized as a valuable approach for Connected and Automated Vehicles (CAVs) because it facilitates collaborative model development across a multitude of vehicles in a decentralized manner. However, numerous studies on FL algorithms only assessed their performance through experiments conducted in simulated client-server configurations (e.g., where both server and clients run on the same machine) or simplified scenarios that do not account for client downtime. In this paper, we aim to conduct more realistic evaluations for CAV applications leveraging FL. We present a preliminary experimental study as well as offer insights into potential future directions.