Real-Time Cyber-Human-Vehicle Systems for Driving Safety Enhancement

In this poster, we briefly present the overview of this CPS project and our research progress in its first two years. The overall goals of this project are to develop onboard-adaptable and personalizable human driver models, create revolutionary driver-specific and personalized vehicle active motion control systems, design dynamic onboard real-time computation task scheduling methods, and integrate real-time V2V communications with driver-vehicle-pair-specific inter-vehicle motion control methods. We also aim to inspire interests of young people and attract them to pursue their higher education in engineering and science. We take a multidisciplinary research approach to yield synergistic innovations at the intersection of human factor, vehicle control, onboard real-time computation and communications for effectively utilizing the newly available vehicle cyber resources to maximize the likelihood of real-world driving accident avoidance. In the past two years, a vehicle controller parameter selection method considering both vehicle system physical constraints, motion control performance, and ECU computational load has been proposed. The method can help a vehicle track its reference path under both the stability and computational capacity constraints. Vehicle active safety control methods using V2V communications with dynamic channel selection mechanism were developed for both vehicle longitudinal control and lateral control scenarios. The methods were evaluated in both simulations using high-fidelity models and experiments using scaled vehicles and results show improved performance in comparison with those using existing V2V communication approaches. We have improved the design of MC-Safe, a multi-channel V2V communication framework with a channel model adaptation to dynamically adapt our model parameters based on the real measurements. An interference suppression was added to reduce the interference from non-safety messages. Those improvements have resulted in a 12.31% lower deadline miss ratio and an 8.21% higher packet delivery ratio. The distribution of driver attention to preview while tracking a winding roadway was measured by perturbing the display of the roadway with sinusoidal observation noise. A different sinusoidal frequency was used at each preview location, and the driver's control movements were analyzed to determine which of the perturbation frequencies were present. The signal-to-noise ratio of each perturbation frequency was the measure of attention to the corresponding preview position. Three studies investigated the sensitivity and structure of the attentional distribution to preview. We developed a new platform for testing the performance of human subjects during simplified driving tasks. The software is written in Java and is intended to be used in coordination with the MiniCave display and driving seat to create a more immersive and high-fidelity version of the experiment compared to previous methods. In addition, we also started research dissemination and educational outreach efforts and published four peer-reviewed papers so far.