Author: Evyatar Luvaton

Publisher:

Published: 2021

Total Pages:

ISBN-13:

Adaptive cruise control (ACC) and cooperative adaptive cruise control (CACC) are two kinds of vehicle automation and communication systems (VACS) that modify the traffic flow. These cruise systems assist equipped vehicles to follow other vehicles safely and comfortably. Both systems must satisfy the string stability specification, i.e., guaranteeing that errors (acceleration and velocity errors) not amplify from one vehicle to another. There are various control algorithms that handle string stability in case all vehicles are equipped with these types of systems. Nevertheless, until all vehicles on the road will be equipped with CACC systems, it is reasonable to assume that our roads will be shared by mixture of vehicles with CACC systems and manual vehicles. Orki and Arogeti (2019) offered a new approach of controlling the abovementioned mixed traffic environment, by looking at the mixed traffic as single entity, termed, mixed platoon. This mixed platoon is controlled by the autonomous vehicles embedded in the platoon. The authors' goal was to achieve a string stable mixed platoon by eliminating the disturbances caused by manual vehicles. While the previous study was focused on the control algorithm (i.e., on the side of the automation) the purpose of this study was to mitigate the disturbances caused by the human drivers by helping them in temporally anticipating the acceleration of the lead vehicle such that the overall string stability of the platoon is improved. Using a novel car-following model, we hypothesized that the best way to improve the manual driver's performance, and hopefully the string stability, is by supporting the manual driver's temporal anticipation capabilities, which mainly consist of estimating the lead vehicle acceleration. For this propose, a car-following assisting 5system named the rear window notification display (RWND) was developed. The RWND presents the acceleration of the autonomous lead vehicle on its rear window. A simulator-based experiment with 10 participants was designed to examine the influence of RWND on individual driver performance and string stability. A velocity profile of the autonomous lead vehicle included three different velocities and two different disturbance frequencies which reflected in the lead vehicle's velocity profile. The participants were required to follow the lead vehicle with and without RWND. The results have shown that participants tended to maintain a longer distance from the lead vehicle when driving with the RWND than without the RWND. The measurement of string stability was improved when driving with the RWND at higher velocities. There was no improvement in string stability when driving without the RWND. A post experiment questionnaire revealed that the display did not increase the perceived workload of the participant during the car-following task.The results of this study suggest that the RWND can be used to aid the manual driver in temporally anticipating the acceleration of the lead vehicle and by that contributing to the overall string stability of the mixed platoon instead of the driver being consider as a disturbance. Afterwards, a modified car-following model was developed and evaluated for further research. The model was aimed for improvement of the autonomous vehicles' controllers while driving beyond human drivers using RWND. The model was based on two LSTM models which were trained on two different datasets - real driving data and simulator' data. Ensemble of these two models was the modified car-following model.