Addressing Adaptive Cruise Control Limitations via Human-in-the-loop Connected Cruise Control

Abstract: Modern vehicles are equipped with adaptive cruise control (ACC) that allows longitudinal control of the ego vehicle based on the speed or distance measured to its preceding vehicle. One of the limitations is that the ego vehicle takes off when the preceding vehicle changes its lane, and the ACC requires 2+ seconds of time headway to be string stable, which often induces cut-ins by nearby lane vehicles. Due to these, many drivers do not use the ACC daily.
In addition, when human drivers follow their preceding vehicles, their acceleration and deceleration are not optimal, mainly due to human perception and reaction. They typically do too much or too little acceleration or deceleration. Given more than 50% of new vehicles have connectivity (e.g., GM’s OnStar), the ego vehicle will likely run into a connected preceding vehicle. Thus, there is a need to design and develop an algorithm that can use the connectivity from its preceding vehicle and help a human driver in charge of their driving improve their driving.
To this end, a human-in-the-loop connected cruise control (hC3) algorithm that takes advantage of vehicle connectivity and keeps driver’s desire to be in control of their vehicle was developed. When the preceding vehicle shares its intended acceleration or deceleration, the ego vehicle can implement the proposed human-in-the-loop connected cruise control (hC3) algorithm. The algorithm takes the ego vehicle driver’s acceleration or deceleration and adjusts it to be optimal based on the preceding vehicle’s acceleration or deceleration. Before conducting field experiments on the performance of the proposed hC3, we designed a virtual simulation environment using an open-source platform, Carla (www.carla.org).
Based on 8 human drivers’ experiments, the hC3 reduced 36.8% acceleration variations and 15.8 % fuel consumption compared to human driving. Our immediate plan includes a field operational test based on a real vehicle environment using comma.ai device, an open-source advanced driving assistance system.