Location and Driver-Specific Vehicle Adaptation Using Crowdsourced Data

This paper presents a method that adjusts the operation of advanced driver-assistance systems (ADAS) to a specific location and driver. The method uses crowd-sourced data collected from multiple drivers in multiple lo-cations/environments to predict the vehicle behavior of an individual driver in a previously unseen location/environment. This prediction can in turn be used for adapting the calibration of ADAS to the specific location/environment, as well as to the individual driver. We describe an algorithm for making predictions, which uses probabilities and quantile functions of empirical cumulative distribution functions to relate an individ-ual driver to the population. The paper reports a simulation study in SUMO (Simulation of Urban MObility), where an emergency braking system is adapted to individual drivers and to different road surface conditions. The results show that the algorithm is quickly able to make accurate predictions and consequently adjust ADAS to the specific location and driver.