This paper presents a real-time driver characterisation algorithm that models the driver based on 10 parameters governing their control of vehicle longitudinal speed and acceleration. It utilises an open-loop longitudinal model in conjunction with an Unscented Kalman Filter (UKF). The algorithm can operate in real-time using velocity and acceleration measurements, in addition to a-priori knowledge of the path curvature and road features such as speed limit locations. The UKF also enables a system identification process to estimate average response of each driver. These identified parameters are subsequently fitted to independent measures of fuel economy and safety to demonstrate possible uses for the characterisations. In addition to the comprehensive characterisation of the drivers’ behaviour the work is also novel in its unique approach to identifying time-based parameters within a time-varying dynamic system (UKF). Possible applications include insurance black box assessments of driver aggressiveness and safety, and in autonomous driving algorithms by providing a model to closely replicate the specific driving styles of various users.


    Access

    Download


    Export, share and cite



    Title :

    Characterisation of driver longitudinal behaviour using an Unscented Kalman Filter


    Contributors:


    Publication date :

    2023-12-01




    Type of media :

    Article (Journal)


    Type of material :

    Electronic Resource


    Language :

    English





    Unscented Kalman Filter

    Zarchan, Paul / Musoff, Howard | AIAA | 2015


    Unscented Schmidt–Kalman Filter Algorithm

    Stauch, Jason / Jah, Moriba | AIAA | 2014


    Unscented Schmidt-Kalman Filter Algorithm

    Jason Stauch | Online Contents | 2015


    Unscented Partial-Update Schmidt-Kalman Filter

    Kevin M Brink | Online Contents | 2017


    Square-Root Unscented Schmidt–Kalman Filter

    Geeraert, Jeroen L. / McMahon, Jay W. | AIAA | 2017