Determination of extreme values in autonomous driving based on multifractals and dynamic scaling

Abstract

This talk will be a review of two methods based on the dynamic scaling and multifractal statistic of corner radar sensor time series with the ultimate aim to estimate extreme values for highly autonomous driving test signals. It is assumed that the extremity of the influencing parameter of the sensor correlates with the probability of collision of the vehicle.The method of dynamic scaling is derived originally from the surface or interface physics. The sensor parameter time series can be considered as fractal surface or interface and grows adequately to a specified scale invariant dynamical equation. It is demonstrated that multifractal statistic can be useful searching for statistical analogous time series, disposed similar multifractal spectrum as the original sensor time series.