Abstract
A voluminous set of longitudinal road profiles gathered from the Long Term Pavement Performance (LTPP) program was processed using median filtering to separate individual large obstacles from the basic quasi-homogeneous random road unevenness. The shapes of 16590 obstacles were analysed partitioned into four sub-groups according to the asphalt-concrete vs. cement-concrete road surfaces, and positive (bumps) vs. negative (potholes) vertical elevations. The empirical obstacle shapes were fitted using seven symmetric and two asymmetric analytical shapes. Based on the root mean square error (RMSE) criterion and accounting for some practical aspects connected with the possible use for the simulation of road profiles with obstacles generalized parabolic shapes (power function) and half-sine shapes seem to be recommendable as adequate approximation functions. Asymmetric obstacle shapes of cosine and parabolic shapes were also briefly discussed and their impact on the vibration transient response of a quarter-car model outlined.