Abstract
A road profile estimation method based on vehicle responses measured by a smartphone is proposed. A half-car model, which can represent vehicle bouncing and pitching motions, is employed; the road profiles at the tire locations are included as state variables in the augmented state vector; these variables are estimated by the combination of a Kalman filter, the RTS smoothing, and the Robbins-Monro algorithm. Because the estimation accuracy depends on the vehicle modeling accuracy, an algorithm to calibrate the model is also proposed. The algorithm optimizes the vehicle parameters by minimizing the difference between identified road profiles at the front and rear tire locations. The rationale is that the two profiles are estimates of one physical profile while they are dealt with as two independent state variables. Through numerical simulation, the accuracy of the profile estimation and parameter identification are clarified. Drive tests with three rental cars showed that the proposed algorithm can compensate for the vehicle and drive speed differences; the estimated IRI values are consistent among the three vehicles and with reference profiler.
