Robust recursive least squares algorithm for automotive suspension identification 2005-01-3996

This paper studies the behavior of a new recursive least-squares (RLS) adaptive algorithm for automotive suspension system identification. RLS algorithms tend to increase the identification error when the input power is low. Like the Variable Memory Length algorithm (VML) [10], the new algorithm, called Robust VML (RVML), is robust in system identification applications in which the input power is significantly reduced during operation. However, RVML is more robust then VML to variations in input power. Simulations with input signals generated according to models that represent the road profile show that the RVML algorithm has the same convergence speed as both the VML and the VFF [6] algorithms, the latter being the one of the most referenced ones for automotive suspension system identification. In steady-state, the RVML algorithm outperforms the other algorithms for any condition of input power. It should encounter application in automotive suspension fault detection systems and in adaptive control for semi-active suspension systems. In both cases, considerable periods of input power variation during operation are common.