Abstract:

This paper addresses the center-of-gravity height and dynamics estimation problem, which is key for rollover prevention systems in automotive. We model the vehicle as a spring-damper system and develop a Bayesian method that outputs estimates of the center-of-gravity height, suspension stiffness and damping coefficient. We leverage the model structure to design a computationally efficient particle filter, which, combined with Bayesian optimization for particle initialization and a particle-size adaptation scheme, leads to an implementation that provides accurate, smooth estimates of CoG height, stiffness, and damping. A Monte-Carlo simulation study on a standardized maneuver shows that the method almost instantaneously provides reliable estimates that represent well the true parameter values.