Event

Orthogonal residual learning for sideslip angle and cornering stiffness estimation

1 reports · 1 sources

Kyoungtae Ji, Dongyeol Won, and Kyoungseok Han propose a sensitivity-orthogonal residual learning method, using null-space parameterization to constrain the model corrector to be orthogonal to the cornering stiffness sensitivity, preventing the corrector from absorbing stiffness variation information and thereby interfering with online adaptation, for joint estimation of vehicle sideslip angle and cornering stiffness. In full-vehicle simulations, the method reduces sideslip angle error by 12.7% to 22.0%, and maintains the same adaptation speed as the uncorrected filter under step changes in stiffness; the stiffness response of an unconstrained corrector of the same scale decreases as the number of parameters increases. On public real-vehicle data, the 430-parameter orthogonal corrector retained more response to nominal stiffness offset than an unconstrained corrector of the same scale, but its sideslip angle error was slightly higher than that of the uncorrected filter. The paper has been submitted to Vehicle System Dynamics.