Trade-offs in automotive active suspension design

Abstract

In this thesis, a strongly consistent subspace algorithm for the identification of discrete-time, linear time invariant systems from nonuniformly spaced power spectrum measurements is proposed. A byproduct subspace algorithm to construct analytic functions from evaluations of their real or imaginary parts on finite subsets of the unit circle is developed. A connection between the subspace identification and the Lagrange-Sylvester interpolation problems is established. Pointwise constraints and trade-offs on closed-loop frequency responses are derived for a quarter-car active suspension model. The influence of tire damping on the design of an active suspension system is analyzed. The rms and the rms gain constraints for the quarter, half, and full-car suspension models are studied in the H2-optimal and multi-objective control frameworks. For the quarter and half-car models, the dependance of closed-loop rms responses on the tire damping is investigated. The multi-objective suspension control problem is formulated as a convex mixed H2/H? synthesis problem for the quarter, half, and full-car models and solved by using linear matrix inequalities. Next, the problem is re-formulated as a non-convex and non-smooth optimization problem for the quarter and full-car models and is solved by using HIFOO toolbox. Then, for the quarter and half-car models the assumption that tire damping coefficient is exactly known is relaxed and robust controllers to cope with polytopic tire damping uncertainties are designed. Finally, a prototype three-degrees-of freedom cabin model for a commercial truck is derived and an active suspension system is designed by using the linear-quadratic-Gaussian design methodology.