Abstract - This paper aims to outline the main steps
of the dcsign of a sloxv-active suspension controller
able to deal with parametric uncertainties of the plant.
The most relevant noise inputs to a suspension system
are caused by road surface roughness (fast varying
inputs) and by cornering and fore and aft acceleration
(slow varying inputs). A previous paper was devoted
to the design of a semiactive suspension which consisted
of a 'backup passive system managing fast-varying
inputs and of an active one managing slow-varying
inputs. The suspension design was developed assuming
a relatively simple model - the two-degree-offreedom
quarter-car model - and replacing the conventional
components, spring and damper, with a simple
hydraulic device. A proper control law, designed
by means 0% the linear quadratic frequencyshaped
(LQ) methodology, achieved the result of minimizing
the sprung mass movements caused by a slow varying
downforce acting on it.
Hn practice, however, all suspension states are not
directly measurable, thus a KaPman filter has to be
introduced for state estimation: it yields a linear quadratic
Gaussian (LQG) controller. Both the fact that
the suspension dynamics depends on the static load
and the fact that parameters may vary while the
suspension works have suggested to compare the EQ
and the LQG active suspension horn the point of view
of stability robustness. This paper demonstrates that
the robustness properties of the LQG active suspension
are not necessarily bad and depend strongly on
the design of the backup passive suspension.