Design of Luenberger State Observers using Fixed-structure H-infinity Optimization and Its Application to Fault Detection in Lane-Keeping Control of Automated Vehicles
Soichi Ibaraki (Kyoto University)
Shashikanth Suryanarayanan (IIT Bombay)
Masayoshi Tomizuka (University of California at Berkeley)
Abstract
Lane-keeping control forms an integral part of fully-automated
Intelligent Vehicle Highway Systems (IVHS). Therefore, its reliable
operation is critical to the operation of an automated highway. In
this paper, we present the design of a fault detection filter for the
lane-keeping control systems on-board vehicles used by
California-PATH, USA in its automated highways program. We use a
Luenberger structure for the fault detection filters and tune the
observer gains based on an H-infinity based cost. Such a
choice of cost was motivated by the need to explicitly incorporate
frequency domain based performance objectives. The LMI-based
formulation of an H-infinity optimization problem of
Luenberger state observers does not allow for the augmentation with
dynamic performance weightings in the optimization objective, since it
makes the problem a non-convex optimization problem. We present an
algorithm to locally solve the problem of design of Luenberger state
observers using H-infinity optimization by transforming
the problem into an H-infinity static output feedback
controller problem. Experimental results demonstrate the efficacy of
the tuning methodology by comparing the fault detection performance of
filters that use H-infinity Luenberger observers versus
those that use Kalman Filters. Implementation issues of the observers
are also discussed.
Key Words: State estimation, H-infinity Optimization, Frequency shaping, Fault detection, Integrated Vehicle Highway Systems.
|
