dc.description.abstract | Abstract
In practical applications such as uncertainties in the system model and signal statistics,
the H∞ filter has been proven to be a robust tool for signal prediction, parameter
estimation, and output feedback control system design. In this dissertation, we consider
a novel application of the H∞ filter and the development of the H∞−filter-based output
feedback control system.
(i) The application is to predict the burst and long-range dependent MPEG traffic flow
in a modern wideband network so as to improve the related QoS of the communication
network. The trend and periodic characteristics of MPEG video traffic are fully captured
by a proposed stochastic state-space dynamic model, which includes traffic parameters
in the state vector, to improve prediction accuracy. As the statistics of the underlying
processes are either unavailable or uncertain in real-time applications, a recursive H∞
filtering algorithm is proposed to estimate traffic parameters for long-range prediction.
Unlike previous prediction schemes, which predict I, P and B frames separately, the
proposed scheme predicts the composite MPEG video traffic. Simulation results based
on real MPEG traffic data show that the time-varying trend, the periodic components,
and the long-range dependence property can be splendidly predicted and captured by the
proposed method. The proposed scheme has superior performance and lower complexity
than some other adaptive neural network methods, such as TDNN, NARX, and Elman
neural networks, in long-range prediction. With accurate and fast long-range prediction
of MPEG video traffic, it is useful for dynamic bandwidth allocation with better network
utilization and less queue occupancy to improve traffic management in high speed packet
networks.
(ii) Finally, we shall study the robust H∞ output feedback control problem for nonlinear
stochastic continuous-time time-delay systems with state-dependent noise represented
by Takagi and Sugeno fuzzy model. Based on the fuzzy approach, the fuzzy controller and
the fuzzy state estimator which guarantee H∞ robustness performance for the considered
nonlinear stochastic systems can be obtained by solving bilinear matrix inequalities.
| en_US |