Design of Bacterial Foraging Optimization Algorithm Based Adaptive Sliding Mode Controller for Inverted Pendulum

摘要

1 Introduction Traditional Sliding Mode Control, also called variable structure systems (VSS), was designed for the systems with uncertainties due to its excellent robustness. But it also presents some drawbacks such as undesired large control authority and chattering problem. These drawbacks were overcome by Slotine and Sastry [1] by replacing control switching at a fixed sliding surface by a smooth control interpolation in a boundary layer neighboring a time varying sliding surface. This eliminated the excitations of high-frequency un-modeled dynamics and leads to an explicit trade-off between the model uncertainty and controller tracking performance. In recent years a large amount of research into the area of sliding control has been developed. Fernanderz and Hedrick [2] generalized the sliding mode approach to a larger class of multivariable systems. Shyu and Tsai [3] introduced a multiple sliding surfaces in the sliding controller design to eliminate control chattering. In Oh and Khalil [4] a VSS controller with a high-gain observer was designed as a function of the state estimates to ensure attractiveness of sliding manifold. Won and Hedrick [5] developed a 'multiple-surface' sliding control for a class of single-input single-output (SISO) non-linear systems whose uncertainties do not satisfy the matching condition. Edwards and Spurgeon [6] presented a controller/observer pair based on sliding mode ideas, provided robust output tracking of a reference signal. Chen and Toshio [7, 8] used the transfer function method based on the VSS theory to estimate the disturbance. The estimated disturbance was employed to construct a VSS-type state observer. Further, the estimated disturbance and the state observer were applied to a controller to place the desired stable poles and to cancel the disturbance [8]. A common assumption of these studies was that the bounds of model uncertainties must be known. If the bounds of model uncertainties were unknown, traditional sliding