Robust Model Predictive Control Of Networked Systems

Model predictive control which was originated in the mid 70s of 20th century has been widely applied in complex industrial process control, especially in industrial sections of petroleum, chemistry, metallurgy, mechanism and so on, for it has many merits such as simple model, convenience of computer implement, effectiveness of handling hard input and output constrains. Besides, because of the approximation and uncertainty of mathematical model, the effect of model nonlinearity, the inaccuracy of measurement, the in-house parameterization of the control system and so on, engineering systems in actual operation are always affected by the uncertainties. Robust control emerged because of demand of solving the control problems of these kinds of control systems with uncertainties. In addition, between the late 1980s and 1990s, computer network was more and more widely applied in area of automatic control. In contrast to traditional control systems, networked control systems have many merits, such as low cost, simple and convenient to install and maintain, high reliability, having good modularization and flexibility and so on. But analysis and design of the whole system become to be complex, for the in-house problems of time delay and data packet dropout, etc. So, how to design networked control systems which can deal with these problems has became one of the most popular issues in control area recently.Based on the existing research results of robust model predictive control, first this paper investigates the algorithm of robust model predictive control for systems with data packet dropout, A sequence of stochastic variables satisfying Bernoulli random binary distribution is adopted to model the data packet dropout. Robust model predictive controller which guarantees some closed-loop stochastic stability for all system parameter uncertainties and admissible data packet dropout is designed. The proposed algorithm is applied to a classical angle positioning system to show that our algorithm has better performance than that of N. Wada et al. In addition, the problem of design of robust model predictive controller for uncertain systems with state and input delay is studied, respectively. The system uncertainty is assumed to be polytopic type, the units of time is also assumed to be an unknown constant, but one of its upper bounds is assumed to be a known constant. and time delay based on the existing research results. The proposed algorithm has better performance than that of S. C. Jeong etc, for the parameter-dependent Lyapunov function method is adopted. At last, we apply the robust model predictive control algorithm with data packet dropout to the water level control of the three-tank system. The simulation result shows the effectiveness of our proposed robust model predictive control algorithm.