Bicriterial Dual Control with multiple linearization

A suboptimal dual controller for discrete stochastic systems with unknown parameters based on the bicriterial approach is proposed and discussed. It is supposed that all the random quantities are non-Gaussian. This assumption induces that a global estimation method has to be used. The Gaussian sum method with multiple linearization technique was chosen and applied in the bicriterial control approach. The probing part of the control law is determined for each local node of estimated probability density function separately and respects accuracy of each local estimate inherent in the estimated probability density function. A comparison of the proposed modified bicriterial controller and the bicriterial controller which uses global point estimate only is shown in some numerical examples.