Precise closed equations of optimal nonlinear filtering for a class of bilinear systems (Q1881943)

The \((n+ l)\)-dimensional state vector \(x\), \(y\), \(x= (x_1,\dots, x_n)^\top\), \(y= (y_1,\dots, y_l)^\top\), satisfies the system \[ dx= Fx\,dt+ G\,dw_1,\qquad dy= \Biggl(\sum^n_{k=1} x_k B_k\Biggr)\,y\,dt \] and the first component \(x\) of the vector is observed: \(dz= Hx\,dt+ Q\,dw_2\). Here \(w_1\in \mathbb{R}^r\) and \(w_2\in \mathbb{R}^s\) are independent standard Wiener processes, \(B_1,\dots, B_n\) are upper triangular \(l\times l\) matrices, \(F\), \(G\), \(H\), \(Q\) are matrices of the corresponding dimension. The authors propose a method for constructing precise finite-dimensional equations for the optimal estimation of the vectors \(x\) and \(y\) (clearly, the optimal estimate \(\widehat x\) for \(x\) is defined by the Kalman-Bucy filter).