No label defined (Q4246547)

The authors are concerned with an optimal observation problem on a discrete dynamical system \[ x(t+1)=Ax(t),\quad t\in T=\{0,1,\dots,N\},\quad x\in\mathbb R^n, \] by means of a piecewise linear measuring device. An initial condition \(x(0)=x_0\) is known up to the `a priori' distribution \(\check X_0,\) \(x_0\in\check X_0=\{x\in\mathbb R^n:b_\star\leq Fx\leq b^\star\}\), \(F\in\mathbb R^{m\times n}.\) The solution to the optimal observation problem consists in finding among the initial conditions from the `a posteriori' distribution \(\widehat X_0\) such an initial condition which delivers the maximal value to the objective function. This problem is reduced to a linear programming problem with piecewise linear constraints on part of the variables. The criterion of local optimality for the admissible trajectory is given by using the adaptation method.