Design of Practical Control Algorithms for Nonlinear stochastic systems.
DANIEL L. ALSPACH · 1977
The method of aligning the dual goals of the general optimal stochastic control as a design tool is discussed. When the two goals are exactly aligned, the certainty equivalence control is optimal and no additional intentional probing is required. If these goals are anti-aligned, demanding opposing controls, the certainty equivalence control can be, locally at least, the worst control possible. The case of a nonlinear filter in a feedback loop with dual goals aligned is shown to have performance very close to that of an optimal stochastic control. A new nonlinear filter was developed for the adaptive Kalman filtering problem where only the measurement noise is unknown. (Author)