Risk-Sensitive Optimal Control

by Peter Whittle

Focusing on two major themes--risk-sensitive control and path-integral or Hamiltonian formulation--it features in-depth treatment of the risk-sensitive certainty-equivalence principle, the extension of the conventional LQG (linear/quadratic/Gaussian) treatment and the path-integral formulation. Looks at the exploitation of the equivalence between policy improvement and the Newton-Raphson method to yield fast guaranteed iterative methods of canonical factorization. Covers the deduction of the natural relationship between Riccati recursions and canonical factorization, the deduction of the form of the canonical factors and the integration of LEQG theory with several methods.

256 pages, Paperback

First published May 11, 1990