The role of information in the context of control is a deep issue. To get at this, we review Witsenhausen’s notions of information patterns for control problems. While staying in that basic framework, we then use ideas from traditional information theory as we re-examine Witsenhausen’s famous “counterexample”. In the process, we construct a family of nonlinear “quantizing” control laws that can perform infinitely better than the best linear policies.
Conditional Distribution Gaussian Random Variable Quadratic Cost Stochastic Optimum Control Nonlinear Controller
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.
This is a preview of subscription content, log in to check access.