On the numerical synthesis of optimal feedback controllers

Jorge Silva, João Sousa

We investigate the performance of feedback control laws obtained via dynamic programming techniques for deterministic continuous-time systems with continuous state-space. The control law relies on a value function which is calculated numerically with a semi-Lagrangian scheme. Both the numerical solver and the control law require interpolation. We propose a new algorithm that minimizes the impact of the discontinuities introduced by the state constraints for finite and infinite horizon optimal control problems.

File: download