Approximate dynamic programming, local or global optimal solution?
Ali Heydari, Sivasubramanya Nadar Balakrishnan · 2014
The problem of global optimality analysis of approximate dynamic programming based solutions is investigated in this study. Sufficient conditions for global optimality is obtained without requiring the state penalizing terms in the cost function or the functions representing the dynamics to be convex functions. Afterwards, the theoretical results are confirmed through a qualitative analysis of an example problem.