powered by:
MagicWare, s.r.o.

Neuro-Dynamic Programming for the Exploration of Unknown Graphs

Authors:Baglietto Marco, University of Genoa, Italy
Battistelli Giorgio, University of Genoa, Italy
Scardovi Luca, University of Genoa, Italy
Zoppoli Riccardo, University of Genoa, Italy
Topic:2.4 Optimal Control
Session:Optimal Control Theory and Design Methods
Keywords: Neural networks, Dynamic programming, Graph, Control, Entropy

Abstract

In this paper, the problem of exploring stochastic graphs is addressed. The definition of the entropy related to the a-priori unknown parameters (the lengths of the a-priori unknown links) leads to the formulation of the problem as a stochastic optimal control one. The application of exact Dynamic Programming suffers the so-called curse of dimensionality. To overcome this drawback, an approximate technique is proposed making use of Neuro-Dynamic Programming. Exploiting the concept of frontier, any approximate solution of the problem is shown to generate a "proper" policy.