Performance of throughput-based Q-routing

Abstract

This paper proposes a scheme to reduce an initial learning period of Q-routing. Q-routing is a routing scheme to guide a packet on the fastest route. A neighbor node with the lowest Q value, which is determined by a remaining time to the destination, is selected to send the packet. Before the routing is stable, the node learns the Q value at the initial period. The Q-routing performs well in only a high traffic network, compared to a shortest path routing scheme. The proposed scheme reduces the initial learning period of Q-routing by considering a throughput of the node.