Reinforcement learning for routing in communication networks

[摘要] ENGLISH ABSTRACT:Routing policies for packet-switched communication networks must be able to adaptto changing traffic patterns and topologies. We study the feasibility of implementingan adaptive routing policy using the Q-Learning algorithm which learns sequences ofactions from delayed rewards. The Q-Routing algorithm adapts a network's routingpolicy based on local information alone and converges toward an optimal solution. Wedemonstrate that Q-Routing is a viable alternative to other adaptive routing methodssuch as Bellman-Ford. We also study variations of Q-Routing designed to better explorepossible routes and to take into consideration limited buffer size and optimize multipleobjectives.

[发布机构] Stellenbosch University