| The purpose of my dissertation will be to present a new routing strategy in packet switching networks in which intelligent capabilities for adaptive routing are concentrated in the packets, rather than in the nodes (or interchangeably routers) and protocols. Networks which contain such Cognitive Packets (CP) are called Cognitive Packet Networks (CPN). Cognitive packets route themselves. They are assigned Quality of Service (QoS) goals before entering the network, and pursue these goals adaptively. They learn to avoid congestion and to avoid getting lost or being destroyed through their own observations about the network and through the experience of other packets with whom they exchange information via mailboxes. Cognitive packets rely minimally on routers, so that network nodes only serve as buffers, mailboxes and processors. Starting from the original reinforcement learning based adaptive routing algorithm design, which was accompanied with an extensive literature survey, we went through simulations, evaluations by mathematical modelling, and we tried other possible approaches including genetic algorithms, finally we made CPN implemented over a testbed and took plenty of measurements. These whole series of steps form a complete development process and eventually converge to this dissertation. |
