Users and Services in Intelligent Networks

We present a vision of an intelligent network in which users dynamically indicate their requests for services, and formulate needs in terms of quality of service (QoS) and price. Users can also monitor on-line the extent to which their requests are being satisfied. In turn the services will dynamically try to satisfy the user as best as they can, and inform the user of the level at which the requests are being satisfied, and at what cost. The network will provide guidelines and constraints to users and services, to avoid that they impede each others' progress. This intelligent and sensible dialogue between users, services and the network can proceed constantly based on mutual observation, network and user self-observation, and on-line adaptive and locally distributed feedback control which proceeds at the same speed as the traffic flows and events being controlled. We illustrate these concepts via an experimental test-bed at Imperial College, based on the cognitive packet network (CPN), that embodies some of these functionalities thanks to "smart packets" and reinforcement learning. At its edges, CPN is fully compatible with the IP protocol, while internally it offers routing that is dynamically modified using on on-line sensing and monitoring, based on users' QoS needs and overall network objectives.