Towards a unified view of inference, communication and control

Technological advances in the control of processes over a communication network have necessitated the development of a unified theory of communication and control. This is currently not available in full generality but steps towards this can and have been taken. In this lecture, I attempt to present, in summary form, some of these developments. Shannon's Information Theory undoubtedly provides the fundamental limitations for reliable communication over a noisy channel in the point-to-point case. The Noisy Channel Coding Theory states that reliable communication in the sense that the error probability tends to zero (in the limit of infinite delay) is attainable if the source is transmitted at a rate below “capacity.” Conversely, reliable transmission is not possible if the source is transmitted at a rate above capacity. It follows from the Noisy Channel Coding Theorem that the coding and decoding necessary to achieve reliable transmission can be usefully separated into source coding and channel coding, and correspondingly channel decoding and source decoding, without losing reliable transmission. This leads to a universal architecture through a digital interface for reliable point-to-point communication.