The key expected innovation of the next generation of radio-access networks (such as cellular) is their ability to service vast numbers of active devices as envisioned in the so-called "Internet-of-Things". Unfortunately, current networks were designed with the human-type communication in mind, and this led to a focus on the operating regime of a (relatively) few simultaneously active users. More specifically, present systems employ centralized resource allocation, thus orthogonalizing the access from different users. This solution is not acceptable for machine-type communication, as it relies on a significant control-layer overhead thereby incurring a significant penalty in latency and energy efficiency. Consequently, there is a strong economical demand for a new solution in both the unlicensed spectrum (so called, low-power wide-area networks) and the licensed spectrum (5G).
The goal of this work is to provide theoretical guidance for the design of the multiple-access layer in the next generation of wireless networks. Classical work on the topic lacks several specific details, making it inadequate: ignoring the control-layer overhead in network analytic literature, and ignoring delay in information theory. Consequently, this work aims to provide necessary contemporary modifications: (a) a gigantic number of idle (inactive) users; (b) a still large number of active users; (c) short packets; (d) high energy-efficiency (low energy-per-bit). This project introduces a new paradigm of random-access coding that separates data communication from user identification, for which fundamental limits are going to be derived and the low-complexity practical solutions studied. Performance of the currently available solutions will be contrasted with the non-asymptotic fundamental limits and new solutions developed. In addition to information-theoretic and communication-theoretic parts, the work involves a combinatorial-theoretic component in the form of constructing Sidon sets, B2-sequences and superimposed codes.
