Coding for random access in wireless networks

Abstract

Wireless networks in the near future face a formidable challenge of accommodating a dense set of infrequently communicating devices characterized by small data payloads and strict latency and energy constraints. In such a scenario, providing energy efficient random-access access becomes a challenge. Information theoretic analysis of such systems becomes imperative to understand the gap from optimality of the methods of random-access currently employed. In this thesis we discuss the trade-off between the required energy-per-bit to achieve a target probability of error (per-user) and the number of active users. Previous works in this regard focused on the AWGN channel model. In this thesis we consider the issue of Rayleigh fading. Specifically, we use random coding with a subspace projection based decoder to get finite blocklength bounds from which we arrive at the trade-off. Further we justify the use of our decoder by proving its asymptotic optimality for the channel under consideration. We also show that the required energy-per-bit increases from around 0-2 dB (for AWGN) to around 8-12 dB under fading.