Analysis of Wireless Communications with Finite Blocklength and Imperfect Channel Knowledge

With the rise of critical machine-to-machine (M2M) applications, next generation wireless communication systems must be designed with strict constraints on the latency and the reliability. A key enabler for designing low-latency systems is the availability of accurate and tractable analytic models. Unfortunately, many performance models do not account for the effects of channel coding at finite blocklength and imperfect channel state information (CSI) due to time-limited channel training. These models are therefore inaccurate for low-latency systems. In this work, we derive a closed-form approximation for the transmission error probability while considering both of these effects. The approximation provides an inverse mapping from the error probability to the achievable rate. Using this approximation, we analyze the queuing delay of the system through stochastic network calculus. Our results show that even though systems with rate adaptation must spend resources on channel training, these systems perform better than systems operating at fixed rate. The enhancement becomes even greater under stricter delay constraints. Moreover, we find that adapting both the training sequence length and the code rate to the delay constraints is crucial for achieving high reliability.