Event-Triggered Communications Over AWGN Channels

Event-triggered communication (ETC) has attracted considerable recent attention because it holds the promise of saving bandwidth and energy in real-time monitoring, networked control, distributed optimization, federated learning, etc. However, its performance limits remain open from a communication-theoretic perspective. This paper investigates event-triggered communications over AWGN channels, where events are modeled by a semi-Markov process. An ETC-oriented code is conceived by adding an all-zero codeword that represents a non-triggered state without any additional energy cost to the random codebook. We present the maximum achievable channel coding rates in both finite and infinite blocklength regimes. Even though the newly added codeword may incur a higher error probability, the rate of infinite blocklength coding is found to remain unchanged, thereby assuring the applicability of Shannon’s formula in ETC. Furthermore, we present two multi-user ETC protocols based on statistical multiplexing and joint coding respectively. Borrowing the idea of information freshness and queueing analysis, we reveal the distortion-power tradeoffs of single- and multi-user ETC. The ETC with Lebesgue sampling is shown to outperform the conventional periodic communication with Riemann sampling. The number of users in statistical multiplexing is also optimized. Numerical results validate our theoretical analysis and demonstrate the substantial power-saving gain of ETC.