Throughput Maximization With an Average Age of Information Constraint in Fading Channels

• Published in: IEEE transactions on wireless communications Vol. 20; no. 1; pp. 481 - 494
• Main Authors: Bhat, Rajshekhar Vishweshwar, Vaze, Rahul, Motani, Mehul
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.01.2021; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Age; Age of information constraint; age-independent stationary randomized policies; Broadband; Coherence; Fading; Fading channels; Freshness; Information flow; Maximization; Optimization; perfect and no CSIT; performance bounds; Policies; Reliability; Throughput; throughput maximization; Transmitters; Wireless communication; Wireless communication systems; Wireless communications
• ISSN: 1536-1276; 1558-2248
• DOI: 10.1109/TWC.2020.3025630

In emerging fifth generation and beyond wireless communication systems, communication nodes are expected to support information flows that are freshness-sensitive , along with broadband traffic having high data rate requirements. Freshness-sensitive flows, where freshness is quantified by a metric called the age of information (AoI), are naturally assigned priority over resources. Motivated by this, we consider long-term average throughput maximization in a single user fading channel, subject to constraints on average AoI and power, and knowledge of channel state information at the transmitter (CSIT), which is the realization of channel power gains. We consider two scenarios: (i) when Perfect CSIT is available and (ii) when CSIT is not available. In both scenarios, the channel distribution information is available. We consider a generate-at-will model, in which update packets can be generated in any block of interest, at the transmitter. We propose simple age-independent stationary randomized policies (AI-SRP), which allocate powers at the transmitter based only on the channel state and/or distribution information, without any knowledge of the AoI. We show that the optimal long-term average throughputs achieved by the AI-SRPs are equal to at least half of the throughputs achieved by optimal policies, independent of all the parameters of the problem. Furthermore, we provide an expression that bounds the difference in throughputs achieved by the optimal policies and AI-SRPs. Finally, we provide extensive numerical results to illustrate the performance of AI-SRPs.