Fitting Empowered Cross-Layer Scheduling for Real-Time Wireless Communications

Real-time wireless communication techniques are expected to play a vital role in emerging Time-Sensitive Networking, deterministic networking, and tactile internet. Although there has been much work on the optimal cross-layer scheduling with discrete-state fading models, how to minimize the average power given both throughput and deadline constraints over wireless channels with continuous fading states remains open. In this paper, the optimal joint channel and queue-aware scheduling policy for continuous fading channels is presented based on a joint quantization and fitting approach. In particular, by quantifying both channel and queue states of the fluid-model-based communications over continuous fading channels, we formulate its approximate Markov model, the optimal scheduling policy of which is then revealed by the Constrained Markov Decision Process (CMDP). As our previous work showed the above discretized policy has a threshold-based structure, it can be easily approximated by a deterministic scheduling policy. As a result, we leverage the fitting approach to find the asymptotically optimal deterministic scheduling. The deadline-constrained capacity, as a function of both the average power and maximum delay, is also determined. Numerical results validate the effectiveness of our analysis.