Power Efficient Scheduling under Delay Constraints over Multi-user Wireless Channels

TL;DR

This paper proposes a power-efficient uplink scheduling algorithm for multi-user wireless channels that learns system dynamics online and satisfies delay constraints, outperforming existing algorithms in power consumption and delay management.

Contribution

It introduces a novel learning-based scheduling algorithm that handles unknown system statistics and reduces power usage while meeting delay requirements in wireless networks.

Findings

Algorithm satisfies user delay constraints in simulations.

Power consumption is significantly lower than existing methods.

Learns system parameters online without prior knowledge.

Abstract

In this paper, we consider the problem of power efficient uplink scheduling in a Time Division Multiple Access (TDMA) system over a fading wireless channel. The objective is to minimize the power expenditure of each user subject to satisfying individual user delay. We make the practical assumption that the system statistics are unknown, i.e., the probability distributions of the user arrivals and channel states are unknown. The problem has the structure of a Constrained Markov Decision Problem (CMDP). Determining an optimal policy under for the CMDP faces the problems of state space explosion and unknown system statistics. To tackle the problem of state space explosion, we suggest determining the transmission rate of a particular user in each slot based on its channel condition and buffer occupancy only. The rate allocation algorithm for a particular user is a learning algorithm that learns about the buffer occupancy and channel states of that user during system execution and thus addresses the issue of unknown system statistics. Once the rate of each user is determined, the proposed algorithm schedules the user with the best rate. Our simulations within an IEEE 802.16 system demonstrate that the algorithm is indeed able to satisfy the user specified delay constraints. We compare the performance of our algorithm with the well known M-LWDF algorithm. Moreover, we demonstrate that the power expended by the users under our algorithm is quite low.