Provision of Maximum Delay Guarantee at Low Energy in a Multi-User Environment

Abstract

In this work a simple scheduling scheme is proposed for wireless sensor networks in presence of a maximum tolerable delay constraint. The scheduler performs the dual task of schedul- ing the users experiencing high channel gains to minimize the energy consumption while at the same time, it takes into account the maximum allowed buffer length of each user to provide an upper bound on the maximum tolerable delay. The idea of a threshold depending on the buffer occupancy is proposed to schedule the users opportunistically unless the deadline for transmission is reached. A maximum delay guarantee is provided in the proposed scheme in addition to average delay guarantee at almost no additional energy cost. The results show that multilevel recursive optimization can minimize energy subject to maximum bounded delay for schedulers that empty the buffer. I. INTRODUCTION Energy saving and delay constraints are one of the most demanding requirements for the state of the art wireless com- munication networks. Specifically, wireless sensor networks (WSN) put an emphasis on the energy saving aspect of the system. WSN consists of a large number of nodes with the sensing, computation and communication capabilities merged together. One of the fundamental and most important tasks in designing protocols for WSN is minimization of energy expen- diture to increase the life time of the network. Sensor nodes can save the measured data locally for some duration and wait in sleep mode before transmitting it to the data collecting node, called fusion node. When they find good channel conditions, they wake up and empty the buffer by transmitting the whole data. Some applications explicitly require the transmission of sensed data before a hard deadline and therefore, often an upper delay bound for each node needs to be provided. This work deals with the dual task of minimizing the energy of the system while providing an upper delay bound for each node. The work in (1) deals with maximization of the information capacity by scheduling the users having the instantaneous channel quality near the peaks. This form of diversity in which different users experience independent channels at the same time is called multiuser diversity. In wireless systems, channel fading has been treated as a source of uncertainty but in the context of multiuser diversity, it can be considered as randomness that can be exploited by scheduling the users experiencing a good channel. Reference (2) discusses a scheme to increase the random fading in a slow fading environment by using multiple antennas on the transmitter side of the downlink. The larger the number of the users in the multiuser environment, the greater is the chance that some of the users will be experiencing channel near the peaks. In (2) an oppor- tunistic scheduling scheme called proportional fair scheduling (PFS) is proposed to provide the fairness guarantees to all users. Reference (3) deals with the tradeoffs between average delay and average power. In (4), a scheduling policy has been proposed to maximize the expected data throughput by using dynamic programming. Similar work in (5) considers a scheduling policy without a centralized scheduler and discusses the energy delay trade off with full and partial shared information about the queue lengths of all the users. In (6), an exact solution for the average packet delay under the optimal offline scheduler has been presented. The results of (3) have been extended to the multiuser context in (7) and the result yields that to achieve an average power within the O(1/L) of the minimum power required for the network stability, there must be an average