OFDMA Resource Allocation Based on Traffic Class-Oriented Optimization

Nararat RUANGCHAIJATUPON  Yusheng JI  

Publication
IEICE TRANSACTIONS on Communications   Vol.E92-B   No.1   pp.93-101
Publication Date: 2009/01/01
Online ISSN: 1745-1345
DOI: 10.1587/transcom.E92.B.93
Print ISSN: 0916-8516
Type of Manuscript: Special Section PAPER (Special Section on Networking Technologies for Dependable Networks)
Category: 
Keyword: 
OFDMA,  cross-layer scheduling,  QoS,  proportional fairness,  multi-user diversity,  multi-channel,  IEEE 802.16,  

Full Text: PDF(446.2KB)
>>Buy this Article


Summary: 
Orthogonal Frequency Division Multiple Access (OFDMA) is the technique for the next generation wireless networks, whose enhanced capacity is to serve a combination of traffic with diverse QoS requirements. To realize this, the resource allocation scheme has to be carefully designed so that the instantaneous channel condition, QoS provision, and the network utilization are integrated. In this paper, we propose the resource allocation scheme for downlink traffic of 2 classes; guaranteed and non-guaranteed, having different traffic contracts. We provide guaranteed throughput for the guaranteed class by considering the cost incurred from serving this class. Then, we formulate the assignment problem with the objective of minimizing this cost. For the non-guaranteed class, we aim to maximize network utilization and to maintain throughput fairness, by employing Proportional Fairness (PF) utility function and emphasizing on the portion of network resource that the user received and the individual user's queue length. We use a heuristic approach to schedule users' data into the downlink subframe by exploiting multi-user multi-channel diversity to utilize system's bandwidth efficiently. Intensive simulation shows that our scheme differentiates classes of traffic and provides satisfied throughput, lower packet drop rate, and lower queuing delay to the guaranteed class, comparing with those of the non-guaranteed class. Furthermore, the results also show that the scheme is fair to users in the same class in both throughput and service time.