Neural-based downlink scheduling algorithm for broadband wireless networks
Computer Communications
| Hi-index | 0.00 |
The unique capabilities of new cellular systems are expected to provide users with integrated multimedia services. Since the air interface still represents the system bottleneck, this paper proposes novel scheduling techniques to integrate efficiently the support of real-time traffic (i.e. voice and video) and data bursty traffic under quality of service (QoS) guarantees. Prioritization among traffic classes is adopted and a polling service discipline is employed within a class, where the permission rights of each traffic source are determined on the basis of token bucket regulators. Two polling-based approaches are compared to serve the sources of a class: (i) when a source is enabled to transmit, a burst of packets is sent at once; (ii) within the time interval destined to a traffic class, a cyclic service of the sources is allowed on a packet basis. With realistic assumptions on both radio channel conditions and protocol signaling overhead, this paper compares these two different approaches and the dynamic slot assignment++ (DSA++) scheme appeared in the literature. The obtained results highlight that our second scheme (case ii) allows increasing the number of supported video traffic sources of many units with respect to DSA++. Finally, an analytical approach has been proposed for our second polling scheme. Copyright © 2004 John Wiley & Sons, Ltd.