494 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 49, NO. 2, MARCH 2000 Adaptive Quality of Service Handoff Priority Scheme for Mobile Multimedia Networks Wei Zhuang, Brahim Bensaou, Member, IEEE, and Kee Chaing Chua, Member, IEEE Abstract—For various advantages including better utilization of radio spectrum (through frequency reuse), lower mobile transmit power requirements, and smaller and cheaper base station equipment, future wireless mobile multimedia networks are likely to adopt micro/picocellular architectures. A consequence of using small cell sizes is the increased rate of call handoffs as mobiles move between cells during the holding times of calls. In a network supporting multimedia services, the increased rate of call handoffs not only increases the signaling load on the network, but makes it very difficult for the network to guarantee the quality of service (QoS) promised to a call at setup or admission time. This paper describes an adaptive QoS handoff priority scheme which reduces the probability of call handoff failures in a mobile multimedia network with a micro/picocellular architecture. The scheme exploits the ability of most multimedia traffic types to adapt and trade off QoS with changes in the amount of bandwidth used. In this way, calls can trade QoS received for fewer handoff failures. The call level and packet level performance of the handoff scheme are studied analytically for a homogeneous network supporting a mix of wide-band and narrow-band calls. Comparisons are made to the performance of the nonpriority handoff scheme and the well-known guard-channel handoff scheme. Index Terms—Adaptive QoS, handoff priority scheme, mobile multimedia networks. I. INTRODUCTION I N RECENT years, there has been a tremendous effort of research and development expended to realize future mo- bile communication networks that will support integrated and multimedia services. For example, a large number of projects funded under the European Commission’s Advanced Commu- nication Technologies and Services (ACTS) program are fo- cused on research related to third-generation mobile commu- nication systems [1], such as the universal mobile telecommu- nications system (UMTS) and wireless asynchronous transfer mode (ATM). Due to the scarce bandwidth as well as errors on the wireless channel, efficiency of resource utilization is a key issue in the design and implementation of wireless mobile net- works in general and mobile multimedia networks in particular. For various advantages including better utilization of radio spec- trum (through frequency reuse), lower mobile transmit power requirements, and smaller and cheaper base station equipment, future mobile multimedia networks will adopt micro/picocel- lular architectures. A consequence of using small cell sizes is Manuscript received September 11, 1997; revised April 27, 1999. W. Zhuang is with the Department of Electrical Engineering, National Uni- versity of Singapore, Singapore. B. Bensaou and K. C. Chua are with the Centre for Wireless Communications, National University of Singapore, Singapore (e-mail: cwcbb@cwc.nus.edu.sg). Publisher Item Identifier S 0018-9545(00)02547-0. the increased rate of call handoffs as mobile terminals move be- tween cells during the holding times of their calls. In a network supporting multimedia services, the increased rate of call hand- offs not only increases the signaling load on the network, but, and more importantly, may adversely impact the quality of ser- vice (QoS) of the calls due to handoff failures. Note that in this paper we are interested in intercells handoff (i.e., handoff due to mobility between cells) rather than intracell handoff where the mobile hands off due to radio interference. From a user’s perspective, having a connection terminated in the middle of a call because of a handoff failure is far more annoying than having a new call attempt blocked occasionally. To reduce handoff failures, various schemes have been devised to prioritize bandwidth (channels) allocation to handoff calls rather than accepting new calls into the network [2]–[6], [11]. Mainly, two generic schemes are considered in the literature. In the so-called guard channel (GC) scheme (e.g., [7]), a number of channels in each cell are reserved for exclusive access by handoff calls. It is well known that the GC scheme, although simple to deploy, has many disadvantages. First, reserving chan- nels continuously translates into limiting drastically the total carried traffic. Second, it has been shown in [7] that the GC scheme is unable to provide fair QoS to different types of ser- vices efficiently. For instance, more bandwidth has to be re- served for wide-band 1 handoff calls; doing so, however, leads to a very low utilization of the scarce radio bandwidth. The second scheme, the so-called handoff-queuing (HQ) scheme, exploits cell overlaps to allow handoff calls to queue and wait for a cer- tain time for channels to become available. This approach is not practically feasible for real-time multimedia services in pico- cellular networks, since the time interval available for queueing handoff requests might not be sufficiently long enough for band- width resources to become available, especially for wide-band handoff calls. In [8], a new handoff scheme which exploits channel sub- rating has been proposed for personal communications service (PCS) systems with one class of service. The scheme improves system performance in terms of call blocking and handoff drop- ping probabilities by allowing an occupied full-rate channel to be temporarily divided into two half-rate channels, one to serve the existing call and the other to serve a handoff call when there is no idle channel available at the time the handoff call arrives. For PCS systems, the scheme clearly exploits the ability of users of telephony services to tolerate the use of half-rate vocoders with some degradation in speech quality. Indeed, this is also true for most multimedia services involving audio and video appli- 1 Wideband here refers to calls with high-bandwidth requirements such as those carrying high-quality audio and video traffic. 0018–9545/00$10.00 © 2000 IEEE