UC San Diego

Future time-division-duplex (TDD) systems operating over small wireless networks will utilize intelligent base station (BS)-coordinated dynamic channel-allocation algorithms in order to support high-bandwidth asymmetric traffic in adjacent cells. In this paper, we use extensive measurements of wireless Internet traffic from a large 802.11b network to create two random traffic models. One model, called “binomial,” is memoryless and the other, called “dynamic,” is based on an event-driven Markov state model with bidirectional flows and deterministic residence times. We then develop a two-BS two-zone wireless TDD interference model that describes the spatial features of interference between cochannel mobile stations (MSs) in adjacent BSs. This is a simplified precursor to more sophisticated models for multiple BSs and/or multisector BSs. We present a set of candidate TDD channel-allocation algorithms, which vary in their level of time-slot coordination and intelligent allocation between BSs. Lastly, we combine the three components (i.e., traffic models, interference models, and channel-allocation algorithms) to demonstrate the capacity for evaluating dynamic channel-allocation algorithms in realistic interference and Internet traffic scenarios. The results show that, for active MSs, the dynamic traffic model has a higher number of packet requests per time frame than the binomial traffic model, given the same mobile activity factor. Additionally, fixed channel-allocation algorithms generally perform much worse than pseudorandom and intelligent BS-coordinated algorithms, especially for asymmetric BSs. The pseudorandom algorithm performs well at low traffic, but suffers from severe interference blocking at high traffic. The intelligent BS-coordinated algorithm performs best, as it avoids MS-to-MS interference blocking from nearby users in adjacent cells and maximizes the overall throughput by attempting to allocate up- and downlink packet requests in corresponding time slots matched to the incoming uplink–downlink traffic demand for each time frame.