Rate adaptation for bursty data transmission in CDMA networks

We study rate adaptation for the uplink of CDMA networks with many ON/OFF data users. We propose a novel multi-access interference (MAI) model that takes into account simultaneously the burstiness of data traffic and time-varying fading. We characterize the MAI from a stochastic process perspective. Our finding reveals that the MAI exhibits scale-invariant burstiness and is self-similar. The self-similarity implies the long range dependence of MAI and hence a predictive MAI structure at coarse time scales, which can be exploited for interference management. Built on this finding, we devise a multiple time-scale MAI predictor, and explore it for rate adaptation, which predictor utilizes variable spreading gain and adaptive coding. Simulation results show that significant throughput gain can be achieved using our rate adaptation scheme. We also investigate the impact of feedback delay, heavy-tailedness, and fading on the system performance.