Traffic modeling and tracking for multiuser detection for random access networks

Traffic burstiness results in the predictability of user activity at the individual source level, and the exploitation of such predictability in the receiver design for wireless packet switching random access/CDMA networks is investigated. It is shown that the conventional approach of assuming all users are active results in substantial performance loss when a linear multiuser detector, and in particular, a decorrelating detector is implemented. A two-stage receiver is proposed where the first stage tracks active users and prior to the second stage symbol detection. Two different user trackers are presented and it is demonstrated that, with the help of traffic predictability, accurate estimate of active set of users is possible, even with a simple matched filter bank implementation at the first stage.