In this paper we study the downlink of a TDD (TimeDivision Duplex) single cell system where the Base Station (BS)employs multiple antennas to serve the users taking into accountthe traffic patterns. The BS chooses each slot the users to beactive, and serves them using Zero Forcing (ZF) precoding. Thisrequires the knowledge of the users’ channels which is assumedto be performed e.g. via uplink training. Due to the channelacquisition overhead, only a subset of users must be active at eachtimeslot (depending on traffic patterns and channel states). In thispaper, we develop an active user selection strategy where the basestation sets a given threshold for the channel gain of the users.Then, only the users that have their channel gain higher than thethreshold send their training sequence and are then considered tobe served. The base station estimates the channel states of theseusers and, due to channel reciprocity, uses these channel statesto transmit data via ZF precoding. With appropriate signalingand threshold selection, which adapt to the queuing behavior ofthe users, we prove that our proposed method achieves a largerstability region than the baseline centralized policy where the BSselects the users based on channel statistics and queue lengths.The performance of the threshold-based method is illustratedvia simulations, where we can observe a tradeoff between theexpansion of the stability region and delay performance.