The paper describes a low complexity tree-structure based user scheduling algorithm in a MLD-based multi-user multiple-input multiple-output (MIMO) wireless systems. An M-branch selection algorithm, which selects M most-possible best branches at each step, is proposed to maximize the whole system sum-rate capacity. To make this work more feasible, some implementation problems such as complexity reduction, power control and performance distortion of channel estimation are also discussed in this paper. Both theoretical analysis and simulation results show that the complexity of proposed scheme, linearly increases with the number of user candidates and the number of antennas to be selected and is much lower than full research algorithm. Furthermore the algorithm can also be performed independently with power control and is robust to the channel estimation error.