Complete Communications Engineering

User Selection for Multiple Antenna Broadcast Channel can improve Multiple-input Multiple-output (MIMO) system throughput by selecting users with proper channel conditions using dirty-paper coding (DPC) and zeroforcing (ZF) beamforming strategies.

MIMO systems have drawn a lot of attention in recent years due to their great potential to achieve high throughput in wireless systems. In a single-user case, it was shown that the MIMO channel capacity increases linearly with the minimum number of transmit and receive antennas. In multi-user systems, multiple antennas can be easily deployed at the base station to benefit the capacity gain. However, due to the cost and size limitations, the numbers of antennas deployable at a mobile terminal is very limited. Thus, the MIMO channel capacity gains are limited by the number of receive antennas at each user, especially in the time division multiple access (TDMA) strategy, where the base station transmits data to only the best user at a time. Thus, multiple-antenna capacity gains with the TDMA strategy are very limited.

To achieve the MIMO broadcast channel capacity improvement, the base station should serve multiple users simultaneously. One of the important multi-user strategies is called dirty-paper coding (DPC). DPC is based on interference pre-subtraction with a complex successive encoding scheme. It can achieve a linear increase of capacity in the number of transmit antennas at the base station regardless of the number of receive antennas at the mobile users. However, the DPC technique is based on non-causal knowledge of each user’s interfering signal at the base station. Optimal capacity-achieving DPC for multi-user MIMO downlink also requires optimization of the transmit signal covariance matrix and optimized transmit power allocation.

The most common linear precoding schemes involve zeroforcing (ZF). In the ZF precoding, the base station transmits signals in a way to isolate the users’ transmissions and hence decouples the multiuser channel into multiple independent sub-channels. Normally the ZF precoder is assumed to implement a pseudoinverse of the channel and to be null space basis of the linear space spanned by other users’ channels. Optimized ZF precoding to maximize the system’s throughput is combined with suitable user scheduling (user selection).

It is well-known that in the ZF beamforming scheme the total number of receive antennas (for the users being served at a time) must be less than or equal to the number of transmit antennas at the base station. Although this is a limitation on the scheme, user selection can be used to take advantage of multiuser diversity. When the number of users is large, the base station can schedule its transmission to those users with proper channel conditions to improve the system throughput. This is why simple suboptimal schemes like orthogonal random beamforming or ZF beamforming are performing fairly well with a large number of users. In fact, it has been proven that ZF beamforming with a semi-orthogonal user selection asymptotically achieves maximum throughput of optimal DPC.

In particular we use an eigenvalue waterfilling method to maximize throughput of ZF beamforming. Thus, the eigenvalues of the block diagonal sub-matrix of the aggregate users’ channel are playing the main role instead of the diagonal elements, as in the single-antenna receiver case. By selecting the proper users the throughput can be maximized. The semi-orthogonal user selection algorithm is extended to the multiple-antenna receiver case. Indeed the semi-orthogonal user selection algorithm is a special case of the algorithm proposed, with the eigenvalues the same as the diagonal elements for the single antenna receivers.

More Information