Macrodiversity system is a communication architecture where base stations (BS) act as distributed nodes of a multiple-input multiple-output (MIMO) antennas. It has many promising features that can improve system performance from a network perspective, such as improving the weak signals of users affected by shadow fading, or users at the cell-edge.
They also allow multiple users to share the same resource in time and frequency, improving the overall user capacity.

Traditionally, evaluating the link quality of resource-sharing users requires instantaneous channel state information (CSI). However, finding compatible users to share resource in macrodiversity systems is a challenging task. For macrodiversity systems, instantaneous CSI could be passed to the backhaul processing unit (BPU) through the network backhaul. This creates a delay in the signal, and makes instantaneous CSI a less accurate reflection of the channel environment at the time. Passing instantaneous CSI of all users also creates a significant amount of network overheads, reducing the overall efficiency of the network. Compared to MIMO systems with co-located antennas, macrodiversity systems cover a larger geographical area and more users. For this reason, the number of user selection combinations can become extremely large, making scheduling decisions in real time an even more challenging task. These problems limit the realisation of the user capacity potential of macrodiversity systems.

This thesis presents a low complexity method of resource allocation for up-link macrodiversity systems. In particular, it uses long-term power to estimate the link quality of resource-sharing users. Using long-term power bypasses the issue of channel estimation error introduced by the network delay, and it also reduces the communication overhead on the network backhaul. In this thesis, we use Symbol-Error Rate (SER) as the measure for link quality. Using the method developed by Basnayaka [1], we are able to estimate SER of resource-sharing users using long-term power. Using the SER estimation method, we further proposed a user compatibility check (UCC), which evaluates the compatibility of users sharing the same resource. Users are only considered compatible with each other if all of them meet a pre-defined SER threshold.

We attempt to reduce the complexity of user selection by using heuristic solution-finding methods. In our research, we found that greedy algorithms have the least complexity. We propose four low-complexity user selection algorithms based on a greedy algorithm. These algorithms are simulated under different environment parameters. We evaluate the system performance in terms of utilisation and complexity. Utilisation refers to the percentage of allocated users compared to the theoretical user capacity. Complexity refers to the number of SER calculations required to find a resource allocation solution. From the simulation results, we observed that with the proposed user selection algorithms, we can achieve moderately high utilisation with much lower complexity, compared to finding user selections via an exhaustive search method. Out of the proposed user selection algorithms, the Priority Order with Sequential Removal (PO+SR) and the First-Fit (FF) algorithm have the best overall performance, in terms of the trade-off between utilisation performance, and complexity performance. The final choice of the algorithm will depend on the processing power and the system performance requirement of the macrodiversity system.