Automatic cell planning for data networks is still an open research subject. Its unique characteristics justify the efforts to extend current voice models to obtain new models. We have found few models oriented to the design of multiuser wireless data networks based on WiMAX like technologies. A complete solution that considers traffic models, link channel quality, sectorized antennas, channel allocation, and variable transmission profile, as far as we know, has not been found.
Right now we are considering a medium complex traffic model. There are several things to do to complete this model. Even though granted services require a fixed data rate, there might be some multiplexing gain based on statistical usage of this flows by end users. In voice model, there is an extension based on silence suppression which reduces capacity requirements for the system. In other types of flows, there are more realistic traffic models based on variable bit rate sources, which are harder to include. There is a need to include models that are nearer the operation scheduler in the design process. Finally, in BE traffic, there is a serious problem with the uplink scheduling. This model is more suitable for downlink transmissions, where the base station assigns transmission opportunities to contending flows. In the uplink, the backoff process defined in IEEE 802.16 standard requires a low usage of the available capacity for it to be effective. It is known that many users contending for the channel simultaneously would cause a low equivalent data rate per user. This model is not included in this work. We need to keep an equilibrium between precision and complexity. As this is a design process, there would be a prohibitive computational cost if the traffic model is far more complicated.
Another issue to be extended, is that users could dynamically change the base station which they connect to. Temporarily, during congestion periods of a base station, a node could connect through another base station which is not optimal in terms of propagation but that has available capacity. Our model only considers static assignment of users to base stations.
Another extension of our model is based on uplink channels estimation. We do not consider this in our model because of its complexity. Usually, uplink data rate requirements are lower than downlink ones, which gives a margin for uplink link budget. We consider symmetric channels for propagation between a user and a base station. Under this consideration, uplink channel quality should be similar to downlink quality. The issue we find difficult to include is the interference. The estimation of uplink received power can be easily done. When we consider interference signals at the base station, we have to consider transmission power of other base stations and other users. The interference from other base stations can be easily considered, but we have no method to estimate the interference caused by other users because we do not know a priori the base station they are connected to, their transmission channel, nor their antennas orientation. We are working on better ways to improve this part of the model.
We are working right now on simplifications of this model to account mathematically tractable solutions. We are considering an optimization problem based on activation and deactivation of candidate base stations, but also to include power control and data traffic models. Variable transmission profile based on adaptive modulation and coding is a really complex part, because of discontinuity. We are looking forward to find the simplifications needed to write it as a mathematical model and compare it with this solution.
