Phil Whiting, Macquarie University, North Ryde, NSW Australia
Heteregeneous networks extend the capacity of existing LTE networks by introducing small cells which are placed so as to offer coverage in traffic hotspots. In this way capacity is gained because the spectrum can be simultaneously used in multiple such small cells.
To make more effective use of the small cells, two additional features have been added. The first is the so called Almost Blanking Subframes (ABS) whereby the supporting Macro cells are ``muted'' so that there is negligible interference to the small cells. The second is that the range of each small cell can be adapted so as to offer service to more or less mobiles and so get a better match to resources.
This talk presents Throughput Stable algorithms which can be used to set the above HetNet parameters against an unknown but in principle supportable offered traffic. Two such algorithms are discussed one which directly adapts both parameters using only the current distiribution of mobiles within the network and given service requirements. The second works by adapting only the ABS fraction on a fast time scale and sets the range according to an auxiliary learning algorithm. For this second algorithm some performance results can be produced based on Processor Sharing constructions originated by Bonald and Massoulie.
Finally we consider the question of how much traffic can be supported if more or less arbitrary schedules are used to control the HetNet parameters. We show that under our model the solution to a ``continuous LP'' determines what the maximum supportable arrival rate can be if the location traffic density is specified.