Configuring Dynamic Heterogeneous Wireless Communications Networks using a Customised Genetic Algorithm

Wireless traffic is growing at an exponential rate due to the prevalence of smart devices, surging demand for multimedia content and the advent of the "Internet of Things". To satisfy demand, network operators are deploying Small Cells alongside their traditional Macro Cells. The resulting Heterogeneous Networks (HetNets) are highly spectrally efficient because both cell tiers transmit across the same scarce and expensive bandwidth. However, load balancing and cross-tier interference issues arise in co-channel operation. Capacity can be increased by intelligently configuring Small Cell powers and biases, and setting the muting cycles of Macro Cells. This paper develops a customised Genetic Algorithm that can reconfigure an entire network within O(10) minutes. Thus, the proposed evolutionary framework yields tailored solutions in real time. The GA boosts cell-edge (5th %-tile) rates by 28.4% versus static baseline settings that are used in practice. HetNets are highly dynamic environments. However, wireless traffic is cyclical since hotspots arise at predictable locations during busy periods. An explicit memory of previously effective solutions is maintained and used to seed runs. Simulations show that 5th %-tile rates are improved by a further 2.4% to 30.8% over baseline when prior knowledge is incorporated.