Grammatical Genetic Programming for Scheduling in Realistic Heterogeneous Communications Networks

Grammatical Genetic Programming is used to evolve Small Cell schedulers on a realistic simulation of a 3GPP 4GLTE heterogeneous cellular network. Extensive analysis of the solution and its performance across a wide range of metrics reveals evolution has uncovered a new human-competitive Small Cell scheduling technique which generalises well across cells of varying sizes. Furthermore, the evolved solution is shown to conform to accepted scheduling frameworks without evolution being explicitly told the form of the desired solution. The evolved solution is shown to out-perform a state-of-of-the-art benchmark.