Resource allocation for block-based multi-carrier systems considering QoS requirements

Future 5G and beyond mobile networks target at services with a high degree of heterogeneity in terms of their communication requirements. To meet these requirements, different PHY numerologies would provide a better performance; still, all services must be served by a single network technology. Generalized Frequency Division Multiplexing (GFDM) is a good candidate for PHY virtualization where the dimensions of the data block can be dynamically configured in time and frequency. Allocating these blocks in a common spectrum every scheduling period leads to a ""packing"" problem, in which the QoS demands of the data flows need to be acknowledged. In this paper we consider the optimization of the data block allocation as a Knapsack problem. We incorporate the flows' QoS demands by means of utility theory, where utility functions provide a metric of urgency for a flow to be scheduled and the data block to be allocated. For the resulting two-dimensional geometric Knapsack problem we propose a heuristic solution, assess different design options and evaluate the performance in terms of data rate and queuing delay.