Risk-Averse Traffic Allocation for Multi-RAT Connectivity in HetNets

We consider the problem of multiple-radio-access technology (multi-RAT) connectivity in heterogeneous networks. Recently, multi-RAT connectivity has received significant attention both from industry and academia because of its potential as a solution to increase throughput, enhance communication reliability, and to minimize communication latency. In this paper, we propose a new risk-averse-traffic-allocation scheme that allows trading off expected throughput for risk measured in throughput semivariance. Semivariance is a mathematical quantity used originally in finance and economics to measure the dispersion of a portfolio return below a risk-aversion benchmark. We adopt semivariance as a measure of throughput dispersion below a risk-aversion-throughput benchmark. We formulate the multi-RAT connectivity problem as a semivariance-optimization problem. There are computational issues associated with semivariance optimization. However, there are approximations to the semivariance that overcome these issues. We study these approximations and develop our algorithm based on one of them. We evaluate the performance of our algorithm through simulation of a system with three radio-access technologies, namely, 4G LTE, 5G NR, and WiFi. Simulation results show the potential gains of using our algorithm.