Delay-aware Predictive Network Selection in data offloading

We tackle the network selection problem in operator-initiated mobile data offloading by considering both the predictive and non-predictive cases. For the non-predictive case, we use the Lyapunov optimization technique to design the Delay-aware Network Selection Algorithm (DNSA), which yields an operation cost within O (1/V) bound of the minimum value, for any V > 0. For the predictive case, we propose a novel approach to incorporate prediction into Lyapunov optimization. The proposed Delay-aware Predictive Network Selection Algorithm (DPNSA) efficiently exploits the potential benefit of predicting traffic load and users´ mobilities, and avoids the state space explosion issue due to information prediction. Numerical results show that DPNSA with prediction window size 5 offers roughly 30% delay saving over DNSA.