User-Network Association in a WLAN-UMTS Hybrid Cell: Global & Individual Optimality

TL;DR

This paper investigates optimal user-network association in a hybrid WLAN-UMTS cell using global and individual criteria, employing stochastic decision models and game theory to derive optimal policies.

Contribution

It introduces a novel formulation of user association as a semi-Markov decision process and analyzes individual optimality through a non-cooperative game framework.

Findings

Global optimal policy lacks convex or concave switching curves.

Threshold and symmetric switching curves observed in homogeneous cases.

Individual optimality results in a staircase threshold policy with increasing mobile arrivals.

Abstract

We study optimal user-network association in an integrated 802.11 WLAN and 3G-UMTS hybrid cell. Assuming saturated resource allocation on the downlink of WLAN and UMTS networks and a single QoS class of mobiles arriving at an average location in the hybrid cell, we formulate the problem with two different approaches: Global and Individual optimality. The Globally optimal association is formulated as an SMDP (Semi Markov Decision Process) connection routing decision problem where rewards comprise a financial gain component and an aggregate network throughput component. The corresponding Dynamic Programming equations are solved using Value Iteration method and a stationary optimal policy with neither convex nor concave type switching curve structure is obtained. Threshold type and symmetric switching curves are observed for the analogous homogenous network cases. The Individual optimality is studied under a non-cooperative dynamic game framework with expected service time of a mobile as the decision cost criteria. It is shown that individual optimality in a WLAN-UMTS hybrid cell, results in a threshold policy curve of descending staircase form with increasing Poisson arrival rate of mobiles.