Optimization of Handover Algorithms for Wireless Networks

Abstract: We are witnessing a continuous development of heterogeneous wireless networks, such as cellular systems (e.g., the LTE, or long term evolution of the third generation wireless system), sensor networks, and satellite networks. The coexistence of all these networks requires the design of control mechanisms to allow the seamless communication from one network to another, and even between communication standards. The handover is one such control mechanism. Specifically, the handover is the mechanism of transferring the connection of a mobile node from an access point (which could be a base station or a network using some communication standard) to another AP. During the handover, a node that is being disconnected from an AP and connected to another AP may experience a sudden degradation of the bandwidth or outage of the communication. This leads to frequent and unnecessary handovers that may reduce significantly the quality of service perceived by the node. Therefore, in order for the node to achieve high quality of service, an optimization of the handover algorithm is essential. In this thesis we study some optimization algorithms for the handover procedure. We show that the variables available at the node to control the handover are the hysteresis margin, which is used to compare the quality of signals of the APs, and the estimation window length, which is the number of samples to estimate with a desired accuracy these signals. Then, we study two optimization problems in which the parameters that affect the handover are optimized by considering as a cost function a convex combination of the probability of outage and the probability of handover. The first problem is based on a static optimization, whereas the second problem on a dynamic optimization. We show that both problems offer better performance with respect to existing algorithms from the literature. We  show that the dynamic optimization gives better results, but at the cost of an increased computational complexity.