Mobility and Spatial-Temporal Traffic Prediction In Wireless Networks Using Markov Renewal Theory
Abu Ghazaleh, Haitham
MetadataShow full item record
An understanding of network traffic behavior is essential in the evolution of today's wireless networks, and thus leads to a more efficient planning and management of the network's scarce bandwidth resources. Prior reservation of radio resources at the future locations of a user's mobile travel path can assist with optimizing the allocation of the network's limited resources. Such actions are intended to support the network with sustaining a desirable Quality-of-Service (QoS) level. To help ensure the availability of the network services to its users at anywhere and anytime, there is the need to predict when and where a user will demand any network usage. In this thesis, the mobility behavior of the wireless users are modeled as a Markov renewal process for predicting the likelihoods of the next-cell transition. The model also includes anticipating the duration between the transitions for an arbitrary user in a wireless network. The proposed prediction technique is further extended to compute the likelihoods of a user being in a particular state after $N$ transitions. This technique can also be applied for estimating the future spatial-temporal traffic load and activity at each location in a network's coverage area. The proposed prediction method is evaluated using some real traffic data to illustrate how it can lead to a significant improvement over some of the conventional methods. The work considers both the cases of mobile users with homogeneous applications (e.g. voice calls) and data connectivity with varying data loads being transferred between the different locations.