Movement Prediction Assisted Fast Handovers for Seamless IP Mobility

Abstract

The use of cross layer communication in the form of link-layer triggers is a growing trend in mobility management schemes because traditional movement detection techniques are no longer suitable for fast handover protocols. Achieving seamless mobility is a major challenge for current and future wireless networks and is the focus of much research. This paper presents a prediction assisted fast handover protocol (PA-FMIP) that uses a data mining based prediction algorithm to replace the need for a pre-trigger. It is a continuation of previous work that aims to improve the packet loss problem inherent in reactive handovers. We further evaluate PA-FMIP through simulation and focus on the capabilities of the mobility prediction algorithm, particularly when faced with random mobility. The impact of a PA-FMIP handover for real-time and TCP application performance is also thoroughly evaluated. It is found that the data mining based prediction algorithm can achieve 100% accuracy for regular mobility. It is also found that the type of TCP used by the user significantly influences throughput. The improvement in the number of lost packets by PA-FMIP increases the overall throughput for any TCP variant.