PA-FMIP: a Mobility Prediction Assisted Fast Handover Protocol

This paper introduces a prediction assisted fast handover protocol (PA-FMIP) that integrates a mobility prediction algorithm into a fast reactive handover scheme. A simple data mining approach is used to implement the prediction algorithm based solely on the user's mobility history between wireless IP subnets. Through simulation, the accuracy of the algorithm is evaluated over a city section of West University Area in Texas by a single user traveling with vehicular mobility. The optimal prediction parameters are found for an average accuracy of 92.5%. The handover performance of the PA-FMIP scheme is compared to MIPv6, proactive FMIPv6, reactive FMIPv6 and simultaneous bindings. Simulation results show a 43% decrease in handover packet loss over reactive FMIPv6 and an improvement in TCP throughput. The overhead incurred as a result of the packet forwarding is 50% less than that of Neighbor Casting in the same scenario