Design and Development of a Handoff Management System in LTE Networks using Predictive Modelling

Handoff management is a key issue in mobile network to provide an efficient and low-cost service. Providing seamless connectivity in high speed data networks is a challenge due to user mobility and varying user traffic patterns. The major challenges with high user mobility are (1) Frequent handoff which causes call dropping (2) Unavailability of resources in the target base station which also may lead to call dropping. If the user behavior can be predicted, the network may proactively start the handoff procedure and thereby minimize the handoff delay, reduce call dropping and effectively manage network resources. In this paper, a Handoff management system is designed and developed for LTE (Long Term Evolution) Networks using Predictive Modelling. The existing handoff procedure specified by 3GPP (3 Generation Partnership Project) is modified to include a prediction part which will significantly reduce handoff overhead time, call dropping and utilize channel allocation. A Multilayer Feed forward Neural Network (MFNN) model is designed and developed as a Predictive Model whose prediction is based on the history of mobility pattern of a mobile user recorded over several days. The MFNN is trained with the data obtained from the mobility pattern of a mobile user for making predictions. To test its functionality, this MFNN model is integrated with the LTE prototype code on ns2 (network simulator 2) by creating user movement scenarios. The performance of the developed LTE Handoff Management System has been verified for prediction accuracy by considering mobility patterns of mobile user. The system simulation has achieved an average prediction accuracy of 93%. Handoff time overhead has reduced to 1 second which is significant as the current handoff process takes 6 seconds. The realized reduction in handoff time overhead has enhanced the Quality of Service (QoS).