Self-optimization of LTE Mobility State Estimation thresholds

This paper describes an algorithm for self-optimizing Mobility State Estimation (MSE) thresholds in heterogeneous Long Term Evolution (LTE) networks using Minimization of Drive Testing (MDT) measurement traces. The algorithm is using the MDT measurements to construct statistics of reselection and handover distributions for different mobility categories to learn how network topology and UE velocities are correlating in local geographical neighborhood. The distributions are used to self-optimize LTE Release 8 MSE thresholds by employing a standard score linear classifier. This allows simple, backward compatible and cost-efficient optimization of MSE thresholds. Moreover, such self-optimization results in a high classification accuracy of UE mobility states and decreases operator's manual parameter configuration complexity. Performance evaluation of the proposed algorithm was done by conducting extensive system simulations. The performance results indicate that in the studied sparse and dense heterogeneous networks the average classification accuracy was 72.2% and 78%, respectively.