Partially Blind Handovers for mm Wave New Radio

Partially Blind Handovers for mm. Wave New Radio Aided by Sub-6 GHz LTE Signaling Faris B. Mismar and Brian L. Evans faris. mismar@utexas. edu and bevans@ece. utexas. edu I MOTIVATION III MODEL & HANDOVER PROCEDURE * SON: Partially Blind Handovers Self-Organizing Network Algorithm: Partially Blind Handover Success Estimation § 5 G New Radio (NR) promises very high data rates § mm. Wave has limited coverage radius relative to sub-6 GHz § A failed handover to NR lowers data rates & customer satisfaction § LTE measurement gaps: • LTE e. NB configures it for a UE to measure another tech • Data transmission and reception seize during gaps §Goal Predict the handover success instead of opening a gap 1) Obtain radio measurement and handover data Xi for UE i for all time T. 2) Generate the supervisory labels based on number of executed handovers. 3) Split the data 4) Train the XGBoost classifier and optimize hyperparameters 5) Obtain the area under the ROC curve using the test data. § Execute a handover from LTE to NR only if likely to succeed. 6) If area Approach § Collect data from users (UEs) who performed measured handovers from LTE to NR over a certain period T Use the handover estimates B. Optimize these steps Proposed § Use data to learn two-classifier: will handover to NR be executed successfully (y = 1) or will it fail (y = 0)? PARAMETERS fc Bandwidth Tx Power 2. 1, 28 GHz 20, 100 MHz 46 d. Bm Antenna pattern Prop. model Tx Ant. Height Rx Ant. Height Synth. Omni COST 231, [8] 20 m 1. 5 m to send 7) Else: Use the baseline algorithm. 8) Repeat procedure for all UEs. II Radio Network § Single co-located macro cell with single antenna. § UEs are scattered per a Poisson point Radio Network Parameters (LTE, NR) Cell radius 350 m Geometry Circular process. then: IV RESULTS Machine Learning § XGBoost Classifier: Machine Learning Hyperparameters Inter-RAT Success Rates CONCLUSIONS V Machine Learning Features § Improved inter-RAT handover success rate § Machine learning classifier was used: • • Reporting coordinates requires a modification to the standards body to enable UEs to report them over RRC. They can be obtained by GNSS, GPS, OTDOA, etc. May, 2018 It learned to predict the inter-RAT handover success for served UEs It used both sub-6 GHz and mm. Wave prior measurements. Cross validation is required to perform a grid search in the hyperparameter space. Personal homepage https: //www. linkedin. com/in/farismismar/