Delay management in LTE/5G dynamic spectrum sharing via modified Kalman filter with linear regression

Abstract

The 5G system supports various services such as enhanced mobile broadband (eMBB), massive machine-type communications (mMTC), and ultra-reliable and low-latency communications (URLLC). Depending on characteristics of these services, they require diversified demands such as increased capacity, high data rate, ultra-low latency, and so on. Therefore, a 5G usage scenario adopts the multi-layer frequency approach in order to meet these demands. For instance, the multi-layer frequency approach uses below 2 GHz frequencies to enable a wide coverage area as well as mmWave frequencies to support a high data rate. Since below 2 GHz frequencies are already utilized by the preexisting LTE system, the LTE and 5G systems need to share the limited resource. This study covers dynamic spectrum sharing between the LTE and 5G systems focusing on delay issues that arise with the inter-site base station scenario. First, we formulate the delay issue when dynamic spectrum sharing is adopted in the inter-site base station scenario. Second, we propose the traffic prediction methodology to compensate for the delay via traffic prediction. Third, we propose Modified Kalman Filter with linear regression which is an algorithm used to predict future traffic amount. By using Matlab simulations, we verified that the proposed Modified Kalman Filter is more accurate than moving average (MA) and recurrent neural network (RNN) in terms of root-mean-square error (RMSE). Also, we proved that the proposed traffic prediction methodology mitigates the performance degradation of both LTE and 5G users due to delays.