Information in the database is then used to assist either power, frequency, and antenna tilting adaptations of the satellite or in adaptation of the cellular system. Spectrum sharing is expected between cellular and satellite systems, e.g., in the 3.4-3.8 GHz band and in the millimeter wave bands such as the 24.25-27.5 GHz band. There definitely are bands and areas where dynamic time domain allocations and prediction should be applied. Machine learning and especially deep learning techniques applied in long term spectrum analysis will help in revealing patterns in the spectrum use and making justified decisions both by the regulator and in the operative network management. Proposals in Table 2 Location Speed Direction Used Channel/Frequency Application Requirements Channel Quality Indicator (CQI) provide some ideas for further research and developments in this area. Predictive Local-Aware Spectrum Allocation for the V2X Use Case The concept for a single car supporting automated driving in a heterogeneous network is shown in Fig. 4. The database-assisted system tells in advance which base stations (BS) or access points (AP) to connect to at which frequency giving also transmission power limits for the operation. Due to increasing number of wireless devices, the optimization procedure will become quite complicated and prioritizations are needed to guarantee BS/AP to Connect Frequency to Use Power Limit Spectrum Database (a) Predicted CSI Values for a 10 km/h User 15 Error Due to Erroneous Prediction for 60 km/h User 2 1.9 1.8 10 RMSE 1.7 1.6 1.5 5 1.4 Real Predicted 0 0 100 200 Time (ms) (b) 300 400 1.3 0 100 200 300 400 Training Window (ms) (c) 500 Figure 4. Predictive local-aware spectrum allocations with mobility management. (a) shows a database management system for mobility prediction. (b) shows how the local channels can be predicted with moving users. (c) reveals that there is a CSI prediction error due to fast changes in channel behaviour. 42 IEEE CIRCUITS AND SYSTEMS MAGAZINE THIRD QUARTER 2019