Predicting Channel Quality Indicators through Deep Learning: A Location-Aware Approach Utilizing Limited Network Parameters

In the dynamic landscape of vehicular communication environments, adequate resource allocation and management are critical for ensuring seamless communication, particularly in scenarios involving constantly moving vehicles. To efficiently allocate resources, the base station requires real-time Channel State Information (CSI) from devices. Consequently, vehicles must continuously report their Channel Quality Indicator (CQI) to the base station. The base station, relying on the reported CQI, determines the Modulation and Coding Scheme (MCS). This study introduces a context-aware CQI prediction framework utilizing bidirectional Long Short-Term Memory (Bi-LSTM) techniques. A distinguishing feature of this framework is the strategic incorporation of location information, precisely latitude and longitude, and two key network parameters, namely Received Signal Strength Indication (RSSI) and Reference Signal Received Power (RSRP), tailored to adhere to IEEE 802.11p Wireless Access in Vehicular Environments (WAVE) standards. The simulation results demonstrate the superiority of the proposed scheme, showcasing lower Mean Squared Error (MSE) when compared to alternative methods and previous studies. This strategic inclusion of geographical data not only highlights the model’s adaptability to spatial dynamics but also positions it as a comprehensive solution for accurate CQI prediction, thereby contributing to the efficiency of resource management in intelligent transportation systems.