Friday 28 February 2025
Scientists have developed a suite of machine learning models that can accurately predict wireless radio metrics in urban environments, paving the way for better network planning and quality of service optimization.
The researchers used crowdsourced data and local environmental features to train their models, which were designed to predict three key metrics: received signal strength indicator (RSSI), reference signal received power (RSRP), and reference signal received quality (RSRQ). These metrics are crucial for assessing the performance of wireless networks in real-world scenarios.
The team’s approach involved refining an initial estimate of RSRP using a path loss model, which takes into account factors such as distance from the transmitter, transmitter height, and obstacles along the signal path. This refined estimate was then used to predict RSSI and RSRQ, which are important indicators of network quality.
To train their models, the researchers combined data from multiple sources, including crowdsourced measurements from mobile devices and open-source mapping data. They also incorporated features such as day of the week, hour of the day, and geographic location to account for temporal and spatial variations in signal strength.
The resulting models were tested on a diverse range of datasets, including urban environments in Canada and Europe. The results show that the models are able to accurately predict RSSI, RSRP, and RSRQ, with root mean squared errors (RMSEs) ranging from 9 to 11 decibels.
One of the key benefits of these models is their ability to generalize across different environments and scenarios. This means that network operators can use them to optimize their networks in a wide range of urban settings, without having to collect data specifically for each location.
The development of these machine learning models has important implications for wireless communication systems. By enabling more accurate predictions of radio metrics, they can help network operators improve the quality and reliability of their services, which is critical for supporting a growing number of devices and applications.
In addition, these models could be used to optimize the deployment of new infrastructure, such as cell towers and antennas. By predicting the performance of different configurations in advance, engineers can design more efficient and effective networks that meet the needs of users.
Overall, this research demonstrates the potential of machine learning to improve our understanding of wireless radio metrics and optimize network performance. As the demand for wireless services continues to grow, these models will play an increasingly important role in ensuring that networks are able to keep up with changing demands.
Cite this article: “Machine Learning Models Predict Wireless Radio Metrics with High Accuracy”, The Science Archive, 2025.
Machine Learning, Wireless Radio Metrics, Urban Environments, Network Planning, Quality Of Service Optimization, Received Signal Strength Indicator, Reference Signal Received Power, Reference Signal Received Quality, Path Loss Model, Crowdsourced Data.
