Automatic Performance Assessment — Step Toward Autonomous Mobile Network Management Systems

Despite significant advances in telecommunications technologies, as well as the use of artificial intelligence in 5G and the anticipated reliance of 6G on trustworthy artificial intelligence, there are still insufficiently automated, inefficient, and unreliable areas of wireless network operation and maintenance that rely on human expert decisions. In this article, we point out the gap in automated performance monitoring and prediction as a factor preventing the design of autonomous mobile network management systems. In addition, we systematize the basics of modeling KPIs measured in a mobile network and conditioned by their configuration and environmental states. KPI modeling recommendations summarized for different network levels were demonstrated for two applications using datasets collected from operator networks, that is, quantifying the performance change in a population of 350 5G cells with unique configuration settings updates, and predicting the performance profile for an unknown configuration in a model trained with historical data associated with different configurations applied to a population of network cells.