Japan challenge (network failure detection & location estimation) and advances in machine learning for wireless communications
* Register (or log in) to the AI4G Neural Network to add this session to your agenda or watch the replay
This webinar, will introduce 2 problem statements for the challenge hosted in Japan as well as expert talks.
ITU-ML5G-PS-016: Location estimation using RSSI of wireless LAN
The demand for location information is becoming more critical due to the emergence of map applications and augmented reality (AR). GPS is the leading steam of localization; however, its accuracy significantly degrades when the number of satellites that can be seen from the receivers decreases or due to reflection from the structures. As a substitution, localization techniques utilizing the radio signal received from Wi-Fi AP and cellular BS is promising. However, the typical triangulation approach suffers from the multipath fading channel, and hence it cannot achieve high accuracy. This challenge aims to verify if the AI/ML aided localization utilizing RSSI observed at the terminal can achieve similar accuracy as the GPS-based localization. This challenge explores the possibility if the data-oriented localization can replace the model-based location with the help of powerful AI/ML techniques.
As 5G mobile networks are getting to be spread globally, the stable and high-quality operation is a must to minimize the social impact caused by 5G service failure. In conjunction with 5G deployment, not only NFV (network virtualization function) but also CNF (cloud native
function) is being deployed in service provider networks, adding complexity and uncertainty to operational environment. In that situation, network automation is a key to accelerate 5G network penetration, although highly experienced operators can tackle affected network failure and the anomaly detection is additionally desired to be automatically and rapidly performed by AI/ML. In this problem, the data sets in a 5G core network are provided along with network status information such as normal, a failure, mis-operation and so forth, as normal/abnormal labels. Participants are required to create the model to pinpoint the network status of failures and mis-operation using those data sets and evaluate the performance of the developed model.
2 Expert talks: Machine Learning for Wireless LANs
The two expert talk provides applications of deep supervised learning and reinforcement learning for microwave and mmWave wireless LANs (WLANs). Furthermore, the talk will address how to apply the machine learning techniques to challenges in WLANs based on mmWave, received power prediction and handover and advances in Machine Learning research applied to wireless communications.