| Title | : | Bad Wi-Fi Killer: Robust, Self-Healing, Intelligent and Data Driven Next Generation Wi-Fi Networks with Machine Learning techniques |
| Speaker | : | Kavin Kumar T (IITM) |
| Details | : | Mon, 3 Jul, 2023 3:00 PM @ MR - I (SSB 233) |
| Abstract: | : | Wireless Local Area Networks (WLANs), based on IEEE 802.11 standards (also called Wi-Fi), become an integral part of our day-to-day life. Wi-Fi networks enable end-user computing devices such as desktops, laptops, tablets, smartphones, and television units to connect to the Internet. At the same time, Machine Learning (ML) techniques become an essential and sometimes de facto tool for many real-world applications such as computer vision, Natural Language Processing (NLP), and speech processing. Both industry and academia have also considered ML techniques for computer networking systems. The proposed Ph.D. research is in the theme of Bad Wi-Fi Killer: Robust, Self-Healing, Intelligent and Data-Driven Wi-Fi Networks. The main objective of the research work is to address the so-called "Bad Wi-Fi problem" in the next-generation Wi-Fi networks using machine learning techniques. These problems include but are not limited to; i) Users are typically unaware of the location of best Wi-Fi zones in the home/enterprise Wi-Fi network, ii) Users are also unaware of why their Wi-Fi network suddenly becomes slow and which of their applications are more bandwidth-hungry. iii) Users are unaware that Wi-Fi Mesh routers must be appropriately placed for the best performance. iv) Users typically have no idea about their home Wi-Fi environment, such as interference from nearby devices operating at the same frequency. At a high level, the Bad Wi-Fi Killer proposal considers the home's virtual deployment area or floor plan as a pre-requisite. Once the Wi-Fi mesh APs, clients and IoT devices are in place and operational, we collect real-time metrics. This data includes Wi-Fi traffic pattern, L1/L2 metrics, timestamp, approximate device position, and client static or moving. Once the required data are collected, either an on-device or offline ML techniques are applied to determine the weights on each zone. These weights will result in patterns or insights to solve the bad Wi-Fi problems. With minimal customisation, this research will be expanded to more extensive Wi-Fi networks, such as enterprise and campus. |
