The objective of this work is to develop ML algorithms for dynamic channel assignment and power control in software defined Wi-Fi APs to increase the reliability of communication in Wi-Fi Networks. Reinforcement Learning (RL) would be employed at the PHY layer to take channel and Tx power decisions to maximize network capacity in SDN based 5GEmpower controlled Wi-Fi Network.
Enabling a wireless network to operate at their best requires channel and Tx power selection in a coordinated manner to control interference and improve reliability of communication. RL has the capability to steer network towards desired performance objectives based on feedback of the actions taken in different states and learning optimal policy. In this project, 5GEmpower controller would be used to learn optimal policy for channel and Tx power assignment to multiple APs in an industrial environment to enable reliable communication between associated users.
The student will use both a state of the art Linux-programmable Wi-Fi access point and some Linux end devices. They will:
a) formulate States, Action and Reward functions for Reinforcement Learning (RL) algorithm
b) develop RL based Channel and Tx Power assignment algorithm for 5GEmpower controller and
c) develop real-time statistics collection for measuring network performance on the go and evaluating the performance of RL algorithm.
The student will work on actual hardware in the lab: they will make use of the very versatile 5GEmpower platform to build a truly programmable Wi-Fi Access Point.
20% Theory, 60% Implementation, 20%Writing
Kamran Zia (email@example.com)