Learning-Based Interference Mitigation for Wireless Networks

Abstract

Wireless networks have raised great attention in the past decades because they provide tether-free connectivity. Although much of the effort in wireless network research has been spent on reducing the interference among the communication nodes, the problem remains open. In this dissertation, we propose a learning-based approach to alleviate wireless interference. The principle of the learning-based approach is based on the observation that although wireless networks are usually complex and dynamic, information can still be extracted from the data measured in the past. By learning from what was observed in the past we can select the desired operational parameters, react intelligently, and achieve substantial performance gain. In particular, we, show that interference mitigation can be achieved in three different aspects: (1) collision avoidance, (2) channel rate adaptation, and (3) spatial reuse.