Feedback and Learning in Cognitive Radio Systems

Cognitive radio is the key enabling technology for future generations of wireless systems that address critical challenges in spectrum efficiency, interference management, and coexistence of heterogeneous networks. In this project, we aim to establish fundamental structures and performance limits of cognitive radio systems.

Our technical approach rests on a systematic exploration of the role of feedback and learning through stochastic control. We focus on three types of learning crucial to cognitive adaptation in a complex and dynamic communication environment: learning from observation history, learning from mistakes, and learning from properties of co-existing heterogeneous network components. Feedback, indispensable to any learning mechanism, is explored jointly. How to introduce feedback, what and how much information to be exchanged are issues under investigation.

Key innovations of this research include exploiting the heavy tail and self similar nature of primary traffic processes in the design of cognitive radio systems. Our approach uses theories and techniques in traffic modeling, distributed statistical inference and consensus learning, and dynamic optimization and stochastic control.