Joint channel probing and proportional fair scheduling in wireless networks

The design of a scheduling scheme is crucial for the efficiency and user-fairness of wireless networks. Assuming that the channel quality information (CQI) of all users is available to a central controller, a simple scheme which maximizes the sum-log utility function has been shown to guarantee proportional fairness. This work studies a more general problem which takes both the CQI acquisition and the user scheduling into account. First, in case the statistics of the channel quality is available to the controller, a joint channel probing and proportional fair scheduling scheme is developed based on the optimal stopping time theory. The convergence and optimality of the scheme is proved. Next, the problem is further studied in the case where the channel statistics are not available to the controller, and a joint learning, probing and scheduling scheme is designed by solving a generalized bandit problem. Furthermore, it is shown that the multiuser diversity gain does not always increase as the number of users increases. Numerical results demonstrate that the proposed scheduling schemes can provide significant gain over existing schemes.