Rate-percentile-optimal sequential channel sensing and probing in cognitive radio networks under spectrum uncertainty

In this paper, we study the quality-of-service (QoS) support for realtime traffic in cognitive radio (CR) networks when spectrum availability and quality is not known a priori. A resource-constrained CR relies on sequential channel sensing and probing to resolve spectrum uncertainty and search for good transmission opportunities in real time. We are interested in maximizing the effective throughput the CR can achieve with a desired confidence (success probability) under a delay constraint. This quantity can be interpreted as the QoS-compliant capacity of the CR link under the uncertain spectrum environment. The optimization is formulated as a finite-horizon optimal stopping problem under the objective of maximizing a given percentile of the rate of return at the stopping time. This formulation cannot be directly solved by classical optimal stopping theory, because the latter only supports a mean-reward objective function. A novel transformation is developed to convert the problem into solving a series of sub problems, each of which optimizes a transformed mean-reward of the original problem and therefore can be solved using classical optimal stopping method. We prove the monotonicity of the sub problems, based on which we develop a fast algorithm to efficiently find the unique solution to the original problem. Extensive simulations are performed to verify the effectiveness and significance of the optimization. We show that significant gains (e.g., over 30%) on the QoS-compliant capacity can be achieved by the proposed algorithm when compared with the counterparts.