Multiuser cognitive access of continuous time Markov channels: Maximum throughput and effective bandwidth regions

The problem of sharing multiple channels owned by primary users with multiple cognitive users is considered. Each primary user transmits on its dedicated channel, and its occupancy is modeled by a continuous time Markov process. Each cognitive user is capable of sensing one channel at a time and it transmits according to a slotted structure. The transmissions of cognitive users on each channel are subject to a prescribed collision constraint. Under tight collision constraints, the maximum throughput region is obtained by a policy referred to as Orthogonalized Periodic Sensing with Memoryless Access (OPS-MA). Characterizations of the maximum throughput region are also provided when the collision constraints are loose. It is shown that the OPS-MA policy achieves the maximum sum-rate under all collision constraints when the number of cognitive users equals to that of the primary users. Inner and outer bounds for the effective bandwidth region are formulated as a pair of convex optimizations. When there are only two channels, corner points (the single user scenario) of the optimal effective bandwidth region are also obtained.