Perfect versus imperfect monitoring in multi-agent opportunistic channel access

This paper investigates the impact of imperfect information on the performance of secondary users (SU) attempting to opportunistically exploit spectrum resources in a distributed manner. We design two channel selection strategies that leverage different levels of information about past channel activity. In the case of perfect monitoring we assume each SU is able to distinguish signals transmitted by a primary user (PU) from signals transmitted by an SU and collision events. In the case of imperfect monitoring we assume SUs to be unable of discriminating between SU or PU activity in the channel. We formally prove that both strategies converge to a static orthogonal channel allocation. Our simulation results show that the impact of imperfect information is to reduce the rate of convergence; the performance loss depends on the level of PU activity and on the amount of competition between SUs.