Efficient Distributed Channel Probing for Opportunistic Scheduling Over Wireless Fading Channels

It is energy-consuming and operationally cumbersome for all users to continuously estimate the channel quality before each transmission decision in opportunistic scheduling over wireless fading channels. This observation motivates us to understand whether and how opportunistic gains can still be achieved with significant reductions in channel probing requirements and without centralized coordination amongst the competing users. To that end, we first study a basic scenario with symmetric arrivals to explicitly characterize the maximum achievable throughput as a function of the allowable probing rates in symmetric and independent ON-OFF fading channels. This result indicates that a natural randomized strategy for distributed implementation cannot exploit the full opportunistic gains. This finding inspires us to design an optimal centralized joint probing and transmission algorithm under the probing constraints. Noting the difficulties in the implementation of the centralized solution, we develop a novel Sequential Greedy Probing (SGP) algorithm by using the maximum-minimums identity, which is naturally well-suited for physical implementation and distributed operation. We show that the SGP algorithm is optimal in the important scenario of symmetric and independent ON-OFF fading channels. Then, we study a variant of the SGP algorithm in general fading channels to obtain its efficiency ratio as an explicit function of the channel statistics and rates, and note its tightness in the symmetric and independent ON-OFF fading scenario. We further expand on the distributed implementation of these greedy solutions by using the Fast-CSMA technique.