Recently, it has been shown that a simple, distributed CSMA algorithm is throughput-optimal for wireless networks. However, throughput-optimality is established under some unrealistic assumptions such as (i) perfect collision avoidance and (ii) ideal carrier sensing. In this talk, I will introduce our novel model for a class of CSMA algorithms that address these two assumptions, and some results on their throughput performance.

Specifically, in the first part of this talk, I will introduce a distributed throughput-optimal CSMA for wireless networks, which is called Preemptive CSMA. Distinguished from other throughput-optimal CSMAs in the literature, it achieves the optimality even with the throughput loss caused by discrete backoff time, non-zero carrier sense delay, and data packet collisions.

In the second part, the achievable throughput of Preemptive CSMA will be investigated under imperfect carrier sensing. Through the analysis on both false positive and negative carrier sensing failures, it will be shown that Preemptive CSMA can achieve an arbitrary fraction of the capacity region if certain access probabilities are set appropriately.

Although our main results are still based on some other idealized assumptions, I will also discuss the practical implication of the results and potential extension.