Specifically, we investigate the maximum stable throughput attainable with decentralized control using channel sense multiple access (CSMA), which has not previously been studied with MPR. We show CSMA provides throughput gain over slotted ALOHA (S-ALOHA), the non-channel-sensing protocol of choice. However, we also find that this gain diminishes as the physical layer strength increases, thereby diminishing the need for channel sensing.
Searching for improvements, we investigate the generalized CSMA protocols which allow new transmissions to begin even when the channel is already in use, provided the usage level is below capacity. In addition, we also analyze the effects of implementing collision detection (CD). We find that generalized CSMA provides a moderate improvement in throughput compared to the significant improvement that CD affords. This is attributable to CD's shortening of the time that corrupted transmissions occupy the MPR channel, thus increasing any new transmission's success probability. As generalized CSMA/CD offers little improvement over the simpler classical CSMA/CD, we conclude the latter suffices for satisfactory improvement. We also present a novel CD method applicable not only to wireline but also to wireless networks.
Finally we investigate the general multiaccess protocol's capacity
with MPR. We find that, if the expected number of successes is
maximized in the limit of infinitely many simultaneous transmissions,
then S-ALOHA and CSMA are optimal. We also establish an upper bound on
the throughput of decentralized multiaccess protocols operating on
channels not satisfying this condition. The gap between the channel
capacity and said upper bound is informative about the cost of
decentralization. This gap also narrows as the physical layer's ability
to separate packets gets stronger, thus corroborating existing theories
that suggest there is less need for multiple accessing for strong MPR