IEEE 802.11 LAN capacity : incentives and incentive learning

For an ad hoc IEEE 802.11 WLAN we investigate how stations' noncooperative behavior in the form of a backoff attack (configuring small minimum and/or maximum CSMA/CA contention windows at some stations in pursuit of a large goodput share) affects a proposed realistic network capacity measure. We argue that such a measure should reflect both the total goodput and fairness, to factor in contention overhead and favor sustainable, i.e., fair, scenarios where each station pursues a large goodput share. Therefore, we consider a capacity-fairness index (CFI), the product of total goodput and the Jain index of the stations' goodput shares. If the backoff mechanism is mandatory, the CSMA/CA game that arises has a unique Nash equilibrium. In the opposite case there is no compelling outcome; we envisage that a station then calculates backoff attack incentives to predict imminent play. We link CFI to the network size, "power awareness", a stations perception of the other stations' susceptibility to incentives, and the way of learning how the stations perceive the other stations' susceptibility to incentives. We show that noncooperative behavior reduces CFI more than does contention overhead, yet if the stations are few and "power aware" enough then its impact may be lessened.