Medium Access DiversityEdit
Medium Access Diversity is a technique intended to utilize higher data transfer rates during better transmission conditions. As quoted by the authors of the paper
"...has encouraged the development of multi-rate adaptors, where the PHY layer data rate can effectively respond to wide variations in channel conditions; the difference between the lowest and highest data rates is expected to widen even further with emerging PHY layer technologies. This is evident by comparison of IEEE 802.11a with legacy 802.11 standard: the lowest data rate increased by a factor of 6 while the highest rate increased by a factor of 27."
Medium Access Diversity (or MAD, as it will be referred to henceforth), tries to exploit time and space varying channels at the MAC layer. It seems to achieve high throughput rate improvements too.
"The simulation results in various wireless LAN scenarios show that MAD achieves 120%∼452% throughput improvement over ARF, the rate adaptation scheme used in Lucent’sWaveLAN II networking devices, and on average more than 50% throughput improvement over the best known rate adaptation scheme, OAR."
How it WorksEdit
Any wireless network consists of a sender and receivers. When the instantaneous channel conditions between the sender and every receiver are known, the sender can exploit multiuser diversity by increasing the transmission rate when the channel conditions are favourable. Also, instead of continuing retransmission to a receiver when the channel conditions are poor, the sender can schedule data transmissions to another receiver.
Thus, in short, with complete knowledge of channel conditions, a sender can select a receiver with the best channel condition, and also decide the highest feasible data rate for the receiver it selects for transmission.
There are, however, some caveats to be kept in mind regarding this scheduling method.
- Every time the sender queries several receivers, it incurs a data overhead. So it must limit the number of receivers it queries to a subset (i.e. if there are 300 receivers, it can query; say; 30 at a time regarding their condition.) Even with a limited subset, it must be able to improve overall performance
- It must maximise the effective throughput for each data transmission. This can compensate for increased overheads for querying
- It must balance the need for improvement of throughput with maintaining fairness expectation among traffic flows.
IEEE 802.11 uses Request To Send (RTS) and Clear To Send (CTS) signals before actual data transmission. (More information about these can be found at http://en.wikipedia.org/wiki/IEEE_802.11_RTS/CTS) The MAD method simply enhances these signals to obtain feedback about the channel conditions.
New RTS = GRTSEdit
Instead of an RTS, MAD uses something called a Group RTS (GRTS) control packet. A GRTS control packet contains a list of RAs (Receiver addresses) about which the sender wishes to receive feedback. The rank of a particular RA in the RA list indicates the order in which the receiver should respond to the GRTS.
New CTS = CTS with FeedbackEdit
The CTS control signal is also modified by including an additional two octets labelled Feedback. This consists of the Rate and Gain subfields; the rate, consisting of 4 bits, and the gain, consisting of 12 bits. The receiver measures the SNR of the GRTS packet, and selects an appropriate transmission rate such that the bit error rate is less than 10^-5 for the measured SNR of the GRTS packet. The 12-bit Gain field stores the current of the the receiver relative to its average value.
For any receiver i, the new average value of the SNR can be estimated by
SNR(avg) = 0.8*SNR(avg) + 0.2*SNR (current)