| Summary |
In a mesh network using IEEE 802.11 radio, each node can use multiple transmission rates defined by the combination of modulation schemes and coding rates.
By selecting the transmission rate according to the communication environment, the throughput can be maximized.
The conventional algorithm aims at maximizing throughput by referring to Received Signal Strength Indicator (RSSI), a transmission success rate, and an acknowledgment frame.
However, realizing maximum throughput is difficult because of various factors such as one-to-one correspondence between RSSI and throughput, unnecessary transmission rate control due to a hidden terminal problem, and the ability to acquire more than one RSSI, for multiple antennas.
In this study, we treat the RSSI and transmission rate collected by experiment as a data set, and use machine learning to maximize throughput by selecting the optimal transmission rate for each node. Training is carried out by using support vector machine (SVM) and k-neighbor method (kNN) about the created data set. From a result of testing, it is reported that the value of 83.2% of the optimum value of throughput can be obtained at the maximum. |
