Achievable Throughput of Multiband Wireless LAN using Simultaneous Transmission over Multiple Primary Channels Assisted by Idle Length Prediction Based on PNN

The authors have studied a multiband wireless local area network (MB-WLAN) which can effectively detect and exploit unused radio resources scattered in time and frequency domains. The MB-WLAN sets one or more primary channels (PCHs) in multiple frequency bands, and each station (STA) carries out random back-off process on the multiple primary channels to obtain a transmission opportunity (TXOP). Once a STA obtains a TXOP on any PCH, it checks whether or not another TXOP can be obtained on any other PCH in near future. If the STA judges that it can obtain another TXOP, it pends its transmission until another TXOP is obtained on any other PCH, and then a channel-bonded frame is transmitted. A suitable pending duration depends on the level of congestion on each PCH because the STA lose its TXOP more frequently to other STA’s frame transmission as the PCH gets more crowded. This paper, therefore, proposes a method to control the maximum pending duration with the aid of idle length prediction based on probabilistic neural network (PNN). This paper also proposes a method to control the timing to invoke learning of channel usage for PNN in order to get rid of the impact of self-transmission on the characteristics of channel usage. In order to validate the effectiveness of the proposals, this paper evaluates the achievable throughput of the MB-WLAN by computer simulation assuming IEEE 802.11n/ac-based WLAN operated in the 2.4GHz and 5GHz bands and 4-antenna STA. It is confirmed that the MBWLAN with two proposals can achieve almost best performance regardless the level of congestion on PCHs.