Understanding Wireless Interference in the Unlicensed Band

March 12, 2012
Shravan Rayanchu | University of Wisconsin-Madison

Current indoor wireless environments comprise a complex ecosystem hosting a plethora of wireless devices including WiFi-enabled devices, and many other non-WiFi RF devices such as Bluetooth devices, microwave ovens, game controllers, wireless cameras and cordless phones. Such a setting involving interaction between distributed entities (each with a limited view of the spectrum) that often employ incompatible or inefficient protocols, frequently results in severe wireless interference. This interference remains the primary reason why wireless networks, unlike their wired brethren, exhibit unpredictable performance and offer significantly lower throughput. Hence, an important step towards building future wireless systems is to better understand and manage such interference.

In this talk, I will show how we can build systems and tools that help us understand interference in indoor wireless environments. Specifically, I will talk about two systems that we built: Airshark and WiFiNet. Together, these systems can detect, localize, and quantify the interference impact of various non-WiFi interference sources in real-time and using commodity WiFi hardware alone. Airshark is a software solution that can run on top of a commodity WiFi Access Point (AP) and can detect multiple simultaneously active non-WiFi RF devices. It has an average detection accuracy of 91-96% while maintaining a low false positive rate. WiFiNet builds on top of Airshark, and is targeted towards enterprise wireless environments. WiFiNet uses information from multiple WiFi APs running Airshark and can accurately discern an individual non-WiFi device’s interference impact in presence of multiple active non-WiFi devices. Further, it can pin-point the physical location of these non-WiFi interference sources. Our deployment and evaluation of WiFiNet show that interference estimates are within 10% of the ground truth and the median localization error is 4 meters. We believe systems such as Airshark and WiFiNet can empower wireless network administrators with new tools that help diagnose non-WiFi interference issues in enterprise wireless environments.

Speaker Details

Shravan Rayanchu is a PhD candidate in the Computer Sciences department at University of Wisconsin-Madison. His research interests are in networking and distributed systems. His dissertation research focuses on diagnosing performance issues in wireless networks and improving their performance by designing smart control and data plane mechanisms. Shravan has an MS in Computer Science from UW Madison, and a B.Tech. in Computer Science from Indian Institute of Technology (IIT), Guwahati. He is a recipient of Microsoft Research PhD fellowship. He won the best paper award for work on hybrid scheduling at Mobicom 2009, and his work on flexible channelization was nominated for the best paper at MobiCom 2011.