Your smartphone or tablet's Wi-Fi radio lives in the moment. When it's time to connect to a hot spot, all other things being equal, it will attach to the radio that's the best for it at that instant: the one with the most attractive combination of signal strength and throughput. But once your device is on the move, that strategy is far from optimal.
A group at MIT is developing technology that takes actual and predicted device movement into account when connecting to hotspots, to increase overall wireless performance on mobile devices. See PDF: Improving Wireless Network Performance Using Sensor Hints.
MIT professor Hari Balakrishnan explains that his team's software uses sensors on a device that are otherwise unused when a Wi-Fi radio is looking for a connection: the GPS sensor and logs, acceleration sensors, even the compass. From this information, the device can tell not just where it is, but where it's going, and it can then attach to a hotspot that it should be able to stick with for a bit longer than if it's just picking one that's good at the moment. Devices can also proactively select and modify radio data rates based on predicted movements.
Reducing the number of handoffs between hotspots increases connection quality, since each handoff takes time and can interrupt the quality of online activities like audio calls or video streaming.
Balakrishnan says his testing shows an improvement in overall network performance from 40 to 60% for moving devices, and a reduction in hotspot handoffs of about 40 percent. The team is working on Wi-Fi, but he believes that the technology could also help improve cellular connections.
I asked Balakrishnan about the current news around Apple's iPhone keeping location data in memory. Could this information be used in association with his technology, to improve network performance of iOS devices? Balakrishnan say yes. But his team has examined the information the iPhone is recording and says that it appears that it's only used to improve the phone's speed at getting a location fix; it's not being used for WiFi performance improvement. (Watch Reporters' Roundtable on Friday, April 29, at noon Pacific Time for more about mobile phone location recording.)
Balakrishnan is developing a handset version of the technology for Android phones, which is fairly easy due to the open nature of the platform. He said an iPhone version is also possible, but it would require users to jailbreak their devices. Ultimately the technology would work better if it was baked into network protocols and also programmed into Wi-Fi routers. Balakrishnan has not made plans to commercialize this work yet. Some of his previous research (including an indoor geolocation technology) has become commercial. Other technology he's worked on ended up in the open source community, he tells me.
The idea of using predicted device movement to help a Wi-Fi radio choose which hot spot it should connect to seems fairly obvious, and the claimed improvements in throughput bear out that we're leaving performance on the table, as it were, by not using such a scheme.