FRAMINGHAM, Mass. — Wireless networks can serve mobile devices up to 76 per cent faster through the novel use of accelerometers, GPS locators, gyroscopes and compasses that come standard on iPhones, iPads and other smartphones and tablets, researchers say.
With the proper rate-adaptation protocols in place on these mobile devices, performance can be boosted by overcoming some of the problems inherent in networks where the end devices alternately move around and sit still, according to a paper scheduled for the 8th USENIX Symposium on Network Systems Design and Implementation scheduled March 31 in Boston.
Hints provided by the built-in sensors can improve the performance of wireless network protocols, says a team of researchers at MIT’s Computer Science and Artificial Intelligence Laboratory.
These protocols, including one developed by the MIT team, take into account whether the device is standing still, how fast it’s moving and which direction it’s headed with the aim of optimizing performance.
“Protocols can explicitly adapt their behavior and parameters to the current mobility mode,” the team says in “Improving Wireless Network Performance Using Sensor Hints.”
The research suggests two ways to improve performance: maximizing the bit rate at which protocols send data and minimizing the number of handoffs between Wi-Fi access points to better handle real-time applications by reducing delay and jitter.
Currently protocols adapt based on network conditions such as loss rate, bit errors and signal-to-noise ratio, the researchers say.
But they found that if a device is moving and a packet is lost, the likelihood that the next packet will also be lost is much higher than if the device is stationary. In fact the research showed that if a device is static, whether a packet fails to get an acknowledgement tells very little about whether the next packet will also be lost.
So it makes sense when a device is moving and a packet is lost for the protocol to quickly (after about 10 milliseconds) ratchet back its bit rate, they say. And after a short interval the protocol should try a higher bit rate. If that is successful, it adopts the faster rate and samples even higher rates to see if they succeed.
The team incorporated this thinking into the RapidSample protocol, which continuously adjusts bit rate based on the success or failure of packets to make it through.
If hints indicate the device is static, it uses a different bit rate protocol altogether, called SampleRate, because it yields better throughput under that condition.
The hint-aware protocol yielded up to 76 per cent improvement over SampleRate alone, and averaged a 20 per cent improvement, the researchers say.
To minimize access-point hops, the MIT team designed an association strategy based on whether the device is moving. If so, it is more likely to come across an access point with a stronger signal, so it should periodically seek other access points. How often it checks is determined by how fast the device is moving.
When the device stops moving, it should stop scanning, but start up again as soon as it registers that it is moving. It should also start scanning again if it is stationary but the strength of the associated access point drops below a given threshold.
The average throughput improvement using this scheme was 30 per cent, the researchers say.
To reduce the number of access points connected to, the device must first be trained in the Wi-Fi environment. It then takes hints about direction and speed of movement to determine which access points to find the next one to connect to when the signal strength of the current one drops below a set threshold. The next access point chosen is the one that the device can remain connected to the longest, given its current speed and direction.
The number of handoffs was 40 per cent lower when based on hints, the researchers say.
So far software to interpret sensor data has been developed for Android and Linux devices, but it should be straightforward to incorporate it in other mobile platforms, the researcher paper says.
The sensor data is organized by a Sensor Hint Manager and sent to other mobile device and access points either via UDP packets or 802.11 Wi-Fi frames, depending on the network and the capabilities of each mobile device. Possible hints are: movement, true/false; walking, true/false; heading, degrees relative to true north; speed in miles per hour; environment, indoor/outdoor.
