Rice University researchers have developed a method that works twice as well as GPS and at 27 times the power savings. Image credit: Rice University
Researchers at Rice University have developed a new method for interior navigational location detection using existing sensors in mobile devices.
While GPS navigation works sufficiently in the open, it falters under poor signals that quickly deplete battery life when in large indoor spaces such as office complexes or shopping malls. Researchers used machine learning for location detection to increase the speed of calculations and decrease the amount of energy used for location detection tapping into sensors that already exist inside smartphones and other mobile devices.
“The original idea is to just use the gyroscope and accelerometer data for indoor-location detection, but the results were poor," says Chen Luo, a graduate student working on the project at Rice. “After we added in some mapping information to our model, the performance improved significantly.”
Since gyroscopes and accelerometer sensors are already built into most mobile devices, it provides a cheap solution. However, these sensors are noisy because of irrelevant movements, meaning the sensors track walking movements but also swinging arms and waving hands. This resulted in a number of errors when trying to compute the final location.
Researchers instead used the idea of estimating answers rather than working with precise calculations because it is not only energy-efficient but also predictable since most humans walk in a straight line and don’t typically make erratic movements. This was a starting point for the machine learning algorithm, preconditioned with humans travelling in a straight line with limited opportunities for possible left and right turns. Location could then be garnered accurately when someone stops moving even with noisy sensors.
Using this solution, researchers were able to demonstrate a method that is twice as accurate as GPS services while about 27 times cheaper in terms of energy, extending battery life in mobile devices.
The work in location detection could be applied to a number of applications in the marketing field, healthcare and pet care markets. For example, a marketer could extend product offers beyond the current location of where people frequently visit or in healthcare could trigger alarms if patients are near harmful areas or help locate missing dogs or cats.
