Microsoft Research Connections Blog
Next at Microsoft
Social Media Collective
Windows on Theory
Posted by Rob Knies
Conventional wisdom holds that the use of GPS satellites to enable indoor mapping is a non-starter. GPS receivers, it is said, simply don’t work indoors, for a variety of reasons. While Earth’s outdoors environment has been mapped extensively, indoor localization of places such as shopping malls or department stores remains an elusive dream.Conventional wisdom, says Jie Liu, is wrong.COIN-GPS: Indoor Localization from Direct GPS Receiving—a paper from Liu and his fellow Microsoft researchers Gerald DeJean, Bodhi Priyantha, Yuzhe Jin, and Ted Hart, along with Shahriar Nirjon of the University of Virginia—challenges current assumptions about the viability of GPS for indoor localization. That paper just won a best-paper award during MobiSys 2014, the 12th International Conference on Mobile Systems, Applications, and Services, held June 16-19 in Bretton Woods, N.H.The research described in the paper represents the first time that innovation in device design, combined with the use of GPS signal-processed algorithms that capitalize on the power of the cloud, can obtain GPS-based locations in many indoor spaces.Liu explains the significance of this advance.“People spend more than 80 percent of their time indoors,” he says. “Location is key contextual information to enable smart services, such as Cortana. With GPS and Wi-Fi-based solutions, people are taking outdoor location services for granted on mobile devices, but that experience has not been transformed to indoor environments.“Currently, the main approach for providing indoor location services is to profile an indoor space to get a set of Wi-Fi maps and use them to infer user location based on the device-provided Wi-Fi signatures. Profiling is a time-consuming and expensive process, because there is no ground truth to ‘nail down’ the Wi-Fi maps. Profiling is also a continuous process, because the number and the locations of the Wi-Fi access points change over time.”The solution, the paper suggests, is to achieve direct, GPS-based indoor localization by attaching a steerable, high-gain, directional antenna to the front of a GPS receiver. This combination offers a novel way of obtaining location fixes using robust signal processing from the cloud with acquisition results from different directions over time.“This paper is the first effort to provide direct locations in indoor environments just as GPS does for outdoor environments,” Liu says. “If this approach works, then no profiling is necessary—as long as the end device incorporates the rather complex multiantenna configuration. We anticipate that the sizes of the antenna [currently a 10-inch-by-10-inch board with 16 antennas arranged in an array] will shrink over time as technology improves.“The best use case of the technology in the short term is probably profiling machines for indoor environments. Not all indoor places can receive GPS signals, but if we have enough ground truth and combine them with inertial navigations using accelerometers and gyroscopes, we can make the indoor-profiling process much cheaper and quicker.”The researchers tested their system in 31 random indoor locations in single-story indoor environments such as warehouses and shopping centers, and they were able to obtain location fixes in 20 of them, with a median error of fewer than 10 meters. This is a scenario in which all normal GPS receivers fail.This means that acceptable location accuracy is delivered without requiring additional infrastructure. No other GPS receiver is as successful.The work revealed in the COIN-GPS paper indicates that soon, it will be as easy to find a Starbucks while in a shopping mall as it is while standing outside the mall. The GPS ecosystem continues to evolve.“The U.S. government is planning to launch GPS III, which will have stronger signals and more complex code,” Liu states. “It is reasonable to anticipate that GPS III will work for some indoor environments, especially if people incorporate what we’ve shown in this paper.“Before GPS III is a reality, mass-market indoor-location services will continue to be based on various kinds of radiofrequency signatures. Our goal is to make this profiling process more robust. For example, we should only admit GPS-based indoor locations when we are very sure about their accuracy and are able to eliminate outliers. Then we can be confident enough to combine them with inertial navigation for profiling.”
"median error of fewer than 10 meters" is too much for indoor localization.
Maybe iBeacons and BLE are better option to indoor localization