New study uses mathematical analysis of walking patterns for early detection of Parkinson’s disorder.

Using the Crowd to Improve Location Accuracy

Read time: 1 min
Mumbai
5 Oct 2018
Photo: Jpatokal at Wts Wikivoyage via Wikimedia Commons CC BY-SA 4.0

Imagine you book a cab using smartphone app, and it shows your cab is five minutes away. Five minutes are gone, and the app still says the same thing! Sounds familiar? While you could easily blame the rush hour traffic, did you know the culprit could also be the Global Positioning System (GPS) of your or your cab driver's phone? Yes, often phones fail to receive the GPS signal because of the interference from tall buildings. In a recent study, researchers from the Indian Institute of Technology Bombay (IIT Bombay) have designed a solution to improve the accuracy of the location in absence of GPS by making your phone communicate with other phones around.

The Global Positioning System (GPS) is a satellite-based navigation system where navigation signals from satellites are received by GPS receivers fitted in most smartphones available today. Unlike the cellular network that connects our phones to each other using nearby mobile towers, GPS relies solely on direct satellite signals. Although the location of a cell phone can be traced by correlating its signal strength and the distance to the nearby cell towers, this approach is inaccurate compared to GPS. However, GPS has its drawbacks too; besides consuming more power, it often fails to work in densely populated city centres with tall buildings, crowded public transport, or inside tunnels where it fails to catch the signals from the satellites.

In this study, the researchers propose a novel, crowdsourced application called CrowdLoc that makes the best use of the inaccurate location information using cellular technologies to improve the accuracy of the estimated location. Their research shows that the errors cancel out when location information from many nearby phones are combined.

“The main innovation in CrowdLoc is that it turns a 'crowd' into an advantage, by sharing information to improve location accuracy, even as the same 'crowd' makes GPS unreliable”, says Prof. Bhaskaran Raman from IIT Bombay who led the study. The findings of this research are published in ACM Transactions on Sensor Networks (TOSN), a reputed computer science journal.

When a phone has CrowdLoc installed, the algorithm in the application combines bare minimum information of its ‘location fingerprint’ along with such fingerprint of other phones with the same app installed, which are within a distance of five metres. The location fingerprint includes the identity of the closest cell tower, the signal strength and the GPS coordinates. Thus, those phones installed with CrowdLoc take part in 'crowdsensing', mutually helping each other to improve the accuracy of their respective locations.

But, what about the privacy, you ask? The 'location fingerprint' shared by CrowdLoc with other phones does not contain anything that uniquely identifies the owner of the phone! There is no personal information such as names or phone numbers that are shared. Also, the application sends the ‘location fingerprint’ every few seconds over an inaudible radio frequency. This approach is not only fast but also energy efficient. Thus, a phone can communicate with those nearby without turning on its Wifi or Bluetooth.

The researchers tested the application by collecting experimental data from two cities. They chose Mumbai because it is a large and densely populated metropolitan city with multiple means of transportation. The other city was Chandigarh, which is a smaller city with a well-planned road network. The researchers observed that their system worked well in spite of the loud, external audio noise of the vehicles, honking and people—a trait common in many metropolitan cities in the developing world. The accuracy of location as determined by the application was not affected even if the phones were of different brands, carry SIM cards from different operators, or have access to the Internet.

“Using CrowdLoc, we found that the median localization accuracy went from 97 metres to 33 metres over the traditional method of using cell-phone data alone. It works well in places where GPS is expensive or unavailable. We have the CrowdLoc prototype available as a module for use in other location-based apps”, says Ravi Bhandari, a PhD candidate at the Department of Computer Science and Engineering of IIT Bombay and one of the primary authors of this research.

The other advantage of the CrowdLoc application is that it does not have to process a massive amount of computationally intensive data to yield accurate results. It can work well even with two to four phones in the vicinity of your phone, with one of them having commuted in the same route at least once.

CrowdLoc could be a gamechanger in the public transport scenario in India, which are always crowded and the probability of a daily commuter in any given journey is high.

CrowdLoc could be used at many places where a little accuracy can be traded with seamless location availability and less energy usage. Examples of such applications could be predicting arrival time of crowded trains or buses, where GPS is unavailable most of the time. Another application could be a city-wide tourist application, where energy consumption of the application is a concern, but GPS-level accuracy is not required. For CrowdLoc to scale to other commuters, engaging incentive mechanisms tied to tangible/monetary rewards, have to be developed”, signs off Ravi.