If you’ve ever used a smartphone to navigate, you’ll know that one of the biggest problems is running out of juice. GPS sensors are a significant battery drain and so any journey of significant length requires some kind of external power source. Added to that is the difficulty in even getting a GPS signal in city centre locations where towering office blocks, bridges and tunnels regularly conspire to block the signal.
So a trick that reduces power consumption while increasing the device’s positioning accuracy would surely be of use.
Today, Cheng Bo at the Illinois Institute of Technology in Chicago and a few pals say they’ve developed just such a program, called SmartLoc, and have tested it extensively while travelling throughout the windy city.
They say that in the city, GPS has a positioning accuracy of about 40 metres. By comparison, their SmartLoc system pinpoints its location to within 20 metres, 90 per cent of the time.
So how have these guys achieved this improvement? The trick that Bo and pals use is to exploit the smartphone’s inertial sensors to determine its position whenever the GPS is off line.
The way this works is straightforward. Imagine a smartphone fixed to the windscreen of a car driving around town. Given a GPS reading to start off with, the smartphone knows where it is on its built-in or online map. It then uses the inertial sensor to measure its acceleration, indicating a move forwards or a turn to the left or right and so on.
By itself, this kind of data is not very useful because it’s hard to tell how far the vehicle has traveled and whether the acceleration was the result of the car speeding up or going over a humpback bridge, for example.
To get around this, the smartphone examines the section of road on the map looking for road layouts and features that might influence the sensors; things like bends in the road, traffic lights, humpback bridges and so on. Each of these has a specific inertial signature that the phone can spot. In this way, it can match the inertial signals to the road features at that point.
The key here is that each road feature has a unique signature. Bo and co have discovered a wide range of inertial signatures, such as the deceleration, waiting and acceleration associated with a set of traffic lights, the forces associated with turnings (and how these differ from the forces generated by changing lanes, for example) and even the change in the force of gravity when going over a bridge.
Having gathered this data, the SmartLoc program looks for these signatures while the car is on the move. These guys have tested it using a Galaxy S3 smartphone on the city streets in Chicago and say it works well. They point out that in the city centre, the GPS signal can disappear for distances of up to a kilometre, which would leave a conventional navigation system entirely confused.
However, SmartLoc simply fills in the gaps using its inertial signature database and a map of the area. “Our extensive evaluations shows that SmartLoc improves the localization accuracy to less than 20m for more than 90% roads in Chicago downtown, compared with ≥ 50% with raw GPS data,” they say.
That certainly looks handy. And this kind of performance could also help save battery power by allowing a smartphone to periodically switch off the GPS sensor and run only using the inertial sensor.
What Bo and co don’t do is explain their plans for their new system. One obvious idea would be to release it as an app–it clearly already works on the Android platform. Another idea would be to sell the technology to an existing mapping company. Perhaps they’re planning both. Whatever the goal, it seems worth keeping an eye on.
Ref: arxiv.org/abs/1310.8187: SmartLoc: Sensing Landmarks Silently for Smartphone Based Metropolitan Localization
How a Russian cyberwar in Ukraine could ripple out globally
Soldiers and tanks may care about national borders. Cyber doesn't.
Meet Altos Labs, Silicon Valley’s latest wild bet on living forever
Funders of a deep-pocketed new "rejuvenation" startup are said to include Jeff Bezos and Yuri Milner.
A horrifying new AI app swaps women into porn videos with a click
Deepfake researchers have long feared the day this would arrive.
Meta’s new learning algorithm can teach AI to multi-task
The single technique for teaching neural networks multiple skills is a step towards general-purpose AI.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.