Hacker News Reader

Comment by untog | Hacker News Reader

I'd love to know a little more about what's going on here. Does it use accelerometers or just wifi signals? One key question: could I share these floorplans? That is to say, using the sensors on HTC Android device #1, could I sent a floorplan to my friend Samsung device #2 and have them be able to similarly track their location?

Does it only work indoors, or is that just for demonstration?

View on HN

The floorplans are definitely transferable! Once you’ve mapped a location once, it works automatically on all devices.

In fact, if you walk into a location that someone else has already mapped, the “Where Am I” will automatically load that map up on your device and start showing your position.

If you’d like to do more advanced stuff with your floorplans, make a WiFiSLAM SDK account here (access.wifislam.com) and we can get you started.

What's in the SDK? I looked around and there was not even a mention of ios versus android. Is it a set of armv7 ndk android libraries? Or something else? Can I use it on simulators running on OS X?

The current WiFiSLAM SDK is an Android java library, available as an Eclipse project with sample code snippets that allow you to add your location to any existing Android app.

We have an iOS SDK that works, unfortunately since Apple disabled WiFi scanning in iOS 5 we can no longer publish it. It still works on jailbroken devices, but not a big target audience. We have a workaround in the works that will incorporate iOS devices

It will work wherever there are ambient WiFi signals, indoors or outdoors. This is particularly interesting indoors because it’s possible to get GPS outdoors, while you can’t indoors. We have tried mapping around Mountain View and it’s worked great

My specific use case is in tracking very small movements, i.e. ones that a GPS would not pick up. I was wondering if this would provide a solution to that, but it still seems to have a delay attached.

Doesn't take away from some very impressive tech on display. Kudos.

thanks! How small are the movements you’re trying to track? And what kind of time delays are you trying to get under? You can try mapping the same location multiple times (the more mapping the more accurate our system will be)

@untog: It's actually a sensor fusion approach that combines both accelerometers and Wi-Fi signals.

Can you guys give any insight into how you're combining the sensor data? I'm a newb on the topic, but I was trying to do something similar in a different domain and got stuck trying to make a kalman filter. I tried to get an EE friend to explain them to me, but I couldn't get it.

Multiple fields actually invented Kalman filters independently and it turns out they are all actually the same thing, they just have widely varying derivations. My personal favourite explanation comes from the textbook: Probabilistic Robotics (2005) by Thrun, Burgard, and Fox.

Reply on news.ycombinator.com