Join devRant
Do all the things like
++ or -- rants, post your own rants, comment on others' rants and build your customized dev avatar
Sign Up
Pipeless API
From the creators of devRant, Pipeless lets you power real-time personalized recommendations and activity feeds using a simple API
Learn More
-
Oh, also I was researching a method of calculating a user's path based on accelerometer data only (because GPS does not work indoors and first gen iPad has no gyroscope as well). In theory, knowing acceleration one could derive the velocity and then path, but it is pretty much impossible in practice. Accelerometers are too noisy and the path will soon drift away from the true position even with signal smoothing.
-
@iiii Does the noise have any bias? That might be useful.
Can you use orientation of the device too?
Hell if you can recognize bright sunlight through a window via the camera, and associate it with the time (sunrise, sunset) then hypothetically, you can derive at least a rough orientation periodically.
I think "sensor-broken" type software that has a suite of secondary cues it can pick up on from the environment, is an interesting project.
Time of day, orientation via gyproscope, known light source, known sound source, last known location, noise bias in accelerometer.
Don't devices already use known position of static wifi sources as a reference for refining gps now? -
@Wisecrack no, the noise is uniform. No, I could not use orientation because of device constraints (1 gen ipad has no gyro). Also it was impossible to use more sophisticated methods because of hardware constraints as well.
-
@iiii "the noise is uniform"
Did you test that statistically or eyeball it?
What appears to be uniform data could have surprising correlations. -
@Wisecrack measured. It was just write noise of the detector's precision limit.
Pretty much a sort of research work. The first assignment was: "look, we have this CAD viewer, but we would want to eventually optimize the structure of the mesh, so here's this method of minimizing the memory footprint. Try implementing it and integrating it with our application."
PS: the method is using triangle strips, where the next triangle uses two vertexes of the previous one, theoretically reducing the memory footprint of the mesh by 2/3 if the mesh is fully optimized. In the end, due to memory and performance constraints (this had to run on the first gen iPad), and overall application architecture, on the fly striping was unfeasible and gained no benefit, because striping an arbitrary mesh is a fucking hard task.
Another one was an implementation of smooth shading by recalculating vertex normals in runtime.
rant
wk254