I am in the beginning stages of development for a project where I need to continuously track the position of an object for long amounts of time. The object will not move very much, but when it does I want to track it to see when it moves 10 meters from its initial position.
I looked into IMU tracking as I have some past experience with IMUs but they seem to have huge problems with long term position tracking due to bias buildup.
The object will be in an area where people can walk, so something like ultrasound won't really work because someone can get in the way of the sensor and make it look like the object is closer to the wall than it used to be when in reality someone just stepped in front of it. The object in question is a large chair that will be stationary most of the time and can only be moved by humans but it can be lifted up. I am mainly concerned in tracking its location on a 2D surface which is the floor its on, but it CAN be lifted off the surface.
Triangulation has been mentioned but I do not know what technology is involved. Do I need to mount reflectors on the object? There will be other things present and moving in the room like people and furniture so how do I account for that and only focus on the object I am interested in n?
The response to the object moving must be somewhat short, preferably less than 30 seconds but closer to 5-10 if possible (and of course lower is always better).
What are the most common techniques used for long term position tracking? When I say "long term" I mean months preferably but right now I'm just trying to figure out what's possible with current technology.
Best Answer
Since your object seems not to be moving most of the time, you may still use an IMU using the following assumption :
If the object is not accelerating and approximately not moving during a long time, then it's probably not moving at all.
Then, use an IMU and check for this condition ; as soon as this condition is detected, consider the position to be that at the beginning of the static phase until acceleration is detected. This effectively zeroes out the drift bias during all immobility phases.
That's simple programming, rather power-efficient, and very robust to interference if the above assumption is valid.