hi all -
i'm a beginner to accelerometers and i'm using a LSM6DSOX to try and get some baseline positional data.
i am using "numerical integration"/"euler's method' - i multiply the acceleration by the delta time...and then i multiply that result by the delta time again.
as i'm sure many of you already anticipate, the results are subpar. can anyone help me understand better what i can reasonably expect from these sensors? is there anything that can be done without buying more hardware?
thanks
You have already discovered that you can't measure position using consumer grade accelerometers. They are simply too noisy and inaccurate.
what i can reasonably expect from these sensors
You can expect to measure acceleration values to within about 1% accuracy. In conjunction with a magnetometer, 3D orientation can be measured to within +/- 3 degrees or so.
Even if you had a perfect measurement of acceleration, the above method is a misuse of Euler's method, which will never get you a correct result.
Suppose you had a simple situation in which the object starts at rest and then acquires an acceleration that is constant (a). If you take k measurement at a time interval T, you've said that you are calculating the displacement as follows:
x(kT) = Σ aT2 = kaT2
The correct result for this example should be:
x(kT) = ½k2aT2
Thus, your result would be off by a factor of ½k (which grows larger and larger with time).
And this is not even getting into the fact that you need to eliminate the gravitational acceleration (g) from your signal before doing any integration...
This is very helpful to understand the positional data problem.
I am also curious if there are some useful tools and resources out there to help visualize the incoming information and experiment with these devices. I am still learning about them and I haven't been finding many good videos on YouTube.
There is more to the problem than the integration step. You will probably find this archived blog post helpful, as it explains a more serious issue.