ARDUINO UNO-LSM303DLM

Hello Evangelia, 

A few key concepts:

* Integrating acceleration gives speed

* Integrating speed gives position 

So: if you integrate the acceleration twice, you'll get the position, at least if you're starting from a 'no movement' situation.

Integration can be done by adding up signals between two known sample periods, as you have already shown. So (leaving out any scaling factors):

speed += acceleration*dT

position += speed*dT

The dT should be constant for easy calculation, and preferrably not the "wait" time in your loop. Either calculate the time you spend in the loop (using millis() ), and only end an iteration of your loop when you've passed 'x' milliseconds, or unleash the power of the microcontroller and read up on how to use timers and interrupts. That will greatly improve the result.

The bad part of all of the above is that you're also integrating your measurement error twice. If you have a bit of noise / measurement error in your accleration data (or your sensor clips / isn't fast enough during a shock) your speed will not be right and you'll have offset errors in position. In practice, the reliability of the position data will deteriorate over time if you take no other measures. What you could do is to have some reference point where you know what the exact position is, so you can reset your integrators.

To know what units the library produces you'll have to read the source of the library and / or the datasheet.

Good luck!

@Victor, everything you say here is true but it may help the OP if we add some numbers.

v = a * t

s = (a * t^2)/2 (distance = acceleration times time squared all over 2)

now if the OP has an accelerometer with 2g (approx 20m/s/s) full scale and the error is only 0.1% (which would be very good going for the LSM303DLM ) the error is 20/1000 = 0.02m/s/s

after one second the velocity error is 0.02m/s

after one minute the velocity error is 1.2m/s

after one hour the velocity error is 72m/s

after one second the error in distance will be (1 * 1 * 0.02)/2 = 0.01m

after one minute the error in distance will be (60 * 60 * 0.02)/2 = 36m.

after one hour the distance error will be (3600  * 3600 * 0.02)/2 = 129600m

what this means is that you need to think quite carefully about your application - if it was for a car speedometer ther errors will become too great (for it to be useful) after only a few minutes. If it is in an application where you have a way of resetting at frequent intervals (relative to the error from the accelerometer) then estimating velocity or position from measured acceleration can work.

MK

@Victor, everything you say here is true but it may help the OP if we add some numbers.

v = a * t

s = (a * t^2)/2 (distance = acceleration times time squared all over 2)

now if the OP has an accelerometer with 2g (approx 20m/s/s) full scale and the error is only 0.1% (which would be very good going for the LSM303DLM ) the error is 20/1000 = 0.02m/s/s

after one second the velocity error is 0.02m/s

after one minute the velocity error is 1.2m/s

after one hour the velocity error is 72m/s

after one second the error in distance will be (1 * 1 * 0.02)/2 = 0.01m

after one minute the error in distance will be (60 * 60 * 0.02)/2 = 36m.

after one hour the distance error will be (3600  * 3600 * 0.02)/2 = 129600m

what this means is that you need to think quite carefully about your application - if it was for a car speedometer ther errors will become too great (for it to be useful) after only a few minutes. If it is in an application where you have a way of resetting at frequent intervals (relative to the error from the accelerometer) then estimating velocity or position from measured acceleration can work.

MK

Thanks!