Hi guys, i’m new in this world. I’m doing a little research project and i’m using the mpu6050 to receive data from accelerometer. I have to find a way to transform accel data into speed and eventually in position. I’ve read something about kalman filter and complementary filter, but the code that i found, was about something with angle. I need in some way an approximately lecture from accelerometer in x,y,z coordinates so i can integrate twice to determinate position. I also read that a way to correct bias error is to calibrate the accelerometer in a short period of time and subtract it every time in x,y,z coordinates.

Thanks to everyone that can help me or suggest me where i can find some stuff to study.

The MPU6050 is an accelerometer (and a gyro).

I have to find a way to transform accel data into speed and eventually in position.

In theory, you can use numerical integration. In practice, this does not work with consumer grade accelerometers, because of noise and inaccuracy.

Here is the theory and an explanation for why it will not work.

Thank you for the answer, i read about the noise problem, but i don't need an accurance data, just a little bit precise. I read about the kalman filter to filter the noise, but the code that i've found it's just about for calculate angles with arct2. Is it possible to have an approximately lecture by calculate bias error and substract it? What i have to understand is that how much the error is, because if the real valure is 1cm but with error i read 5 cm, for what i have to do, is still accetable. Thank you guys.

What i have to understand is that how much the error is

For the error in velocity and position, see table 1 in this link.

The typical orientation error for MPU6050, with a good Kalman filter is about 2 degrees, so check the table entry for 2 degrees. Beware of errors in the position column.

With 2 degrees orientation error, after just 10 seconds, your expected velocity error is +/- 3.4 m/s and the position error is +/- 17.1 m (error in the table).