2

I'm trying to combine the data from an accelerometer and a gyroscope to accurately measure the pitch and yaw angles of an object. After researching the complementary filter and attempting to implement it, i have a few questions on how it works.

I've read that the filter "trusts" the gyroscope data if there is a lot of angular movement and that it "trusts" the accelerometer data if the object is stable.

http://www.pieter-jan.com/node/11 In this article the complementary filter is described in this way:

*angle = 0.98(angle + gyrData * dt) + 0.02*(accData)*

To me it, seems as if the gyroscope data is being favoured. In the following image, http://www.pieter-jan.com/images/resize/Complementary_Filter.png , found at the bottom of the page, the filtered data seems to "keep close" to the accelerometer data, even though the gyroscope data has drifted. I don't understand why this occurs when the calculation suggests the gyroscope data is being favoured. I have observed this in other photos as well. During my own testing i needed to "swap" the 0.98 and 0.02, suggesting the accelerometer data is being favoured, to obtain similar results. Am i missing completely misunderstanding how this filter works? Is it normal to "favour" the accelerometer data?

Furthermore when the angle of an object needs to be monitored for a long length of time, doesn't the gyroscope data become useless as the drift is so large, how does the filter compensate?

MrBigRig
  • 21
  • 1
  • 3

2 Answers2

0

I Realise where i was going wrong. I had essentially calculated the angle using only the gyroscope data and used that in the filter. i.e

GyroAngle += d°/s * time_between_cycles
FilteredAngle = 0.98*GyroAngle + 0.02*AccelerometerAngle

Instead i should've been doing this:

FilteredAngle = 0.98*(FilteredAngle + d°/s * time_between_cycles) + 0.02*AccelerometerAngle

Doing this has yielded much better results

MrBigRig
  • 21
  • 1
  • 3
0

I'm trying to do something similar to this. I'm implementing a complementary filter in my Arduino code using an LSM9DS0 sensor (It has a gyro/accelerometer built-in https://www.adafruit.com/product/2021

There's a filter value I have been playing around and a calibration method I'm trying to use but I can't seem to get rid of it 100%. There is always a small deviation from the angle and I can never get a 100% filtered angle with no error.

user7838253
  • 1