1

I try to get the Z-angle rotation of a GY-521 ( MPU 6050 ) but the angle rise static also if i don't move the sensor. Is there a way to "filter" the angle that he is correct for my case ?

Code:

float accel_z = accel_t_gyro.value.z_accel;
float accel_angle_z = 0;
float gyro_angle_z = gyro_z*dt + get_last_z_angle();
float unfiltered_gyro_angle_z = gyro_z*dt + get_last_gyro_z_angle();
float alpha = 0.96;
float angle_z = alpha*gyro_angle_z + (1.0 - alpha)*accel_angle_z;

output when sensor not moved:

x = x-roatation-angle z = z-roatation- angel

      x   z
    8.37 4.24
    8.35 4.22
    8.33 4.21
    8.32 4.19
    8.31 4.18
    8.29 4.17
    8.28 4.15
    8.26 4.14
    8.25 4.12
    8.23 4.12
    8.22 4.10
    8.21 4.09
    8.19 4.07
    8.18 4.05
Timo Jungblut
  • 662
  • 6
  • 16
  • possible duplicate of [How do you use a moving average to filter out accelerometer values in iPhone OS](http://stackoverflow.com/questions/2272527/how-do-you-use-a-moving-average-to-filter-out-accelerometer-values-in-iphone-os) – Paul R Oct 17 '13 at 20:41
  • It looks like you already have a low pass filter - maybe you need to increase the value of `alpha` to get more smoothing ? – Paul R Oct 17 '13 at 20:43
  • got a solution for my problem. its not possible to read the rotation on the z axis the angle change because there is no gravity that hold the gyroscope in the right position ;) – Timo Jungblut Oct 20 '13 at 20:34

1 Answers1

1

Timo, I saw that you found a partial solution to your problem (quote: "there is no gravity that hold the gyroscope in the right position"). I would like to elaborate because there are ways to get reasonably good results from a gyroscope only. I don't know if it will be good enough for your application. Here are the results I got using the same hardware as you: http://robokitchen.tumblr.com/post/68625623585/imu-with-gyroscope-only-i-tested-the-three-axes. In this video I am not using the accelerometer at all.

There are two points that could help you get better results:

1) Unless your sensor remains totally horizontal you cannot get the z-axis rotation using this formula:

float gyro_angle_z = gyro_z*dt + get_last_z_angle();

To explain why, imagine this: if your gyroscope turned 90 degrees around the x axis then 90 degrees around the z axis, you would not obtain the same result as if it only turned 90 degrees around the z axis. This means that you should compute the rotation around all the axes and then extract the z-axis rotation. If you don't use the x and y axes your results will be incorrect (unless your sensor remains totally horizontal).

One efficient way to compute the rotation around all the axes is to use quaternions. There is a lot to say about quaternions, so I let you do your own research and ask more questions if required.

2) After using all the 3 axes there will still be a static drift. This is because the gyroscope will not read a perfect 0 even when it does not move. To improve the situation, you could calibrate the gyroscope. An easy way to calibrate is to read a large number (100+) of values when the gyroscope is not moving, calculate the average drift, and subtract it from your measures when the gyroscope is moving. There will still be a slight drift after this, but not as bad as what you may be experiencing now.

Here are the best results I got, using the MPU-6050 gyroscope and accelerometer: http://robokitchen.tumblr.com/post/68626551378/imu-with-gyroscope-and-accelerometer-i-ran-the. As you already know the accelerometer helps stabilise the x and y axes but not the z axis. However the z axis drift is low and remains useable for many applications. For instance there are several quadcopters which do not use a magnetometer and suffer from z axis drift and yet remain useable.

marcv81
  • 850
  • 8
  • 22