search

kriswiner / BNO055

Teensiduino sketch for 9-axis BNO055 motion sensor + MS5637 pressure sensor
54 stars 22 forks source link

BNO055 hardware fusion is wrong at low angular speed #12

Closed schadocalex closed 5 years ago

schadocalex commented 5 years ago

Hi, I use a BNO055 and I have a strange problem. At low angular speed, the intern kalman does not change the yaw value. If I turn manually the imu at about 3deg/s, the angle and the gyroscopes values returned by the fusion are static (there is some very few noise near 0). But above 5-10 deg/s it's working as expecting (and very well).

If I read raw input (without fusion), gyroscope values become good even at low speed.

Important note: I don't use magnetometer because of the huge metallic mass near the imu (the BNO055 is in IMU mode for fusion (acc+gyro)).

kriswiner commented 5 years ago

I have no idea what is happening here. This sounds like a question for Bosch...

You might have to do the fusion on the MCU host using the raw sensor output...

On Mon, Oct 29, 2018 at 6:57 AM Alexis Schad notifications@github.com wrote:

Hi, I use a BNO055 and I have a strange problem. At low angular speed, the intern kalman does not change the yaw value. If I turn manually the imu at about 3deg/s, the angle and the gyroscopes values returned by the fusion are static (there is some very few noise near 0). But above 5-10 deg/s it's working as expecting (and very well).

If I read raw input (without fusion), gyroscope values become good even at low speed.

Important note: I don't use magnetometer because of the huge metallic mass near the imu (the BNO055 is in IMU mode for fusion (acc+gyro)).

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/kriswiner/BNO055/issues/12, or mute the thread https://github.com/notifications/unsubscribe-auth/AGY1qg49VMqqKwhvv1cnS6x4Jm2WXByDks5upwksgaJpZM4X_Opb .

schadocalex commented 5 years ago

Well, after posting this I did ask Bosch directly (I found a official repo). But we can't delete issues so I let this one. They answered me, you can see the answer if you want: https://github.com/BoschSensortec/BNO055_driver/issues/12#issuecomment-433944642

Sorry for the trouble.

kriswiner commented 5 years ago

It is going to be hard (impossible really, without significant drift) to get accurate heading without a magnetometer but if all you care about is roll and pitch then Madgwick's 6 DoF fusion algorithm works well with well-calibrated accel/gyro data with even a modest MCU.

https://github.com/kriswiner/MPU6050/blob/master/MPU6050Library/MPU6050IMU.ino

On Mon, Oct 29, 2018 at 9:32 AM Alexis Schad notifications@github.com wrote:

Closed #12 https://github.com/kriswiner/BNO055/issues/12.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kriswiner/BNO055/issues/12#event-1932745594, or mute the thread https://github.com/notifications/unsubscribe-auth/AGY1qjXdd7XPDtgSFKz5CBBgfXqpRgRXks5upy2ygaJpZM4X_Opb .

schadocalex commented 5 years ago

I have a very special case: I care only about relative yaw (relative since an old yaw position). I only want to know my position relative to a previous one, so I think with gyro it will be ok (I will comment with my tests results If you want).

The heading in my case will be given by others sensors than a magnetometer (that does not work in my case, I have 2t of steel and high DC power (about 10kW sometimes) near the imu).