Should EK3 parameters be flight controller dependent?

Hello!
I’ve been exploring the EKF3 and how EK3 parameters affect state estimation (using the Replay tool) and there is something that puzzles me.

When I compared the default EK3 parameters of a Cube Orange and a CUAV X7+ Pro, these parameters are the same. Even the gyro and accelerometer related parameters. I was under the impression that the default values might be different depending on the board’s IMUs, but it’s apparently not the case. The X7+ Pro is equipped with a “tactical-grade” ADIS16470 IMU, so I would have imagined that the EKF would take that into account.

My question are:

  1. Should the EKF3 take this into account?
  2. Would reducing parameters such as EK3_GYRO_P_NSE, EK3_ACC_P_NSE, EK3_GBIAS_P_NSE and EK3_ABIAS_P_NSE be appropriate for the ADIS16470?
  3. Are these parameters dependent on expected vibration (i.e. if vibrations are high, the parameters should not be lowered)?
  4. How were the default parameters determined?

My supervisor is interested in funding a master’s student (or at least an intern) to work on tuning EK3 parameters depending on the flight controller and equipped sensors (RTK vs non-RTK, better barometers and baro housings, different rangefinders). Maybe developing a procedure to tune these parameters. So I’m just getting the ball rolling and seeing if we correctly understand these parameters, and if they should be adjusted for the sensors’ specifications.

Another aspect about these parameters. There is currently only 1 of each parameter, even though there are multiple different IMUs on a board, even though external barometers or compasses could be different from internal ones. Would there be a gain in defining these noise and bias parameters for each IMU, compass, baro, GPS, rangefinder, etc?

Thanks!

1 Like