Closed RajS999 closed 1 month ago
There are additional sources of noises not captured by the calibration procedure, in the ideal case we wouldn't need any inflation. Noise caused by temp / humidity / motors of the robot etc will all have an impact here.
I would read the original specification if you need to see how they determined the method: [IEEE Standard Specification Format Guide and Test Procedure for Single-Axis Interferometric Fiber Optic Gyros (page 71, section C)]
It ends up being this chart here:
It should be a onetime process, but if you take the drone to a very hot (or cold) or completely different environment, then I would expect all sensors to need to be re-calibrated in this case.
I was referring to calculation of gyro and accel white noise and random walk using the video Visual-Inertial Sensor Calibration - A Complete Tutorial and Discussion for my Intel Realsense. I have following doubts:
T=1, Slope=0, T=3
respectively. Why exactly Angle Random Walk is read at T=1 and Rate Random Walk at T=3 and not at some other values? Is it that these types of noise are "empirically observed" to most dominant in Allan Variance plot at these cluster interval values?