I was going through theory of INS and trying to understand attitude estimation, velocity and position and estimation. Theroy can be summarized as follows,
1) Gyro is integrated to estimate the attitude/orientation, with initial orientation as the initial condition.
2) Accelerations measured in body frame are converted to inertial frame using orientation given by gyro, and then acceleration in inertial frame is integrated to get velocity and again integrated to get position. Initial velocity and position are the initial conditions.
So, above two steps are quite clear. But now the doubt/problem comes into picture,
1) Any gyro, however accurate it is will have bias and moreover changing bias, which will mess up the estimated attitude for sure. In this scenario, estmation of postion and velocity will definitely be erroneous.
2) Error in position and velocity can be tackled by GPS, but what about attitude? How to restrain drift in estimated attitude within acceptable limits?
3) Lets assume a scenario in which a small UAV is going in tight horizontal circiular motion for extended period of time. In this scenario, accelerometer cannnot be used to correct for drift, as there is constant acceleration in addition to gravity. How to tackle this problem? Whether this problem is solvable at all in case of MEMS IMUs?
I understand that fusion softwares are proprietory solutions but I am not asking for any code here. I just want some explanations about how things are done in this respect.
Thanks for your inquiry.
Your summary for the theory of INS and attitude and heading reference system (AHRS) is correct.
The MEMS sensors that we make are for consumer electronics applications which means that the MEMS gyro, accel and mag are in consumer grade. In order to build an INS high-end sensors need to be used that have very little bias drift and very low noise.
In order to have precise AHRS the sensor fusion software needs to take care of accel vibration, mag calibration and gyro bias drift. Again the sensor fusion that we have developed is for consumer electronics applications.
I agree with you that MEMS sensors currently available are not military/tactical grade ones.
But I have purchased one BMI088, one of the recommended applications is drones. It is true that INS needs very accurate sensors especially in GPS denied environments, but implementing an AHRS should not be an issue.
Even if drone is a toy class, for autonomous flying all three euler angles needs to accurate enough (especially roll and pitch) and without any major drift. My only point was in case of long duration external accelerations, accelerometer cannot be trusted for absolute attutude estimation hence there will be nothing to stop the drift in estimated angles. If we see the current scenario, all the flight controller boards use these consumer grade MEMS sensors. So, I just wanted to understand the basics behind the drift problem in case of external acceleration. I would be grateful if you can refer to some books/papers which I can refer to and understand basics.
You can find a lot of articles online when you search keywords "MEMS drone". Here is one article written by a colleague in Bosch Sensortec at https://www.fierceelectronics.com/components/mems-sensors-are-heart-a-drone.