I have a set of 12 IMUs mounted on the end effector of my robot arm which I read using a micro controller to determine it's movement. With my controller I can read two sensors simultaneously using direct memory access. After acquiring the measurements I would like to fuse them to make up for the sensor error and generate a more reliable reading than having only one sensor.
After some research my understanding is that I can use a Kalman filter to reach my desired outcome, but still have the problem of all the sensor values having different time stamps, since I can read only two at a time and even if both time stamps will be synchronized perfectly, the next pair will have a different time stamp if only in the µs range.
Now I know controls engineering principles but am completely new to the topic of sensor fusion and google presents me with too many results to find a solution in a reasonable amount of time. Therefore my question, can anybody point me into the right direction by naming me a certain keyword I need to look for or literature I should work through to better understand that topic, please?
Thank you!
The topic you are dealing with is not an easy one. Try to have a look at the multi-rate kalman filters.
The idea is that you design different kalman filters for each combination of sensor that you can available at the same time, and use it when you have the data from those sensors, while the system state is passed between the various filters.