Support different IMU sensors as input

[louisoutin]

Hello,

First of all, thank you for your work (paper + code), it's very well explained and the approach looks elegant.
I would like to push the work further by trying to make the method support multiple IMUs (you mention it on future works btw). I was wondering if by just keeping the same approach and architecture, but just adding different IMU source as inputs during training would be enough to leverage "multi IMU source"? Did you made some experiments on this already?
I would really like to hear your thoughts on this.

Thanks!

[mbrossar]

Hi,

I did not made experiment with multiple IMUs, but I think you have two possibilities:

• Keep the same approach and architecture, and add different IMU as inputs. The advantage of that method is that it is relatively simple. The problem is that during training the neural networks need to be able to compute the orientation from an IMU to another.
• Have multiple IMU systems with their own systems and them fuse the measurements with e.g. an extended Kalman filter, see e.g. this paper. I would recommend that approach, where the Kalman filter estimates the relative poses between the IMUs (you can also try to train a neural network for doing that)