Problem with providing occasional visual position corrections via a camera attached to the drone

Hello,

We are a group of engineering students working on our senior projects. Our project is developing a system for non-gps navigation. We are planning to use an optical flow sensor and also an image recognition module to correct the drift.
We have so far developed an image recognition module that outputs the drone’s position with a considerable accuracy, however these outputs are occasional since they are dependent on the terrain and landmarks drone is currently flying over.
Our problem is finding a way of feeding these occasional position updates into the ekf. These update are highly accurate but not that accurate that we can reliably just set our new position to them. We thought using the VISION_POSITION_ESTIMATE command on Mavlink would be problematic since we don’t have a fixed update rate.
We would appreciate any idea about implementing these corrections?

Hello @mmert welcome to the community,

It requires an update rate greater than 5Hz.

Thanks. Is there any other way we can try to feed our corrections somehow?

The EKF requires periodic updates. It might work down to 3 Hz, but not less than that.

ArduCopter uses and requires the EFK to be stable, there is no workaround that.

1 Like