After going through some tests we got weird results.
We tried to tune them with the classic Ziegler-Nichols method.
The proportional gain (POS_XY_P) did not seem to have any effect. Even when we put a gain of 2000.
The derivative gain (VEL_XY_P) when little (~0.5-0.8) seemed to work pretty good to maintain position after pushes. Similar to the video uploaded: https://www.youtube.com/watch?v=doYs5kI83jc
However when ouputs where given through the joystick to change the position of the rov, similar as we would do to the depth hold mode to give a new depth, the rov just started moving into a weird direction.
The integrator gains (VEL_XY_I and VEL_XY_IMAX) had to be set to 0 or the ROV also moved weirdly.
I know the position flight modes are still in development an maybe the PID is just not adapted to the ROV however I could not understand why setting the proportional gain to really high numbers did not make the ROV go crazy.
I noticed that the max value was 2.0 but you could force the saving. Maybe there is a security in the code to avoid setting too high numbers.
Maybe I did not take into account some parameters or I misunderstood those I used. If you have got any comments do not hesitate in responding