adaptive gain seems not correct in aqua algorithms

[whoispo]

if alpha equals 0 then, alpha will be 0 forever. Because in the code, alpha updates:

Even if f equals 1, then new alpha will be 0.

I have read the reference article, the gain is not stored and used again.

[Mayitzin]

You are right. The gain wasn't being properly updated.

The commit de6e3632055a57c45b1d66b21113174e66dad37e has modified this part and its calls within the algorithm. The changes are:

• The final computation is made with an optional alpha value named alpha_bar (equal to 0.1) to be consistent with the original article.
• Because it is optional it avoids the problem of being called with the predefined self.alpha as it was.
• As mentioned, the alpha gain is re-defined in the function adaptive_gain() as the second parameter now. The first parameter is now the Measured local acceleration vector, as it always should have been.
• The calls to adaptive_gain() in the methods updateIMU() and updateMARG() are done now with the measured acceleration vector only.
• The computation of the gain factor $f(e_m)$ has been simplified using NumPy's function clip.
• The documentation is updated accordingly using $\overline{\alpha}$ for the constant value that gives the best filtering result in static conditions.

If you have further observations, let me know.

[Mayitzin]

Unit tests for the function adaptive_gain were added with commit fd4ad3b56a0115fabda2b7f5d17453ab5c4c94b5

I think we can close this issue.