Now I'm thinking about
1) detecting when the solution is bad. Now this can be easily done after the solution by detection how it fits the LOBs. In some rare cases it can also be detected from inputs (Ownship course with no legs) and in one very very race case it can be detected inside the formula (determinant=0).
2) Adding more limits, especially speed or distance. That would allow to get solution even when you don't have leg. But at the first glance it means to combine this matrix with quadratic solution .. currently I have no idea how to do it. To set Vx and Vy to certain values would be simple, but to add vx*vx+vy*vy=certain value is a problem.
Anyway this could be easily solved numerically, by simply trying more possible courses and returning solution with best error. It's similar with distance.
|