Adaptive simultaneous evaluation of both compass and GPS direction.
Gathering feedback
Currently, If automatic switching is set ON, compass direction switches to GPS direction or back at a given speed value .
It could make sense to perform processing of both simultaneously, e.g. by well set Kalman filter. It would give an adaptive, parameterless direction provider, decreasing direction error by combining 2 independent sources, especially for the transition speed range.
I like this idea 
Direction error of the compass is more or less constant wrt speed, but can vary due other factors.
Direction error of the GPS progressively grows with decreasing speed.
In current implementation, optimal speed to switch is the speed where both errors are the same. But as GPS direction error depends also on GPS HDOP ( position accuracy ), the optimal speed for switching compass and GPS direction is variable.
Even if we managed the dynamic computing of that optimal speed, the adaptive Kalman direction provider would provide accuracy better even then the dynamic ideal case.*]
At high speed it will have accuracy of GPS direction, at near zero speed accuracy of compass, at speed near switching speed will have lower standard deviation than both compass and GPS.
*] That dynamic ideal case means, if the speed based switching would be dynamically computed from measured compass error and current HDOP.
Direction error of the compass is more or less constant wrt speed, but can vary due other factors.
Direction error of the GPS progressively grows with decreasing speed.
In current implementation, optimal speed to switch is the speed where both errors are the same. But as GPS direction error depends also on GPS HDOP ( position accuracy ), the optimal speed for switching compass and GPS direction is variable.
Even if we managed the dynamic computing of that optimal speed, the adaptive Kalman direction provider would provide accuracy better even then the dynamic ideal case.*]
At high speed it will have accuracy of GPS direction, at near zero speed accuracy of compass, at speed near switching speed will have lower standard deviation than both compass and GPS.
*] That dynamic ideal case means, if the speed based switching would be dynamically computed from measured compass error and current HDOP.
failed graph attempt :)
failed graph attempt :)
Replies have been locked on this page!