Implement ProcessorMotion::processKF
Whenever a processor creates a KF, this KF is initialized with the state according to the PM::getState
.
However, until the next motion capture is processed, this KF is not "linked" to the graph with the motion (motion factor is established with the previous KF and the new KF is set as the new origin).
Then, if the problem is solved before this new motion capture is processed, the state of this new KF can "jump" considering only the factor(s) established by the processor that created this KF. This produces a very discontinuous output of getCurrentState()
.
To avoid this, the ProcessorMotion::processKF()
could be implemented. That would "link" this KF to the graph with motion factors just after its creation and without need of waiting a new motion capture to be processed. Also, it may simplify the implementation of processCapture
.