Combination of Energy and Kalman cost functions
-----------------------------------------------

I wonder if there is some useful and significant way to combine the use
of the energy and Kalman cost functions.  I am thinking that the energy
cost function represents a-priori information about the tracks, in the
sense that a trajectory is likely to use minimum energy to get from one
point to another.

This is certainly ad-hoc, but may be interesting to investigate.

If the real measurements don't include velocity, we wouldn't use the same
energy cost function as in the K-path-viterbi paper.  But Kalman
estimates include both position and velocity, so one possibility would be
to assume that the measured data velocity equals the kalman-predicted
velocity, then compute an energy cost based on the difference in position
between the kalman predictions and the measured values.  That cost could
be scaled and added to the normal kalman cost, and would perhaps be
similar to adding log(P(data)) to a probability estimate.