A helper class that uses an underlying RandomSampleConsensus<TModel> to select the best pair of linear and angular velocities from a buffer of recent timed poses.
RANSACVelocity allows for the stable estimation of velocity under noise-prone circumstances. This is particuarly important when throwing objects because of the unreliability of data around the moment of intended release: if the perceived moment of release is slightly too late, for example, the behavior of the throwing implement (hand, controller, etc.) should not impact the calculated velocity. RANSAC velocity estimation across several frames mitigates this problem.