Analysis of waypoint range uncertainty

• Earlier estimates are more accurate in an absolute sense, but later poses have drifted farther from the absolute reference frame, nulling this advantage.
• Later estimates contain less uncertainty due to the travel from the estimate point to the projected waypoint, but that projection depends on the projection of the waypoint ray into our current FOV. This is a function of the larger relative uncertainty of the estimate point w.r.t. the waypoint base.

• Stereo range uncertainty decreases as the square of distance, so closer ranges are much more accurate than longer ones.
• The only reason that flat earth assumption is at all reasonable is that remaining segment of the terrain increasingly resembles a plane as we approach the waypoint.

Note that strictly speaking, the quality improves as we close on the waypoint, which is not necessarily the same as improving as time passes. We might be moving away from the waypoint due collision avoidance, etc. It nevertheless seems reasonable to approximate this effect by doing a running average that gives more weight to recent projections (e.g. a low-pass filter.)