Relative position control should take into account yaw.

[david-boles]

Confirmed with Prof. Tellex, the math for position control using the previous image needs to account for the yaw of the drone when that image was taken. Also, the current pose's yaw should be updated. I believe, for x, y, theta in the global coordinate frame, and dx, dy, dtheta in the previous image's coordinate frame, the math looks like:

theta' = theta + dtheta
x' = x + cos(theta)dx - sin(theta)dy
y' = y + cos(theta)dy + sin(theta)dx

Scaling between the coordinate frames should be the same because altitude is taken into account before this math happens. Also, this obviously still assumes the camera is pointing straight down.