Initially focused on landmark distortion (or dewarping) effects, and evolved into leveraging a constant curvature constraint. Used to compare a novel "circular motion estimation" method against a naive RANSAC approach, and formed the final part of my DPhil research in improve radar-only egomotion estimation accuracy.