Sparse Point-Plane SLAM

SLAM is a fundamental problem of mobile robotics and its most efficient version models the environment as set of sparse points, against which the camera is tracked. We explore the possibility of using geometric information in the form of planes to further improve the accuracy of tracking and mapping. Plane measurements are considered and planar constraints are introduced between points and planes on which they lie. We integrate planar measurements and constraints in a state-of-the-art feature-based SLAM system and show that this provides better trajectory estimates than points alone, and more meaningful maps, especially in environment with low texture and dominantly planar surfaces.