A Parallelized Iterative Closest Point Algorithm for Attitude Estimation

Abstract

In recent decades, low-earth orbital debris has become a major concern. If the density of objects in orbit is too high, and nothing is done to remove any debris, there could be a cascade of collisions, each generating more debris, increasing the frequency of collisions even further. This would render future space missions very difficult and dangerous, if not impossible. Something must be done to remove space debris from orbit. This thesis attempts to solve one piece of the orbital debris problem – that is, tracking a piece of debris and determining its attitude and position relative to a capture vehicle. A LIDAR camera is used to acquire images of the body to be tracked. To achieve a fast and practical solution, a parallelized Iterative Closest Point (ICP) algorithm and a Kalman filter are implemented to track the attitude and position of a model rocket booster. Additionally, this thesis presents a method to increase ICP's accuracy and reliability by artificially increasing image resolution by algorithmically increasing image size. This work also explores the performance of different variations of ICP and the dependency of their performance on image size.