Pose Estimation Using a Flash Lidar

Abstract

A pose (relative attitude and position) estimation approach suitable for proximity operations to support, for instance, on-orbit servicing missions or the autonomous landing of unmanned air vehicles is investigated in this paper. This approach is designed to harness the recent advances in flash lidar (light detection and ranging) sensors, capable of providing fast frame rate (>30 frames per second), large field of view (> 30 deg square) point cloud measurements. Unlike the scanning lidar, the flash lidar has the ability to capture an entire scene with a single laser pulse and can help reduce distortion in the point cloud data if the object is rotating or translating. Pose estimation algorithms based on the homogeneous transformation (HT), which has been employed extensively in robotics, computer vision, and computer graphics, were developed for translation only, rotation only, and combined translation/rotation cases. The HT-based algorithms provide an alternative to the conventional iterative closest point (ICP) algorithms in defining and evaluating the error metric for the pose estimation, allowing a simpler least squares solution. Using a Swiss Ranger flash lidar in a laboratory testing environment, preliminary pose estimation tests were conducted and promising results were obtained although there remain some unanswered questions such as how to define the initial alignment between the target object and the sensor, how to obtain the truth target movement knowledge in the sensor frame, and how to overcome the small rotation angle limitation in the algorithms.