3D LASSO: REAL-TIME POSE ESTIMATION FROM 3D DATA FOR AUTONOMOUS SATELLITE SERVICING

摘要

The recent development of space flight ready 3D sensors, such as the Neptec Laser Camera System (LCS), allows 3D vision technology to be considered for autonomous missions. These active sensors provide their own illumination and have a small instantaneous field of view, making them immune to dynamic lighting. Harsh and dynamic lighting conditions have severely limited the use of 2D passive camera based space vision systems for mission critical applications. Autonomous robotic servicing missions, such as the Hubble Rescue Vehicle (HRV), will require vision systems that are capable of providing high accuracy pose estimates in real-time while being robust to changes in lighting conditions. This paper describes the 3-Dimensional LCS Algorithms for Spacecraft Servicing On-orbit ( 3D LASSO) system currently under development at Neptec. The project is funded by the Canadian Space Agency (CSA) under the Space Technologies Development Program (STDP). The 3D LASSO system is designed to perform real-time tracking and 6 degree of freedom pose estimation of target spacecraft(s) from sparse and noisy 3D data. The approach is compatible with any sensor capable of providing 3D data. The algorithms have been successfully tested with Neptec’s LCS in a variety of test scenarios. Tracking was performed using the random access capability of the sensor which is used to perform rapid, sparse sampling of the target object(s). The data obtained is aligned to a reference model of the target(s) using a newly developed faster version of the Iterative Closest Point (ICP) algorithm developed at Neptec. The pose estimate obtained is then used to compute the trajectory of the object(s).