Model based pose estimation for autonomous operations in space

Abstract

Presents a method for determining the 3D pose of a known object with the intent of autonomously controlling space hardware systems. Operations of interest include satellite on-orbit servicing and orbit transfer. Using a specially designed, adaptable vision server, stereo images are acquired and edge data is extracted. Corresponding points are found in the left and right edge data and 3D information is computed. This 3D data is then registered to a model of the object in such a way as to minimize the distance between the two data sets. Once the model is aligned with the data, the full 3D pose of the object is known; this information can then be used by a robot controller to compute a path to the object. The process runs continuously, giving the controller up-to-date information. Experimental results presented in this paper show the capabilities and robustness of the system.