| Spacecraft are expensive to build and launch. Once in orbit a spacecraft is very expensive to service and currently the only effective way of accomplishing this task is by sending astronauts to service them. In the case of the Hubble Space Telescope there have been two service missions. These missions are very dangerous to the astronauts as they must leave the relative safety of the well shielded Space Shuttle for space walks to perform repairs. Robotic systems that can be used to extend the operational life of spacecraft without placing astronauts at risk are of considerable value. This thesis presents a vision-based system that addresses a fundamental technical issue related to autonomous spacecraft servicing: It addresses the problem of obtaining accurate pose estimation of a spacecraft. Utilizing the known 3D structure of the spacecraft and edge features extracted from an image captured of the craft, a genetic algorithm is developed to obtain relative pose estimation. Experimental results with mockups of space hardware including a latch from the Hubble Space Telescope and a grapple fixture demonstrate the effectiveness of the approach. |
