Vision-based proximity navigation is an essential technology for space debris removal, in-orbit servicing or space missions around asteroids. It requires high autonomy and robustness due to the challenging for challenging situations of bad illumination, the large uncertainties and if the object is unknown, etc. It’s very difficult to manage with data from a single sensor or from a single data processing technique. This PhD research aims to fuse data from multiple onboard sensors to compensate for each sensor’s limitations and potential data failure, to generate highly robust and autonomous navigation solutions for different range scenarios. Specifically, we will focus on pose estimation including both position and attitude. Multispectral imaging cameras such as Visible, Near and Thermal Infrared ones, and LiDAR will be the mainly considered sensors for the fusion technique. We need to explore the proper fusion framework in terms of fusion level, tailored image processing algorithms, and weights allocations of different sensors, to generate accurate and robust pose estimation. In particular, we will couple related image-processing techniques with cutting-edge machine-learning techniques, to improve the autonomous level of the developed algorithms