Abstract
In this paper, we propose a visual servoing algorithm for the automated docking of unmanned spacecraft. The proposed algorithm detects a set of custom visual markers using pyramid-based template matching for scale-invariance, and tracks the set of markers using a particle filter and the continuously adaptive mean-shift (CAMShift) algorithm for nonlinear movement. For fast execution, the proposed algorithm runs on general-purpose graphics processing units (GPGPUs) using compute unified device architecture (CUDA) and multi-platform shared memory parallel programming. The performance of the proposed algorithm is validated in flight tests using an indoor unmanned aerial vehicle (UAV) and a simulated docking adapters which showed satisfactory results.
