Simulation of Space Debris Capture Using a Game Engine with Physics-Based Flight Software

There is a growing interest in active debris removal technologies, as space debris poses a risk to critical satellite infrastructure and future space missions. A guided net mechanism is in development, wherein each corner of the net is guided to the debris via small CubeSats with cameras. We present a novel method for training and testing an image-based guidance, navigation, and control algorithm using Unity, a game engine. An object identification algorithm is generated using YOLOv4 (You Only Look Once), a popular object detection model, and realistic images of space debris generated by Unity. The algorithm is then used in a software-in-the-loop testing environment, using Unity to generate artificial scenes of debris, identify the debris, then navigate to the debris using a physics driven navigation algorithm implemented in F Prime. This methodology can allow for rapid validation of vision-based navigation systems.