Minimizing Rocket Landing's Trajectory Following Error via Visual Guided Wind Compensator

Abstract

Current rocket landing missions, just as demonstrated by Starship from SpaceX can not rely on GPS alone for a realtime maneuver that needs a very high guidance and control computation, let alone reaching target with pinpoint accuracy. GPS' low update rate must be overcomed by incorporating other navigation devices. A proposed visually augmented precision landing (VAPL) guidance system directs the landing rocket model (LRM) to conduct vertical-takeoff vertical-landing (VTVL) maneuvers. Higher frequency rate, lower cost, and a better ability to recede interference are main benefits of vision-augmented navigation. Moreover when combined with other navigation sensors, such as GPS/INS, LIDAR and ultrasonic, vision-augmented navigation provides higher guidance and positioning accuracy. One of main challenges must be faced by the landing guidance system is to predict and overcome any uncertainties such as wind disturbance that can degrade positioning accuracy severely. The proposed VAPL combines classical flight control system with wind compensator to strengthen the work of the guidance system in following the designated trajectory shape. Combination of OpenCV and YOLOv8 is employed to construct this guidance system by facilitating precise thrust vector control and intensive engine thrust adjustments based on visual feedback. This integration ensures landing pinpoint accuracy to circular error precision < 5 m radius, and minimize the trajectory following error.