Vision-based navigation and hazard detection for terrestrial rocket approach and landing

This paper introduces a compact optical payload enabling vision-based navigation and hazard detection during terrestrial rocket flights. This payload, the Terrain-Relative Navigation & Descent Imager (TRNDI), serves as an add-on module for Draper Laboratory´s GENIE Autonomous-GNC system. GENIE´s inertial navigation system relies on GPS during terrestrial test flights, making it insufficient for GPS-denied, extraterrestrial environments. The TRNDI system includes a horizon-pointing monochromatic camera, two downward-pointing monochromatic cameras, and a robust single-board computer. TRNDI utilizes a modular software framework, which allows multiple algorithms to run in real-time and communicate with the primary GENIE flight computer. This system enables data collection and real-time data processing for further development and evaluation of vision-based terrain-relative navigation, visual odometry, and hazard detection algorithms, with a path toward future closed-loop integration with GENIE´s navigation system. This paper presents an overview of the TRNDI algorithms and hardware, as well as simulated results from two vision-based hazard detection (HD) algorithms. The first HD algorithm searches the camera field of view for suitable landing sites. The second HD algorithm creates a map of the relative safety of the landing region. The simulated results demonstrate that TRNDI can reliably find and deliver a safe landing site to GENIE in realtime during a planetary approach trajectory.