Visual Navigation Advances for Low-altitude, Subsonic Flight using WOSA Design Principles

KEF Robotics proposes to adapt their commercial visual navigation software, Tailwind, for high-speed aircraft and weapons systems, specifically targeting the Air Force's Franklin Affordable Mass Missile (FAMM) and Extended Range Attack Munition (ERAM) programs. The proposal focuses on Modular Open System Architecture (MOSA) and Weapon Open System Architecture (WOSA) compliance, particularly with the All Source Position and Navigation (ASPN) messaging protocol. Tailwind, already commercially available and flown on over 45 aircraft, provides Visual Inertial Odometry (VIO) and map-matching capabilities. However, it currently operates on slower-speed platforms. This Phase II SBIR project aims to enhance Tailwind's performance for subsonic, low-altitude, high-speed scenarios. The technical objectives include: 1) Improving Machine Learning (ML)-based approaches for visual feature tracking and map matching. Datasets will be collected and simulated to train neural networks robust to motion blur, image noise, and low thermal contrast. Algorithms will be optimized for real-time performance on edge computing devices. 2) Integrating visual navigation frontends into a pntOS framework and characterizing system navigation performance. Multi-camera VIO and Adaptive Learned Inertial Odometry (ALIO) will be developed and exported as ASPN messages. 3) Deploying and testing MOSA VisNav modules in simulation and captive carry. Photorealistic simulations will model flight dynamics, and captive carry tests will validate performance on a MOSA-compliant test vehicle. Key results include optical flow tracking with low pixel error and high output rates for trackers and map matching, and improved position error estimation in challenging conditions.