COSINE: Context-aware Opportunistic Sensing for Indoor Navigation Environment

COSINE is at the forefront of innovation in the field of Alternative Positioning, Navigation, and Timing (A-PNT), utilizing machine learning and a diverse array of sensors such as IMU, camera, audio, maps, and Wi-Fi (or Signals of Opportunity) to ensure robust 3D localization, even in the absence of reliable GPS. Currently leveraging the capabilities of Android and AWS (Amazon Web Service), COSINE has proven its real-world applicability through successful pilot projects at a major US casino. Its versatility is further demonstrated by planned deployments and trials across a wide range of sectors, including military operations, convention centers, office buildings, and cruise ships. COSINEÆs ultimate objective is to redefine PNT by delivering sub-meter accuracy with real-time predictions under 5 seconds, across diverse mobility styles and in GPS-contested environments, while also enabling faster and more scalable machine learning deployment and management. This commitment to resilience and efficiency positions COSINE as a transformative force in the A-PNT landscape, a goal that is the central focus of sequential phase II. The innovations brought forth by COSINE include a unique repository of localization and navigation datasets collected in real-world scenarios for robust position estimation, a deep learning framework that leverages raw tri-axial inertial sensors for accurate positioning and navigation across diverse mobility styles, a deep neural network designed to rapidly detect and correct GNSS positioning errors, a deep neural network framework to estimate the position and orientation of various cameras by matching a query image or image sequence against terrain maps, a deep learning-driven framework designed to rapidly detect and correct magnetometer anomalies, and finally, an AI positioning solution for edge devices, powered by ML OperationsĀfor faster, more accurate on-device predictions.