The rapid deployment of large Low Earth Orbit (LEO) mega-constellations is driving the need for a new generation of Precise Orbit Determination (POD) solutions that remain accurate, robust, and computationally scalable across thousands of spacecraft. The ESA-funded POD for LEO Mega-Constellations (POD4LEO-MeCo) project addresses this challenge by investigating advanced architectures for both real-time onboard navigation and high-precision ground-based POD. The project targets sub-decimetre real-time navigation performance and centimetre-level post-processed products, supported by scalable estimation strategies distributed across space and ground segments.

The study evaluates a broad set of enabling technologies, including multi-frequency GPS/Galileo, high-accuracy and high-rate GNSS corrections, Inter-Satellite Links (ISL), Ground-Satellite Links (GSL), and distributed estimation and processing algorithms. Building on these elements, a suite of Advanced POD (APOD) concepts has been defined, prototyped, and benchmarked to overcome the scalability limitations inherent to classical POD methods when applied to future LEO mega-constellations. The APOD concepts are organised into two complementary groups.

The first group comprises three onboard navigation concepts based on reduced-dynamics Extended Kalman Filters (EKF). To determine the minimum achievable threshold with current technologies, the first concept follows a traditional GNSS-based approach using GPS+Galileo code and phase measurements, progressively incorporating broadcast ephemerides, Galileo HAS corrections, and ultimately low-latency high-accuracy GNSS corrections disseminated in near real time. The second concept exploits only one-way ISL and GSL observables, enabling autonomous navigation for communication-type constellations by combining ISL ranges with neighbour-state information and sparse GSL updates, similar to a representative mega-constellation in LEO for communications. The third concept, combining both first two concepts as a potential on-board navigation technology, integrates GNSS, ISL, and GSL to support absolute positioning with the strong geometric relative ranging information provided by the latter two.

The second group focuses on post-processing weighted batch least-squares POD for large-scale network scenarios and also consists of three concepts. The first concept decomposes the mega-constellation into subnetworks (e.g., orbital planes), performing standalone GNSS-based POD per subnetwork before applying ISL-driven network-wide refinement, thus mitigating the cubic computational growth of classical approaches. The second concept generalises the GNSS ground-network POD problem to scenarios involving 150–1,000 tracking stations and optional GNSS-constellation ISLs, using partitioned solutions that are subsequently combined. The final concept integrates the two previous concepts into a system-of-systems solution that jointly estimates GNSS MEO satellites, a LEO-PNT constellation equipped with GNSS and ISL capabilities, and a global ground network.

All APOD concepts are being implemented, simulated, and benchmarked in a dedicated sandbox environment running a high-fidelity simulator based on the GMV FocusPOD library. The paper will present the POD4LEO-MeCo architectures, the APOD concepts as well as their implementation, and performance results, assessed through a set of Key Performance Indicators (KPI).