Justin Kruger, EraDrive, Inc.; Alex Mescher, EraDrive, Inc.; Aviad Golan, EraDrive, Inc.; Sumant Sharma, EraDrive, Inc.; Simone D’Amico, EraDrive, Inc.
https://doi.org/10.64861/WJCZ9805
Keywords: Space situational awareness, space-based assets, autonomy, optical navigation, orbit determination, distributed space systems, space traffic management
Abstract:
The large number of satellites carrying onboard star trackers encourages their usage as opportunistic data collection platforms for space situational awareness (SSA), with the potential to collect more timely, more persistent, and more accurate data than ground-based assets. It is especially valuable if SSA data is processed onboard, such that products can improve autonomy and safety in real time. However, space-based SSA efforts have typically relied on large, expensive, dedicated satellites or on downlinking sensor data for on-ground processing, rather than performing SSA online.
The EraDrive autonomous SSA architecture proposes a novel capability in which distributed networks of space-based observers enable autonomous space traffic coordination (STC), at both the global and individual satellite level. The architecture consists of two layers: an online ‘Edge’ real-time optical sensing and processing module, and an offline ‘Fusion’ data aggregation procedure which maintains a global catalog of Resident Space Objects (RSO).
The Edge process runs in a decentralized fashion on board each observer within the SSA network. It detects RSOs in images acquired by the observer’s camera and compares observations with the RSO catalog available on board. Upon successful identity match, optical measurements are fused with other available sources (e.g., from other Edge modules) to produce refined state estimates for the observer and concurrently maintain the local RSO catalog. If the RSO is unmatched, a batch of measurements is collected over time to compute an on-orbit initialization. Refined state estimates are used to perform autonomous conjunction assessment and collision avoidance maneuver planning for the observer, if operationally allowed. Intermediate SSA products are shared with the ground and with other Edge platforms over a radio frequency network.
The Fusion process runs in a centralized fashion on the ground. Downlinked Edge products from observers are fused with measurements and RSO catalog updates from external third-party providers. Subsequently, an updated RSO catalog is produced considering all sources of measurement information. Fusion products are uplinked to observers at a regular cadence, and may be accessed by operators to reduce conjunction assessment uncertainties.
Key components of the EraDrive autonomous SSA system are being tested aboard the NASA Starling swarm mission. Starling consists of four identical 6U propulsive CubeSats launched in July 2023 into a near-sun-synchronous Low Earth Orbit (LEO). Each satellite carries two commercial off-the-shelf star trackers and S-band software-defined radio crosslink to demonstrate and mature EraDrive’s Edge module. EraDrive’s flight software is embedded into an ARM dual-core 700 MHz microprocessor and offers four core functions: Image Processing (IMP), Batch Orbit Determination (BOD), Sequential Orbit Determination (SOD), and Maneuver Planning (MAP).
When a new star tracker image is received by IMP, the orbits of each RSO in the catalog are efficiently propagated to the current epoch. Detected objects are matched to identities in the catalog using statistical distance thresholds and kinematic multi-hypothesis tracking. The BOD module robustly computes initial target orbit estimates from bearing angle batches, leveraging iterative batch least squares refinement or a linear system solution if multiple observers are present. SOD employs an adaptive, efficient nonlinear sequential filter (NSF) to refine target states, which maximizes observability through nonlinear dynamics and measurement models. RSO measurements are used to refine orbit states within the NSF, which are subsequently stored in the onboard catalog. Finally, MAP utilizes states from SOD to plan attitude and translational maneuvers to achieve control objectives (e.g. centering a desired RSO in the camera FOV, absolute or relative orbit-keeping, or collision avoidance if minimum separation bounds between the observer and a target are violated).
Flight experiments throughout 2025 and 2026 have demonstrated crtical capabilities of the Edge system for autonomous SSA. Thousands of unique RSOs were detected and identified on board across intersatellite distances (ISD) ranging from tens to thousands of kilometers. Measurement errors were less than 100 m per 100 km of ISD on average, indicating significantly better accuracy than typical two-line element (TLE) products. An onboard attitude planner was able to autonomously select RSOs of interest for observation, based on distance and visibility criteria, and compute Euler angle commands to center targets in the star tracker field of view. Tens of candidate RSOs were imaged over a one-day period through entirely autonomous tip-and-cue of the spacecraft, greatly enhancing the field of regard.
Measurements were subsequently used to refine stale catalog orbit information on board. For Starlink satellites, position errors were reduced by 50% on average, compared to on-ground reference TLE products. For persistent swarm targets, along-track position errors of multiple kilometers were reduced to hundreds of meters, with relative position uncertainties falling below a 1% ISD threshold. Measurements were also shared between multiple Edge modules over the crosslink to enable rapid initialization of new RSO orbit estimates in flight. Using short arcs of bearing angle measurements, target relative position errors remained below 3% of the ISD. These first-of-their-kind flight experiments validate the potential of leveraging distributed satellites as autonomous SSA providers with benefits for downstream STC.
The advantages of integrating proliferated Edge modules with the globally centralized Fusion network are demonstrated through high-fidelity simulations of the complete RSO catalog, calibrated using Starling flight data. Approximately 400 Edge instances distributed across LEO altitudes were able to achieve 99% coverage of the LEO RSO population, with mean revisit times of less than 10 minutes and and nominal RSO positioning errors of tens of meters, superior to typical SSA products. The proposed EraDrive architecture therefore provides multilayered benefits for the safety and autonomy of individual satellites, as well as space sustainability at scale.
Date of Conference: September 15-18, 2026
Track: Space-Based Assets