As the value of the space domain has become more widely recognized, nations have begun developing weapons systems that target space assets from the ground and others that are themselves integrated into space-based assets. Increased space domain awareness (SDA) and maneuverability offer benefits that may mitigate certain adversarial strategies, at a relatively small cost to overall size, weight and power. The trajectories of space assets in Earth orbits are largely determined by launch conditions but can be modified at the expense of delta-v. Additionally, a kinetic attack from the ground or adversarial spacecraft requires that the two objects must be near the same location in space at the same time. Because of these two facts, kinetic attacks from the ground or from an adversarial spacecraft are severely constrained by physics and this information can be used to focus SDA coverage. This also implies that agents with more alert SDA and more precise maneuverability capabilities can impose unfavorable conditions on their targets or pursuers, as they can better assess possible intent given observed behaviors and react appropriately to create penalties in position, phasing, or required delta-v. For the most complex operations, modern space-based platforms require onboard imaging systems to detect, track, and navigate relative to other spacecraft with high precision. Under this development Astrobotic Technology, Inc. and Arizona State University will design an optical sensor that aims to improve SDA provided by space-based platforms and is also likely to offer notable performance benefits for reliable, on-orbit, GPS-denied navigation that powers maneuver sequences needed for satellite servicing and similar operations. The proposed sensor, RetiNav, combines computational imaging algorithms similar to methods used in consumer mobile devices, along with event camera sensor technology, and efficient space-grade computing. Event cameras are bio-inspired passive imaging sensors that measure changes in light intensity asynchronously using independent pixels. They generate streams of binary responses for each pixel when it experiences an intensity change. The high sensitivity and low power requirements will make RetiNav especially well-suited for imaging in HDR environments and in SWaP-constrained systems. Our approach will fuse high performance-per-watt embedded computing with a balance of past and contemporary algorithms and machine learning models to provide low latency, robust system-level processing. The proposed RetiNav sensor development focuses on the interaction of the sensor and neural network for processing events and establishing the feasibility of this processing on space-relevant computing hardware. We believe it has the potential to offer unique advantages for SDA and GPS-denied navigation in commercial and government space systems as a dual-use technology, as described in the 2020 Defense Space Strategy Summary.
