Aster Saul, Virginia Tech; Kevin Schroeder, Virginia Tech
https://doi.org/10.64861/AQEQ6266
Keywords: LWIR,microbolometer,hyperspectral,tasking
Abstract:
Current Space Domain Awareness (SDA) sensor networks rely heavily on ground-based optical sensors operating in the visual to short-wave infrared (VSWIR) range. Visual sensors can only operate between dusk and dawn, and can only observe targets that are illuminated by the Sun. Targets in low Earth orbit (LEO) spend up to 40% of their time in shadow, decreasing to approximately 5% at seasonal maximum in geostationary orbit (GEO). These lighting constraints reduce the effective operating time of a sensor for LEO targets to just a few hours a day, increasing the target estimation error, revisit time, and potential for loss of custody. Recent developments in extending visual sensitivity into the SWIR range with uncooled detectors enable some detections when the sensor is in daylight, but the lighting constraint issue remains for most applications. Long-wave infrared (LWIR) sensors unlock the upper end of the optical spectrum where targets emit thermal radiation and the atmosphere is somewhat transparent. LWIR sensors do not have the same lighting constraints as VSWIR sensors because targets emit thermal radiation in LWIR wavelengths independent of their illumination by the Sun. Increased target signal and lower daylight atmospheric noise can also enable daylight observations similar to SWIR. This work will show that LWIR-augmented SDA networks should be able to achieve lower revisit times and state estimation error than current VSWIR-only networks.
Historically, LWIR sensors have not been widely used for SDA due to significant hardware barriers to entry. Traditional cooled LWIR sensors are not new, and their use for SDA has been studied extensively. These are typically exquisite sensors mounted on large telescopes and have significant integration issues with SDA hardware due to their cooling requirements and weight. Additionally, astronomy-sized optics with LWIR-transparent lenses do not exist commercially, requiring such telescopes to be constructed entirely from mirrors. This places an upper limit on several telescope performance factors, where the most significant is the high focal length and narrow field of view (FOV). Information collection rate is proportional to FOV, so narrow FOV systems are typically suboptimal for large-scale SDA search applications. Future SDA networks will likely be composed mostly of proliferated small telescopes with wide FOV VSWIR sensors. This work will examine augmenting these networks with LWIR sensors tasked separately to take advantage of their abilities.
Microbolometers are an alternative type of uncooled sensor that offer easier integration into remote autonomous SDA systems due to their low SWaP and complexity. These are currently underexplored in SDA because commercially available microbolometers have only recently been able to achieve low enough noise to be viable for SDA applications. This work will present a physics-based signal-to-noise ratio (SNR) model for microbolometers to evaluate their effectiveness in SDA networks.
We fully simulate an SDA network as a Partially Observable Markov Decision Process (POMDP), where sensors are tasked on certain optimal boresights to minimize target revisit times and state estimation error. Realistic operational target sets containing mixes of LEO and GEO targets are varied with networks containing a range of VSWIR/LWIR sensors. We present metrics on revisit time and information gain between scenarios and propose an optimal VSWIR/LWIR sensor ratio. We implement SNR models for VSWIR and LWIR sensors and validate our microbolometer models against operationally relevant images of various targets collected with a LWIR microbolometer on a 0.3-meter telescope.
Date of Conference: September 15-18, 2026
Track: SDA Systems & Instrumentation