SHERLOC
Automated Inference, Sensemaking, and Threat Assessment for Space Domain Awareness (SDA)
In order to help space domain awareness (SDA) decision-makers identify and prioritize threats, Stottler Henke’s SHERLOC tools are designed to use automated reasoning from multi-source evidence to help interpret the current situation in real-time based on data and inference from past cases.
SHERLOC prototypes were first developed and tested in the Sprint Advanced Concept Training (SACT) exercises and Dragon Army operational events hosted out of the Catalyst Campus in Colorado Springs, for cases including breakup detection and the investigation of space weather causes for satellite anomalies. This application involves a probabilistic framework for interpreting otherwise hidden interrelationships between multi-source input data. Prior probability distributions for combined evidence factors form the basis for threat assessments in new situations, with the added benefit of explainability in terms of relevant similar prior cases. Different kinds of threats are encoded with a representation of multiple evidence factors, each functioning as an independent variable based on specific inputs, such as the separate evidence coming from different phenomenologies or level one analytics. Inputs include sensor data (electro-optical, passive radio frequency) and human inputs such as characterizations from publicly available information. In operation, these evidence factors are populated either automatically via machine-to-machine interfaces, or manually with operator input.
In a subsequent effort, further development of SHERLOC prototypes has produced a similarity-based inference tool for automated reasoning and visualization using a relational network. Developed in conjunction with the Apollo Accelerator cohorts in the SDA Tools, Applications, and Programming (TAP) Lab in Colorado Springs, SHERLOC has been integrated with multiple subsystems in the SDA TAP Lab’s developmental space battle management framework. This tool is designed to help space domain analysts and operators avoid operational surprise from uncertainties regarding specific objects in space, by making automated inferences from similar objects. It uses a relational network approach to representing space objects, along with a similarity function that works like a “friend recommender” in social network systems. As a component of the larger battle management architecture, SHERLOC ingests machine-to-machine object data from multiple upstream sources, and generates automated inferences for downstream tools. The tool also includes a human-machine interface to support operator-directed queries and browsing in the relational network. Example use cases explored in existing research include identifying objects similar to known threats, resolving unknowns about space objects, and finding relationships between multiple objects.
SHERLOC Similarity-Based Inference Tool, from SDA TAP Lab Cohort 5 Demo Day, Jan 2025