Multi-Object Behavior Modeling of Space Systems

In response to the DoD SBIR 2023.3 Solicitation for the Space Force Phase I Program, topic F233-0014 “Multi-Body Behavior Modeling of Space Systems,” Advanced Space, LLC (Advanced Space) proposes a comprehensive feasibility study applying machine learning (ML) based approaches to investigate and develop tactical-speed analytical capabilities that characterize and assess multiple objects and threats in space. Critical to such applications is an automated intelligence agent (or agents) that can assess real-time data intuitively, make informed and contextual decisions, and recommend the best course of action to a human operator who ultimately decides whether to pursue the recommended action. In other words, we aim to reduce the burden on the human-in-the-loop requirements throughout the entire data analysis process, and strategically automate certain aspects of this process to increase tactical response capabilities.  A fundamental component of ensuring freedom of operation in space is space battle management and command and control (SBMC2), with a core tenet therein being the ability to conduct real-time analysis of the risk to mission. This is necessary to allow action to be taken accurately, knowledgeably, and on appropriate timelines.  Collected data can be used to provide information to form hypotheses, but these hypotheses must have meaningful and critical context for BMC2 decision-making. Traditional approaches to modeling multi-object and/or campaign-level activities is time consuming and labor intensive, carries high levels of bias risk, and does not easily accommodate iteration for comparative and/or “what-if” analysis.  The improved and new capabilities will support more complex analyses on tactically-relevant timelines and enable a robust risk assessment to include likelihood and consequence of events. The technology to be developed will consider multiple behavioral and physics inputs, perform threat and impact analysis beyond single object and event actions, and present assessment results in mission-time for critical decision making.  Ms. Kanak Parmar will be the Principal Investigator (PI) for the proposed project. She has an extensive background in strategic applications of AI/ML algorithms for various phases of standard space mission operations. Ms. Parmar is a Staff Astrodynamics and Satellite Navigation (ASN) Engineer, as well as an AI/ML Engineer, at Advanced Space. Through various projects and prior research, she has experience in applying ML to trajectory station keeping and path planning; robust autonomous optical navigation; soft data characterization schemes; and human-AI cooperative/interactive learning approaches in multi-objective, contextualized, and priority-based situations. She will be supported by Dr. Nathan Ré, Principal ASN Engineer, who is a subject matter expert (SME) in low-thrust trajectory optimization and the use of neural networks for autonomous spacecraft operations.