OUSD (R&E) CRITICAL TECHNOLOGY AREA(S): Trusted AI and Autonomy
The technology within this topic is restricted under the International Traffic in Arms Regulation (ITAR), 22 CFR Parts 120-130, which controls the export and import of defense-related material and services, including export of sensitive technical data, or the Export Administration Regulation (EAR), 15 CFR Parts 730-774, which controls dual use items. Offerors must disclose any proposed use of foreign nationals (FNs), their country(ies) of origin, the type of visa or work permit possessed, and the statement of work (SOW) tasks intended for accomplishment by the FN(s) in accordance with the Announcement. Offerors are advised foreign nationals proposed to perform on this topic may be restricted due to the technical data under US Export Control Laws.
OBJECTIVE: The objective of this work is the development of techniques that will enable the estimation of a space object's behavior/intent in close-proximity scenarios. Space domain awareness (SDA) is often described as the characterization of available information in the space environment in a meaningful way. For instance, measurements of an object in the space environment may yield a ``state estimate" of that object (e.g. some position, velocity, and attitude all with some corresponding uncertainty). The belief state of this object, via orbital dynamics knowledge, can then be further propagated into the future. This actionable knowledge enables decision-making in the space domain, such as a collision-avoidance maneuver or a reorientation of a high-valued asset. Conventionally, these belief states represent the core of SDA. However, it is becoming necessary to analyze the available data in the space domain at a higher level - not only where an object is/going, but why it is there/maneuvering. With the assumption that there is some agency behind the control of a space object, what tools and algorithms can be developed on available data that will enable the precise estimation of that object's intent? The ability to estimate both an object's intent and its state simultaneously will enable more-informed decision-making in the space domain. Furthermore, we seek solutions that enable these methods to be implemented on-board a spacecraft, enhancing its autonomous capabilities.
DESCRIPTION: Although there does not yet exist a collectively-agreed upon definition of autonomy amongst academic circles, the core idea is often some variation on the following: a machine-driven system that (i) receives data, (ii) interprets that data into some form of knowledge representation, and (iii) uses that knowledge representation to make decisions and accomplish some predefined task without human input. As the space domain becomes increasingly contested and congested with a rapidly-growing population of space objects there is a need to improve the on-board autonomous capabilities of high-valued assets. A promising avenue to do this is through informed decision-making based on the state and intent estimates of nearby objects. While state estimation techniques have been studied for decades, intent estimation is a relatively unexplored area. The ultimate aim is to infer the intent of an agent from available data (e.g. sensor observations, process dynamics, historical patterns of behavior). There are many open research questions regarding this topic - what types of uncertainty (aleatory, epistemic) are most applicable to intent characterization? Given that many forms of space-based data offer ambiguous interpretations of intent, i.e. available evidence may point to multiple mutually-exclusive hypotheses simultaneously, how can this be leveraged with current mathematical frameworks, such as Kolmogorov's axioms of modern probability or belief function theory? What information theoretics (e.g. Kullback-Leibler divergence, Mahalanobis distance, etc...) can be utilized to yield intent estimation metrics? Thus, the objective of this SBIR is to investigate these research problems and develop mathematically-rigorous algorithms that supply intent estimates of nearby space objects. Offerers should specify in their proposals what government-furnished property and/or data is required to conduct this effort.
PHASE I: Conduct a comprehensive comparative assessment and trade-off study of various intent estimation approaches. Define metrics that indicate the precision/success of intent estimation techniques and acknowledge the computational complexity of employing investigated techniques on-board space-grade hardware. Investigate the effects of intent estimation for on-board autonomous decision making.
PHASE II: Improve and iterate upon the most promising and effective intent estimation method. Conduct performance analyses on available space asset data and test on-board algorithms on space-grade hardware in the AFRL/RV laboratory environment.
PHASE III DUAL USE APPLICATIONS: Develop flight-ready intent-estimation software that can be employed into future AFRL or government space missions and experiments.
- Balch M.S., Martin R., and Ferson S., (2019), ``Satellite Conjunction Analysis and the False Confidence Theorem," Proc. R. Soc. A.4752018056520180565;
- Lai, Y., Paul, G., Cui, Y. et al. (2022), ``User Intent Estimation During Robot Learning Using Physical Human Robot Interaction Primitives." Autonomous Robotics 46;
- Cramer M., Cramer J., Kellens K., Demeester E., (2018), ``Towards Robust Intention Estimation Based on Object Affordance Enabling Natural Human-Robot Collaboration in Assembly Tasks," Procedia CIRP, Volume 78, Pages 255-260, ISSN 2212-8271
KEYWORDS: Autonomy; Intent Estimation; Statistical Inference; Data-Driven Decision-Making