Jeremy Correa, Katalyst Space Technologies; John Ware, Katalyst Space Technologies; Maruthi Akella, UT Austin; Brandon Jones, UT Austin
Keywords: Dynamic Tasking, Cooperative Decision Making, Cislunar OD
Abstract:
Recent interest and activity in cislunar space have highlighted the need to extend space domain awareness (SDA) capabilities to ensure safe operations for existing and planned missions in geostationary orbit and beyond. Effective SDA in the cislunar regime requires the ability to search for, detect, and predict the trajectories of cislunar space objects, including those previously unknown. This challenge is further complicated by non-cooperative space objects that may execute unanticipated maneuvers or reveal concealed capabilities, making conventional tracking and sensor management strategies less effective. Due to these compounding challenges and increased mission complexity in the cislunar regime, the imperative for system architectures that support automation of tactical decision making and sensor tasking becomes increasingly clear.
Although there are numerous scenarios in the cislunar regime whereby non-cooperative space objects can impede tracking efforts, this paper focuses on a reference scenario that involves tracking multiple space objects in transit between the Earth and Moon with the capability to execute impulsive maneuvers and deploy unknown objects. The paper outlines a framework designed to provide automated sensor tasking and facilitate the maintenance of tracking data for multiple target objects. To establish this framework, research efforts were focused on: (1) developing locally optimal sensor tasking strategies for space-based panchromatic (VIS) sensors, (2) developing multi-target observation association techniques to detect non-cooperative tracking events, and (3) designing a scalable and modular software architecture with open architecture concepts to enable flexibility, adaptability, and robustness in simulating cislunar space operations.
The approach used to construct sensor tasking schedules leverages a semi-Markov Decision Process (SMDP) as the fundamental mathematical framework. Situations are considered where target states are well-known, direct observations are feasible, and localized searches of targets when covariance growth exceeds manageable thresholds. Over a finite planning horizon, we are able to generate locally optimal sensor tasking strategies that aim to minimize tracking uncertainty and improve the quality of state estimates of non-cooperative space objects. The approach enables cooperative tracking of multiple target objects through selective information sharing and aggregation of track data across multiple sensing platforms. The use of kinematic gating techniques and non-kinematic association heuristics were used to distinguish between closely spaced objects, thereby reducing the likelihood of track duplication and improving overall tracking accuracy. In this paper, target state estimation algorithms are limited to ingesting angles-only passive optical measurements and target detectability criteria is computed using a simplified panchromatic (VIS) radiometric model.
A core innovation of the work is the development of a methodology that combines state-driven policy adjustments and consensus algorithms to dynamically generate observation tasking priorities and adapt sensor schedules in real-time. Through state-driven policy adjustments, each sensor agent re-evaluates its actions based on the most current state information, allowing for immediate responses to changes such as reductions in uncertainty volume or new target detections. Consensus algorithms facilitate coordinated decision-making across multiple sensing platforms, ensuring efficient task allocation and minimizing redundant efforts. This method is computationally efficient and proves valuable in rapidly evolving multi-target tracking scenarios or when direct space-to-ground communication links are unavailable for real-time sensor tasking.
Date of Conference: September 17-20, 2024
Track: Cislunar SDA