Multi-Target Tracking using Random Finite Sets for Rendezvous and Proximity Operations with Non-Gaussian Uncertainties
The state-of-the-art in Multi-Target Tracking (MTT) is the use of Random Finite Set (RFS) filters with the recent theoretical innovations including the development of multi-sensor methods and the computational efficient Joint Generalized Labeled Multi-Bernoulli (JGLMB). Random Finite Set (RFS) theory is used in Multi-Target Tracking (MTT) application as a rigorous and “engineering-friendly” Bayesian formulation in which the collection of target state vectors is treated as a single finite set. Thus, this work focuses on implementing the JGLMB to the problem of tracking multiple space vehicles in Rendezvous and Proximity Operations (RPOs). While this has been performed previously with Gaussian Mixtures (GMs), this work developed the algorithms for the general Gaussian Scaled Mixtures (GSMs) in a JGLMB which will allow for multi-modal and heavy-tailed sensor noise modeling. These developments will provide the specific application of JGLMB filters for RPOs involving Earth, cislunar, and lunar orbits incorporating sensor measurements with non-Gaussian noise.

On-Orbit Planning and Guidance for Rendezvous and Proximity Operations Avoiding Debris Enabled by Random Finite Sets
In the increasingly cluttered orbital space about and beyond Earth, accessing target cooperative and non-cooperative Resident Space Objects (RSOs) via Rendezvous and Proximity Operations to service or deorbit them will necessitate precise tracking and planning through debris fields of varying density and size where this debris may be associated with the local environment or due to the breakup of the RSO itself. Thus, it is anticipated that in the future it will be necessary for spacecraft to augment any ground-based tracking and planning with on-orbit capabilities for tracking debris and path planning at high resolution. This work focuses on developing tracking and planning algorithms based on Random Finite Set (RFS) theory for spacecraft to avoid high-risk debris areas measured by its sensors in its path toward the target RSO. This work used two RSF filters in parallel for real-time on-orbit tracking of debris and feature tracking of the RSO which were passed to an Extended Kalman Filter (EKF) for attitude and angular velocity estimation to provide a docking localization and velocity information. These outputs were processed as part of a Receding Horizon (RH) Extended Linear Quadratic Regulator (RH-ELQR) in order to dynamically plan a path and guide the spacecraft through the field of debris to rendezvous with the RSO.

Related Publications
​

• V. Hill and J. D. Larson, “Multi-Sensor Fusion for Decentralized GPS-Denied Robotic Swarm Cooperative Navigation,” in AIAA SciTech 2023 Forum, January 2023
• J. P. Roux, S. I. Sheikh, C. S. Hisamoto, and J. D. Larson. ”Software Toolset for Localization and Control of Swarming Robots using Random Finite Set Statistics.” in AIAA ASCEND 2021, pp. 4025, November 2021
• V. Hill, R. W. Thomas, and J. D. Larson, “Autonomous Situational Awareness for UAS Swarms,” in 2021 IEEE Aerospace Conference 2021, April 2021
• R. W. Thomas and J. D. Larson, “Inverse Reinforcement Learning for Generalized Labeled Multi-Bernoulli Multi-Target Tracking,” in 2021 IEEE Aerospace Conference, March 2021
• SciTech 2021 R. W. Thomas, V. Hill, and J. D. Larson, “Hierarchical GNC for High Cardinality Random Finite Set Based Teams with Autonomous Mission Planning,” in 2021 AIAA SciTech Forum, January 2021
• J. D. Larson, B. Doerr, and R. Linares, “Autonomous Mission Planning for Swarms Using Random Finite Sets,” in 2019 ION GNSS+ Conference, pp. 1753-1761, September 2019