Development and Testing of Autonomous On-Orbit Assembly and Servicing Systems Using the SPHERES Testbed

My time under the support of the NSTRF program can be jointly summarized between my time as a master’s and doctoral student at MIT in the department of Aeronautics and Astronautics. I was awarded the fellowship half-way through my master’s degree when my focus of research was autonomous on-orbit assembly and servicing. During this time I was the state estimation lead of the MIT Space Systems Lab SPHERES team and ISS testbed. These efforts resulted in five live experiments that first academic year as an NSTRF fellow, emphasizing vision-based autonomous docking between free-floating robotic systems and the check-out of new expansion port hardware to enable future autonomous assembly activity between SPHERES. From this work, I completed my master’s thesis titled, “State Estimation of Cooperative Satellites for On-Orbit Assembly and Servicing of Spacecraft,” through which I developed and validated two distinct state estimation frameworks for autonomously docking satellites. The first --- tested aboard the ISS --- was based on an Unscented Kalman Filter. The second was a probabilistic inference approach, modeled via a dynamically updated factor graph. This allowed for the convergence of continuously smoothed estimates of both docking satellite’s states. Since this model incorporates all sensor information as it becomes available, the satellites are able to estimate their own (and their relative states) using IMU data, or global metrology, even when their main sensor, the docking port camera, loses line-of-sight of its target. This algorithm was tested, and show to be successful, by playing-back data obtained from the ISS on a virtual machine. Following the completion of my master’s degree, I began my first NSTRF VTE which began a relationship with the JPL Guidance & Control section that has continued through the present. At JPL, I served as the attitude control systems (ACS) member of the design team for the Habitable Exoplanet Imaging Mission (HabEx) in preparation of the U.S. 2020 Astrophysics Decadal Survey. There, I designed multiple preliminary ACS architectures and simulation environments to meet high-precision (1 milliarcsecond class) observatory pointing requirements for direct imaging of earth-like exoplanets. I continued this work the following summer during my second VTE with the group. My first year as a doctoral student was marked by a blend of coursework and continued development of ISS robotics experiments via the SPHERES testbed --- primarily through the Smooth-Nav campaign. The sequence of three test session sought to further develop robust and precise multi-agent relative-state estimation for space applications by further exploring factor graph based inference methods. It was argued that this methodology more naturally frames the multi-satellite relative-state estimation problem, providing solutions that are inherently agnostic to the timing, synchronization, and potentially inoperable sensor issues that limit the filtering framework, by simply appending observations (i.e., measurements) to the graph (and therefore the joint probability distribution of the state history and measurements) as the information becomes available. As far as our team is aware, we became the first to successfully run smoothing-based estimation via graph based methods, on-board multiple satellites, in real-time. During the second half of this year, I also mentored the control & estimation team for the senior capstone design class, which sought to design a spinning aperture imaging CubeSat concept. During my second year as a doctoral student (3rd NSTRF), I led the final test session for the ISS SPHERES tether-slosh campaign for MIT. This included all management of the MIT SPHERES team, product (software) development and delivery on schedule, testing in simulation and hardware, customer interaction (Airbus Defense & Space), and post-session data analysis. The objective of the session was primarily to collect high-speed IMU data of the SPHERES satellites tugging a fluid-filled (as well as a solid) capsule. Additionally, I joined the in-space assembly of telescopes (iSAT) NASA/JPL study, beginning by mentoring and providing feedback to younger graduate students on astrodynamics, trajectory design, and guidance & control and then fully managing the MIT team (from the student perspective) in the spring semester. Our lab was tasked with performing an independent cost analysis for an in-space assembled telescope concept. We developed and delivered a pareto optimization analysis tool that breaks hundreds of telescope assembly architectures specified by primary mirror diameter, segmentation, raft geometry, and assembly location. The tool then uses linear programming methods to optimize packing of all assembly components (by mass, volume, and cost) into available launch vehicles (including SLS). Efficient trajectories using the unstable/stable manifold of the Circular Restricted Three Body problem are then determined to transport all materials to the assembly location (currently either Earth-Moon L1 or Sun-Earth L2). The architectures are then compared against a developed cost metric based on architecture complexity to determine the pareto front. The results heavily showed favoring of SLS launch vehicles. These activities were independent of the work of my dissertation. During my time at JPL, I became interested in direct exoplanet imaging (through HabEx), trajectory design, and multi-body astrodynamics. The resulting body of research, which culminated in my doctoral thesis (defended May 2020) is summarized as follows: The prolific discovery of habitable zone residing exoplanets via indirect detection methods have spurred many in the astrophysics and space technology community to call for the prioritization of funding for a direct exoplanet imaging space telescope. Though the state-of-the-art in optical technology suggests near-term feasibility, successful and efficient high-contrast imaging remains a problem. A promising solution is formation flying an external occulter in front of the observatory to suppress host starlight and allow for imaging of the obscured exoplanet. However, recent analyses have demonstrated that for the required separation distance between the spacecraft, angular slew maneuvers to retarget the formation line-of-sight between stars in a Design Reference Mission (DRM) demand a significant amount of fuel, restricting the potential science yield of a five year mission. It can be found that many of these analyses use traditional, impulsive control solutions to slew the occulter between points in three-dimensional positional space, or attempt exhaustive search methods to find less expensive alternatives. These approaches are uninformed by the rich and complex dynamical six-dimensional phase space in which the spacecraft truly lie. For this work it is assumed that both the observatory and external occulter are operating near Sun-Earth Lagrange point 2 (SEL2).Researchers across celestial mechanics, nonlinear dynamics, chaos theory, and astrodynamics over the last century have made considerable contributions to shedding light on the families and classes of natural trajectories existing in the phase space about Lagrange points. However, it has only been in the last few decades (and still continuing through the present) that it has been revealed how to use these previously elusive pathways in mission design. All of this points to a rich and underutilized design space for crafting naturally existing, or minimally active-control assisted, low-fuel solutions to solve complex motion problems. The difficulty lies in teasing out trajectories of interest in the often times opaque dynamical structure. However, history has shown that by understanding the basic classes of motion existing in the phase space through the lens of Dynamical Systems Theory (DST) --- which is concerned with qualitatively uncovering the structure of solutions in a system’s phase space through the study of its equilibrium points, their stability, sensitivity to parameters, and the vector flow connecting these points --- it can be done. My thesis investigated the use of natural solutions to frame and solve the formation retargeting maneuvers of an observatory/external occulter exoplanet imaging mission. By illuminating the classes of natural motion that can be exploited, fuel costs can be minimized, but more importantly, the set of all available paths contextualized within the dynamical landscape. This provides a baseline from which solutions can be interpreted and mission design trade-offs analyzed. To this end, a Trajectory Design Methodology (TDM) was developed that guides the spacecraft along the natural periodic and quasi-periodic motion of the CR3BP phase space's center manifold. The TDM determines the fuel-minimizing path, under the constraints of the analysis, that passes the formation line-of-sight through the maximum number of stars within an extended time window. Since the framework is dynamically informed, the incremental costs of deviating from this maximal path, to achieve a specific science objective, can be readily considered. A sample mission analysis demonstrating these contributions was provided.