Many future space missions, such as space-based radar, earth mapping, and interferometry, will require formation flying of multiple spacecraft to achieve their very advanced science objectives. While formation flying offers many performance and operational advantages, there are several challenges that must be addressed, including navigation, control, autonomy, distributed data management, efficient inter-vehicle communication, and robustness. One of the key issues with formation flying of large fleets is selecting the overall system architecture, because it drives the distribution of the various algorithms and the extent to which data must be transmitted.These challenges are particularly evident with the relative navigation. While carrier-phase differential GPS can be used as a highly accurate sensor for LEO formations, it is not sufficient as the sole sensor for missions beyond LEO. If local ranges and range rates are used to augment or replace the GPS measurements, precise estimation can continue into MEO and beyond. However these new measurements complicate the estimator decentralization by coupling the vehicles' state estimates.This thesis explores solutions to many of these challenges within the context of the Orion microsatellite formation flying mission. It also presents the Formation Flying Information Technology testbed, developed to evaluate the communication and computational requirements associated with various system architectures when using augmented GPS. Several architectures and their associated estimation algorithms are also analyzed and compared in terms of performance, computation, and communication requirements. This analysis clearly shows that the decentralized reduced-order filters provide near optimal estimation without excessive communication or computation requirements. Embedding these reduced-order estimators within the hierarchic architecture presented should also permit scaling of the relative navigation to very large fleets. AcknowledgmentsIn carrying out the research that went into this Master's thesis, there were several key individuals that played large roles in helping me make it to the end. This was a long and difficult road at times and I thank everyone whole-heartedly for their kindness and support over the past two years.Firstly, I would like to thank my advisor, Professor Jonathan P. How for directing and guiding me through this research. Professor How's strong will and drive for excellence kept me on track, studying new and exciting topics along the way. Professor How taught me the keys to effective research, and for that, I thank him.

Formation flying spacecraft has been identified as an enabling technology for many future NASA and DoD space missions. However, this is still, as yet, an unproven technology. Thus, to minimize the mission risk associated with these new formation flying technologies, testbeds are required that will enable comprehensive simulation and experimentation. This paper presents an innovative hardware-in-the-loop testbed for developing and testing estimation and control architectures for formation flying spacecraft. The testbed consists of multiple computers that each emulate a separate spacecraft in the fleet. These computers are restricted to communicate via serial cables to emulate the actual inter-spacecraft communications expected on-orbit. A unique feature of this testbed is that all estimation and control algorithms are implemented in Matlab, which greatly enhances its flexibility and reconfigurability and provides an excellent environment for rapidly comparing numerous control and estimation algorithms and architectures. A multi-tasking/multi-thread software environment is simulated by simultaneously running several instances of Matlab on each computer. The paper contains initial simulation results using one particular estimation, coordination, and control architecture for a fleet of 3 spacecraft, but current work is focused on extending that to larger fleets with different architectures. It is expected that this testbed will play a pivotal role in determining and validating the data flows and timing requirements for upcoming formation flying missions such as Orion and TechSat21.

This paper describes several tools and technologies that have been developed for future spacecraft formation flying missions. This includes algorithms to perform autonomous navigation and control, with new results presented for trajectory planning in highly elliptic orbits and fault detection for a distributed spacecraft system. The overall approach is embedded in the OA middleware developed by Princeton Satellite Systems, which provides a seamless integration of networking and process control with C/C++ software development. New results in this paper also demonstrate OA integrated with software that is running in hard real-time. A new multi-computer spacecraft formation flying testbed was created to simulate the performance of the full system. In particular, simulation results of an optimally initialized "recurring tetrahedron formation" on a highly elliptical orbit are shown to demonstrate both the functionality of the new testbed and the potential for creating fuel-efficient formation configurations relevant to future magnetospheric space science.

No abstract

The number of resident space objects (RSOs) in orbit has increased dramatically within the last 10 years. While RSOs pose a serious challenge to our continued use of space, these objects also provide an opportunity to improve on-orbit state estimation by detecting and identifying these objects using star trackers. While star trackers are currently used to determine the orientation of their host spacecraft, their utility could be expanded to both position and orientation estimation by harnessing RSO detections. In order for RSO-based optical navigation to be commercially viable, a reliable filter covariance estimate is required. This paper introduces an Unscented Kalman Filter (UKF) for estimating an observing spacecraft's position and attitude based on RSO observations. To ensure that this filter is reliable, a new assessment technique is introduced where we compare the moving standard deviation and mean of the filter's estimate error with the predicted error based on the filter's covariance estimate. We demonstrate the utility of this assessment method by comparing the covariance trusts for our UKF and a previously developed Extended Kalman Filter.