Dynamics and Controls of Swarms of Femtosatellites

The goal of the proposed research is to create guidance and control algorithms for spacecraft swarms that are both fuel and computationally efficient. Fuel optimal methods, such as Lambert transfers, exist for a single spacecraft but the difficulty of a swarm reconfiguration occurs when trying to find collision-free maneuvers. This may require some of the spacecraft to have suboptimal trajectories in order to avoid collisions. Another challenge of swarm guidance is the limited computational capabilities of the spacecraft in the swarm. For this reason, computational efficiency is an essential quality of a swarm guidance algorithm. In order to meet the computational requirements of the problem, decentralized algorithms will need to be considered. Decentralized and hierarchical approaches provide several advantages including scalability and algorithmic simplicity, higher accuracy and efficiency with a larger number of femtosats, robustness to individual spacecraft failures, and minimal onboard computational requirements. This research deals with the guidance and control algorithms required to enable the flight of spacecraft swarms. The algorithms developed in this research are focused on achieving two main goals: swarm keeping and swarm reconfiguration. The objectives of swarm keeping are to maintain bounded relative distances between spacecraft, prevent collisions between spacecraft, and minimize the propellant used by each spacecraft. Swarm reconfiguration requires the transfer of the swarm to a specific shape. Like with swarm keeping, minimizing the propellant used and preventing collisions are the main objectives. Additionally, the algorithms required for swarm keeping and swarm reconfiguration should be decentralized with respect to communication and computation so that they can be implemented on femtosats, which have limited hardware capabilities. The algorithms developed in this research are concerned with swarms located in low Earth orbit. In these orbits, Earth oblateness and atmospheric drag have a significant effect on the relative motion of the swarm. The complicated dynamic environment of low Earth orbits further complicates the swarm-keeping and swarm-reconfiguration problems. To better develop and test these algorithms, a nonlinear, relative dynamic model with J2 and drag perturbations is developed. This model is used to validate the algorithms using computer simulations. The swarm-keeping problem is solved by placing the spacecraft on J2-invariant relative orbits, which prevent collisions and minimize the drift of the swarm over hundreds of orbits using a single burn. These orbits are achieved by energy matching the spacecraft to the reference orbit. Additionally, these conditions can be repeatedly applied to minimize the drift of the swarm when atmospheric drag has a large effect (orbits with an altitude under 500 km). The swarm reconfiguration is achieved using two steps: trajectory optimization and assignment. The trajectory optimization problem can be written as a nonlinear, optimal control problem. This optimal control problem is discretized, decoupled, and convexified so that the individual femtosats can efficiently solve the optimization. Sequential convex programming is used to generate the control sequences and trajectories required to safely and efficiently transfer a spacecraft from one position to another. The sequence of trajectories is shown to converge to a Karush-Kuhn-Tucker point of the nonconvex problem. In the case where many of the spacecraft are interchangeable, a variable-swarm, distributed auction algorithm is used to determine the assignment of spacecraft to target positions. This auction algorithm requires only local communication and all of the bidding parameters are stored locally. The assignment generated using this auction algorithm is shown to be near optimal and to converge in a finite number of bids. Additionally, the bidding process is used to modify the number of targets used in the assignment so that the reconfiguration can be achieved even when there is a disconnected communication network or a significant loss of spacecraft. Once the assignment is achieved, the trajectory optimization can be run using the terminal positions determined by the auction algorithm. To implement these algorithms in real time a model predictive control formulation is used. Model predictive control uses a finite horizon to apply the most up-to-date control sequence while simultaneously calculating a new assignment and trajectory based on updated state information. Using a finite horizon allows collisions to only be considered between spacecraft that are near each other at the current time. This relaxes the all-to-all communication assumption so that only neighboring spacecraft need to communicate. Experimental validation is done using the formation flying testbed. The swarm reconfiguration algorithms are tested using multiple quadrotors. Experiments have been performed using sequential convex programming for offline trajectory planning, model predictive control and sequential convex programming for real-time trajectory generation, and the variable-swarm, distributed auction algorithm for optimal assignment. These experiments show that the swarm-reconfiguration algorithms can be implemented in real time using actual hardware. In general, this research presents guidance and control algorithms that maintain and reconfigure swarms of spacecraft while preventing collisions between the spacecraft and minimizing the amount of propellant used.