Abstract
Satellites flying in close proximity can provide many advantages over a single monolithic system when performing a mission. However, the balance between minimising fuel and maintaining safe trajectories creates the inevitable dilemma of trading fuel for safety or vice versa to preserve formation life. Multi-satellite trajectory planning for proximity operations is traditionally completed days or weeks in advance of a manoeuvre due to the complex dynamics of relative motion. Allowing satellite formations the flexibility to perform path-planning operations on-board each spacecraft can significantly reduce the ground operations burden and increase the responsiveness of a formation to reconfiguration events. To meet rapid manoeuvre requirements for future multi-satellite missions, collision-free path-planning and execution must be completed on-board a satellite. This thesis presents a novel approach for real-time multi-satellite collision avoidance path-planning and execution which can be implemented autonomously on-board individual spacecraft in a formation. A systematic study of the effects of perturbations during optimal reconfiguration and a heuristic model of reconfiguration in relative motion creates a basis for building multi-satellite collision-free trajectory planning and control tools. Utilising an analytic reconfiguration model, a new semi-analytic collision identification approach is developed which increases the dimensional understanding and allows for focused collision avoidance planning. Implementation of this approach in conjunction with a sequential pareto-optimal trajectory deviation strategy to produces an innovative collision avoidance path-planner. A new analytic model predictive control system is developed which implements collision-free manoeuvre plans in the presence of perturbations and other uncertainties. Additionally, approaches are presented for extending the heuristic motion model. A new relative motion model is developed including J2 perturbations and using cylindrical coordinates which allows for higher-fidelity modelling of long-duration, large-separation relative motion. Such models further decrease fuel usage during the execution of multi-satellite collision-free reconfiguration. Comparisons with traditional methods demonstrate a substantially reduced computational burden allowing these to be the first such path-planning tools to be validated on spacecraft hardware. Controller demonstration also shows a dramatic decrease in fuel usage when compared with traditional analytic controllers at nearly equal computation time. Satellite hardware testing validates that both the semi-analytic collision avoidance and analytic model predictive controller are real-time solutions to safe on-orbit formation reconfiguration.