Algorithmic Foundations for Real-Time and Dependable Spacecraft Motion Planning


The goal of this effort is to devise real-time, efficient and dependable algorithms for spacecraft autonomous maneuvering, with a focus on dynamic and cluttered environments (which may arise, e.g., due to debris or outgassing activity). Specifically, this project is aimed at devising a technology for the online planning of trajectories in proximity operations, which together with reliable environmental sensing and autonomous high-level decision-making is a key enabler for autonomous spacecraft navigation. As a radical departure from traditional methods, we are leveraging recent algorithmic advances in the field of robotic motion planning to spacecraft control. Our research objectives are to:
  1. Address the theoretical underpinnings for the application of robotic motion planning algorithms to the problem of onboard spacecraft maneuvering, with special attention paid to implementability on space-qualified hardware.
  2. Integrate the planning module within the overall spacecraft autonomy module, with a focus on encoding safety modes and addressing environmental uncertainties.
  3. Validate our algorithms on a state-of-the-art test bed that emulates both deep-space and microgravity environments.
This effort is supported by a an Early Career Faculty grant from NASA's Space Technology Research Grants Program (Grant NNX12AQ43G), and involves collaboration with the NASA Goddard Space Flight Center and the NASA Jet Propulsion Laboratory.


The 3-DOF testbeds of the Stanford Free-Flying Robotics Facility, three robots that use compressed air to hover on air bearings to mimic microgravity on a large granite table.  The free-flyers are used for real-time planning demonstrations in a space-like environment.


Autonomous planning in proximity of outgassing material on the surface of Enceladus, an example of a mission that cannot be guided by human oversight and hence demands robust and safe autonomy.


Planning in dynamic and cluttered environments requires fast re-planning and anytime computation, e.g., through tree data structures. Our research focuses on applying an efficient, asymptotically optimal algorithm named Fast Marching Trees (FMT*) to challenging dynamically-constrained scenarios.