2013 Poster Sessions : Sampling-Based Spacecraft Motion Planning

Student Name : Joseph Starek
Advisor : Marco Pavone
Research Areas: Information Systems
Sampling-based motion planning strategies are investigated for real-time spacecraft guidance and control in dynamic, rapidly-changing and cluttered environments, with an emphasis on spacecraft proximity operations. As computing becomes more efficient and sensing more sophisticated, real-time autonomous planning is expected to play an increasing role in the control of modern spacecraft systems, particularly in dynamic environments made hazardous by neighboring spacecraft, debris, and out-gassing activity. Autonomous spacecraft proximity operations is a challenging task, however, that entails the online solution of an NP-hard combinatorial optimization problem with differential and algebraic constraints, possibly uncertain information, and very limited computational capabilities. This appears to be beyond the scope of traditional control approaches, which mainly require static, uncluttered environments and often rely on a ground station in the loop. I am exploring a class of real-time sampling-based motion planning algorithms with "anytime capability" that can guarantee mission safety and robust execution during critical, high-risk mission phases as well as allow online re-planning in the event of control failures or environmental changes. Such techniques could be key enablers for future near-Earth and deep-space missions, including on-orbit satellite servicing, rendezvous and docking, and missions to near-Earth objects and other planetary moons.

The Rapidly-exploring Random Tree (RRT*) algorithm is demonstrated here in generating trees of feasible trajectories between an initial state and goal state for a 3-DOF point-mass spacecraft in several dynamical environments. The RRT* motion planning algorithm is employed in order to quickly explore the free spacecraft configuration space, enabling the circumnavigation of obstacles and determination of an efficient and provably-safe trajectory to the goal. Initial simulation results are presented for three hypothetical proximity operations scenarios: (1) impulsive thrust and (2) piecewise-constant thrust control in Low-Earth Orbit under the Clohessy-Wiltshire-Hill Equations, and (3) autonomous maneuvering about a stable rotating celestial small-body approximated as a very low gravity inverse-square law. Control of a manipulator arm with embedded sensory networks is also presented as a proof-of-concept for real-time, closed-loop control using a similar sampling-based RRT-based algorithm, with specific application to time-varying temperature obstacle avoidance.

Joseph Starek is currently a PhD candidate in the Aeronautics & Astronautics department at Stanford University. He obtained his Bachelor's of Science in Engineering (BSE) and Master's of Science in Engineering (MSE) in aerospace engineering from the University of Michigan at Ann Arbor in 2010, where he researched model predictive control techniques for low-thrust trajectory optimization. Joseph has spent two summers working at NASA Ames Research Center in rotorcraft aeromechanics testing and small spacecraft propulsion system design, later joining the Ames small spacecraft Mission Design Center for nine months before attending Stanford in the fall of 2011. He is now pursuing a doctoral degree under the guidance of Professor Marco Pavone, director of the Autonomous Systems Laboratory (ASL), with a research focus centered on spacecraft dynamics and control, trajectory optimization, and spacecraft motion planning algorithms.