A Hybrid Stochastic Motion Planning Algorithm for Safe and Efficient, Close Proximity, Autonomous Spacecraft Missions

Open Access
Digirolamo, Lawrence J
Graduate Program:
Aerospace Engineering
Master of Science
Document Type:
Master Thesis
Date of Defense:
Committee Members:
  • David Spencer, Thesis Advisor
  • Motion Planning
  • RRT*
  • Obstacle Avoidance
  • CMA-ES
  • Evolutionary Strategy
  • Spacecraft
  • Hill Clohessy Wiltshire Equations
As the International Space Station and Earth orbiting satellites age well past their originally planned operational lifespan, improved monitoring of these spacecraft’s integrity will be critical to the safety of any crew on board and continued functionality. A small autonomous free flying spacecraft could have the ability to monitor for structural instabilities without the need for astronaut intervention. This thesis develops a hybrid offline motion planner that determines a fuel efficient trajectory between user specified waypoints for an inspection spacecraft, while avoiding thruster impingement with the target spacecraft. The planner requires the inspection vehicle’s dynamics model, a thruster model and the obstacle field within which inspection vehicle operates. The algorithm is shown to find trajectories superior to its predecessor. The algorithm is a merging of the Optimal Rapidly Exploring Random Tree algorithm, the Covariance Matrix Adaptation Evolutionary Strategy, and the Hill-Clohessy-Wiltshire equations. The Optimal Rapidly Exploring Random Tree algorithm is a directed tree search algorithm with a theoretical basis in random graph theory that acts as the main search framework. The Covariance Matrix Adaptation Evolutionary algorithm is used as a local optimizer which directs search to low cost solutions. Finally, the Hill-Clohessy-Wiltshire equations are used to find dynamically feasible trajectories through the search space. Simulations are performed for flights around a simple model resembling the International Space Station, with several start and goal points, and several combinations of algorithm parameters. Results obtained from the stand-alone Optimal Rapidly Exploring Random Tree algorithm are compared to results obtained from the hybrid algorithm developed in this thesis. The hybrid algorithm shows improved performance over the stand-alone Optimal Rapidly Exploring Random Tree algorithm. The hybrid algorithm on average is able to find trajectories which require less propellant and less flight time, however the hybrid algorithm tends to require more computation time. In addition, planned trajectories that do not require impingement prevention require less propellant and shorter flight times. As such, cold gas thrusters may prove to be more appropriate for close proximity spacecraft missions, despite their lower efficiency as compared to their counterparts which utilize more volatile propellants. Finally, the propellant use and time of flight of a particular trajectory can be effectively tuned by the user through algorithm parameter modification.