Guided Motion Planning
Jeff M Phillips
Summary
This project plans a path which obeys motion and cost constraints
and avoids obstacles. The increasing importance of autonomous satelite
service and repair which requires optimal control for orbital
rendezvous and docking motivates this research.
A Guided Expansive Spaces Trees technique is developed to search the
continuous state space of free-flying vehicles for an near-optimal and
collision-free path from a start to a goal. This technique employs a
tree-expansion technique to explore the configuration space by biasing
the search towards low cost paths and away from regions of the
configuration space already searched. By balancing both factors, low-cost
paths which obey motion constraints and avoid obstacles are produced.
Because beginings of high-cost paths are biased against being further
searched,
they are in effect preemptively pruned from the search tree before they
violate a constraint without explicitly calculating it.
The path is further refined by following the gradient of a path cost
function which is the sum of a term to avoid obstacles and a term to apply
minimal control. This technique converges to a local minimum in a few
iterations and can be used to avoid dynamic obstacles in real time while
maintaining optimality.
This technique is applied to a space shuttle docking simulation which
takes into account plume impingement and moving obstacles, as well as to a
simulation of a fan-controlled blimp in a factory environment.
Visuals
Guided search tree for space shuttle docking on space shuttle and most
optimal path with plume clouds superimposed.
install
cortona vrml viewer
Top view of guided search tree for fan-controlled blimp in factory
hallway.
Publications
Guided Expansive Spaces Trees: A Search Strategy for Motion- and
Cost-Constrained State Spaces.
Jeff M. Phillips, Nazareth Bedrossian, and Lydia E. Kavraki.
IEEE International Conference on Robotics and Automation.
April 2004.
Spacecraft Rendezvous and Docking with Real-Time, Randomized
Optimization.
Jeff M. Phillips, Lydia E. Kavraki, and Nazareth Bedrossian.
AIAA Guidance, Navigation, and Control. August 2003.
Probabilistic Optimization Applied to Spacecraft Rendezvous and Docking.
Jeff M. Phillips, Lydia E. Kavraki, and Nazareth Bedrossian.
AAS/AIAA Space Flight Mechanics Meeting. February 2003.
jeff phillips