Closed Jmeyer1292 closed 7 years ago
Does this change not affect the high level Sparse and Dense planners?
It does not. The internal point mimics the interface of a JointTrajectoryPoint
(ala Python duck-typing) and no modifications to the other components are required.
+1
This PR replaces the use of
JointTrajectoryPt
inside theplanning_graph
with aInternalJointPt
class that mimics the former's interface, but only includes components used by the planning graph.This means that every vertex no longer stores additional upper/lower/discretization vectors or coordinate frame transforms.
On the ROSCON demo, with discretization of 0.05 radians, maximum memory consumption dropped 36%. With 0.1 radians, it dropped approximately 48%. Program execution time dropped by about 10%.