Link to DRRT paper: https://www.clear.rice.edu/comp450/papers/drrt.pdf

DRRT (Dynamic RRT) is a tree-based motion planning algorithm for dynamic environments. In addition to traditional RRT, it implements a replanning procedure which repairs the RRT when changes have been made to the configuration space. In static environments, DRRT behaves identically to RRT, and thus inherits properties such as probabilistic completeness.

Subsequent calls to the solve method of the planner will prune the tree based on changes to the environment, and keep adding nodes until a solution is found again. In some situations, this is faster than planning from scratch.

This planner was tested using a modification of the Point2DPlanning demo, where the environment and start positions were changed between calls to solve. It can be observed that parts of the tree are reused.

Initial call to solve:

Adding a barrier close to the start, and calling solve again (7x faster than initial solve):

Changing the start position whilst keeping the barrier:

Adding a barrier closer to the goal:

The planner was also tested in Pybullet with a Turtlebot: https://github.com/romand001/pybullet_ompl