抄録
Path planning remains a critical challenge in robot navigation and autonomous driving. While the
Rapidly-exploring Random Tree (RRT) algorithm efficiently explores high-dimensional spaces, its
asymptotically optimal variant RRT* suffers from slow convergence due to high computational overhead. This
study evaluates two heuristic optimization techniques to accelerate RRT* convergence. First, ellipse sampling
constrains random sampling to an elliptical region defined by the start and goal points as foci and the current
path length as the major axis, eliminating wasteful samples. Second, path pruning removes redundant nodes to
generate straighter, more efficient trajectories. Experimental simulations demonstrate that ellipse sampling
significantly improves convergence speed, while pruning enhances path quality by reducing unnecessary
detours. These heuristic optimizations substantially improve the practical performance of RRT* for real-time
applications.
