Academic Work

Master’s Thesis

Abstract

Robotic path planning is a critical problem in autonomous robotics. Two common approaches to robotic path planning are sampling-based motion planners and continuous optimization methods. Sampling-based motion planners explore the search space effectively, but either return low quality paths or take a long time to initially find a path. Continuous optimization methods quickly find high-quality paths, but often return paths in collision with obstacles. This thesis combines sampling- based and continuous optimization techniques in order to improve the performance of these planning approaches. This thesis shows that the advantages and disadvantages of these approaches are complementary and proposes combining them into a pipeline. The proposed pipeline results in better path quality than either approach alone, providing a robust, efficient, and high-quality general path planning solution. The use of collision checking techniques introduced by continuous optimization methods in sampling-based planners is also analyzed and approximation error rates and timing results are provided.

Full Thesis

PDF version

Slides