Project Info
Learning to Plan
Neil Dantam | ndantam@mines.edu
Robot planners offer rich capabilities for autonomous decision-making. However, planners have many variations and parameters to choose from. Selecting the best planning approach is problem-dependent and may require a deep understanding of the planner’s internals. In non-ideal configurations, planning may be slow or not even work. This project will explore techniques to automatically learn and tune the planning process to obtain the best performance. Possible approaches include machine learning, optimization, and parallelization.
More Information
The Open Motion Planning Library (https://ompl.kavrakilab.org/).
The Task-Motion Kit (http://www.neil.dantam.name/papers/dantam2018tmkit.pdf).
Grand Engineering Challenge: Enhance virtual reality
Student Preparation
Qualifications
Student must be familiar with basic data structures and the C or C++ programming language. Necessary Courses: CSCI-262. Desired Courses: CSCI-358, MATH 213.
Time Commitment
20 hours/month
Skills/Techniques Gained
Student will develop an understanding of robot motion planning techniques and libraries. Student will learn various methods and
algorithms for learning and optimization in complex spaces.
Mentoring Plan
Weekly meetings and code reviews.