671: Advanced Topics in Computational Robotics: Foundations and Tools, Spring 2020
Instructor: Jingjin Yu
The course provided a systematic coverage of advanced computational methods commonly used in Robotics, spanning the three main components of sensing, planning, and control. In terms of specific topics, the course discussed foundational elements including configuration spaces and their topological structures; basic state estimation including geometric methods (e.g., trilateration, ICP), Kalman filters, SLAM, and learning based sensor fusion; decision making and control must-knows including task and motion planning and belief space planning.
<!–
The plan is to keep the enrollment relatively small (e.g., ~ 20 students) so that focused discussions can be held.
The course will cover a set of advanced topics related to computational issues in robotics. While the list of topics is still being finalized, some likely choices are:
Foundations: configuration spaces, topological manifolds, connectedness, homotopy, random graphs, percolation theory. State estimation: Kalman filter, SLAM, ICP, learning-based sensor fusion. Decision making and control: task and motion planning, symbolic planning (e.g., PDDL), LQR/LQG, iterative LQR/LQG, POMDP and belief space planning, kinodynamic motion planning, dynamical systems and optimal control, small time local controllability (STLC).
The topics to be covered will also be partly based on student comments/suggestions.
In terms of the teaching format, after an initial 1-2 weeks of coverage on foundations, the course will focus on a computational method for 1-3 weeks (e.g., ICP, iLQR, PDDL, etc.) and dig into it. The goal is that, after the coverage, a student would be able to have both a theoretical understanding and basic working knowledge of the topic. We will allocate sufficient time to cover a given topic/method. Detailed notes will be provided. Significant efforts are expected from the students.–>