Robotics Engineering Colloquium Series: Sihui Li
4:00 pm to 5:00 pm
Infeasibility Proofs in Robot Planning
Abstract: Many real-world tasks are conducted in confined workspaces or cluttered environments. Examples range from organizing household items in storage spaces to engine compartment checking and house framing. To perform such tasks, robots must not only find a path but also determine if it's feasible to reach their destination. Sampling-based motion planners are effective in high-dimensional spaces but are only probabilistically complete. This means the planners can find a plan given long enough time but cannot provide a definite answer of infeasibility if no plan exists. Our research contributes to algorithms that construct infeasibility proofs in motion planning problems. Infeasibility proofs together with sampling-based motion planning provide asymptotic completeness guarantees, meaning the planner can find a plan or prove infeasibility given long enough time.
This presentation will focus on the motion planning infeasibility proof algorithm, its extensions, practical applications, and future prospects. The algorithm combines learning and sampling-based motion planning in a new paradigm by interpreting learning results using computational geometry tools. The result is not only useful for constructing infeasibility proofs but also helps solve feasible, but challenging, motion planning problems. This work has shown application in HRI settings. Future work will focus on extending the current work to higher dimensions, applying it to improve task and motion planning, and leveraging this new capability across diverse applications.
Bio: Sihui Li is a Ph.D. candidate in Computer Science from the Colorado School of Mines. Before that, she had a master’s degree from Worcester Polytechnic Institute, a master’s degree from Rensselaer Polytechnic Institute, and a bachelor’s degree from the University of Science and Technology of China. Her research focuses on infeasibility proofs in robot motion planning that combine sampling-based algorithms and learning methods in an innovative framework. Her work was published in reputable journals such as IJRR, RA-L, and conferences such as RSS, IROS, and ICRA. She joined RSS Pioneers in 2023.
Zoom Link: https://wpi.zoom.us/s/92946413217