Robot Manipulation in Densely Cluttered Environments N00014-23-S-B001
Abstract
Abstract (Approved for Public Release)Real-world environments, whether they be natural or man-made, are typically densely clutteredwith complex and irregular objects filling them. Modern manipulation systems have difficulty operating outside of carefully controlled environments with few obstacles primarily because of the complexity of the spaces; computing motions through these spaces typically involves an optimization problem that must be solved in a high-dimensional space with complex geometry. In this work, we propose a general framework that will enable robots to effectively operate and manipulate items in these densely cluttered environments. The overriding technical solution in this proposed research, called Act-Track-Resolve Trees, focuses mainly on addressing challengesin robot motion optimization. Our primary approach to robot motion optimization here is receding-horizon control, further supplemented by concepts from task and motion planning. This multifaceted approach will enable our solutions to effectively manage the continuous elements of robot manipulation via nonlinear constrained optimization as well as the discrete elements of robot manipulation via logic and symbolic-based reasoning. This research has the potential to enable robots to perform a wider range of tasks in settings where there are many objects and obstacles present. We believe that this research has the potential to significantly expand the operational capabilities of military robots as well as to have significant impact on application domains including disaster relief, agriculture, and any system that must operate in human-built environments.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Mar 08, 2024
- Source ID
- N000142412124
Entities
People
- Daniel Rakita
Organizations
- Office of Naval Research
- United States Navy
- Yale University