(YIP) INFORMATION-GEOMETRIC PATH PLANNING
Abstract
The purpose of this proposal is to introduce the concept of information-geometric path planning (IGPP) and develop its theoretical and algorithmic foundations. The proposed development is motivated by human decision-makers capability of simplicity-oriented" path planning (e.g., simple and long paths are sometimes preferred to short and complex paths), which often contributes to the reliability and the robustness of human decisions. Due to the lack of a standardized mathematical metric for simplicity," it is challenging for robots to reproduce human experts capability of discovering simple" solutions. Consequently, the task of nding the simplest path is often left for human operators, which signicantly restricts the capability of robots in the current practice. To overcome this challenge, we develop the IGPP framework, which has a built-in capability of penalizing both path length and path complexity. This can be achieved by introducing a novel information-geometric distance function in the conguration space. The IGPP concept is closely related to the notion of rational inattention introduced by C. Sims (2003). The proposed research project involves the development of ecient algorithms to nd the shortest path with respect to the new distance function.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 12, 2021
- Source ID
- FA95502010101
Entities
People
- Takashi Tanaka
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- University of Texas at Austin