Information-Geometric Path Planning

Abstract

The purpose of this proposal is to introduce the new concept of information-geometric path planning (IGPP), where we seek the shortest path in the configuration space, not in terms of the conventional Euclidean distance functions, but in terms of more general distance functions that can quantify the path complexity in an information theoretic sense. The proposed development is motivated by realistic motion planning tasks in probabilistic environments (e.g., autonomous navigation of UAVs) in which simple and long paths are sometimes preferred to short and complex paths in view of safety and reliability. Inspired by Sim's notion of rational inattention, the IGPP framework incorporates a distance function that is proportional to the information gain required to follow the given path. We equip the IGPP with highly efficient algorithms for path search, path smoothing and path following, all leveraged by the recent advancements on information gain maximization algorithms in the networked control systems theory literature. This project contains three thrusts. Thrust 1 develops mathematical bases for IGPP, including its characterization using the Finsler manifold theory. Thrust 2 focuses on algorithmic developments, including the modifications of the standard path planning algorithms such as RRT and A-star. Thrust 3 assesses how IGPP can advance the frontiers of the current path-planning technologies. In particular, we consider (i) path planning under sensing resource constraints, (ii) decentralized path planning, and (iii) path planning mimicking human experts. The expected outcomes of this project include a new theoretical framework, algorithms, and insights that add a new dimension to the existing autonomous motion planning technologies. The added dimension will help overcome the existing challenges in unmanned/manned Air Force platforms and enhance their autonomous capabilities.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Apr 30, 2024
Accession Number
AD1230411

Entities

People

  • Takashi Tanaka

Organizations

  • University of Texas at Austin

Tags

Fields of Study

  • Computer science

Readers

  • Graph Algorithms and Convex Optimization.
  • Robotics and Automation.
  • Systems Analysis and Design

Technology Areas

  • Autonomy
  • Autonomy - Autonomous System Control
  • Space
  • Space - Spacecraft Maneuvers