Publication

Autonomy Artificial Intelligence Robotics (AAIR)
2022

PATH-FREE ESTIMATION OF NAVIGATION DISTANCE USING OBSTACLE SHAPE STATISTICS AND DENSITY

by Stephen J. Harnett; Sean Brennan; Karl Reichard; Jesse Pentzer; David Gorsich

Abstract

In the field of ground robotics, the problems of global path planning and local obstacle avoidance are often treated separately but both are assessed in terms of a cost related to navigating through a given environment. Traversal cost is typically defined in terms of the required fuel [1], required travel time [2], and imparted mechanical wear [3] to guide route selection. Prior work [4] has shown that obstacle field complexity and navigation cost can be abstracted into quantitative dimensionless parameters. But determining the cost parameters and their relationship to field complexity requires running repeated path planning simulations [4]. This work presents a method for estimating navigation cost solely from geometric obstacle field complexity measures, namely the statistical properties of an obstacle’s shape and the density of obstacles within an environment, eliminating the requirement to run a path planner in a simulation environment.