Rough Terrain Autonomous Mobility—Part 1: A Theoretical Analysisof Requirements

  • Authors:

  • Alonzo Kelly;Anthony Stentz

  • Affiliations:

  • Robotics Institute, Carnegie Mellon University, Pittsburgh, PA 15213. E-mail: alonzo@ri.cmu.edu, axs@ri.cmu.edu;Robotics Institute, Carnegie Mellon University, Pittsburgh, PA 15213. E-mail: alonzo@ri.cmu.edu, axs@ri.cmu.edu

  • Venue:
  • Autonomous Robots
  • Year:
  • 1998

Quantified Score

Hi-index 0.00

Visualization

Abstract

A basic requirement ofautonomous vehicles is that of guaranteeing the safety of thevehicle by avoiding hazardous situations. This paper analyses thisrequirement in general terms of real-time response, throughput,and the resolution and accuracy of sensors and computations. Several nondimensional expressions emerge which characterize requirements in canonical form.The automatic generation of dense geometric models for autonomously navigating vehicles is a computationally expensive process. Usingfirst principles, it is possible to quantify the relationshipbetween the raw throughput required of the perception system andthe maximum safely achievable speed of the vehicle. We deriveseveral useful expressions for the complexity of terrain mappingperception under various assumptions. All of them can be reduced topolynomials in the response distance.The significant time consumed by geometric perception degradesreal-time response characteristics. Using our results, severalstrategies of active geometric perception arise which arepractical for autonomous vehicles and increasingly important at higher speeds.