Guaranteed Collision Avoidance with Discrete Observations and Limited Actuation

Erick J. Rodríguez-Seda, Dusan M Stipanovic

Research output: Chapter in Book/Report/Conference proceedingChapter

Abstract

A critical operational objective of any autonomous vehicle is to avoid collisions with other agents (i.e. vehicles and obstacles) at all times. To attain this objective, the vehicle must be able to localize other agents and implement a collision avoidance strategy. The localization process of other agents can be continuous or discrete. Although ideal, continuous observations may not be possible or practical in many realistic scenarios due to common sensor limitations (e.g. slow sampling rates, short sensing ranges, and system failures) or because of energy-saving concerns (i.e. limiting the use of sensors to prolong the life of the vehicle's power source). Therefore, the design of avoidance laws with discrete information becomes necessary. In this chapter, we present noncooperative and cooperative avoidance strategies with time-varying discrete observations for a pair of two autonomous agents. We consider vehicles with limited sensing range and bounded control inputs, and develop strategies that guarantee the avoidance of a dynamic obstacle. The observation of the obstacle's position occurs at discrete time intervals that decrease in size as the obstacle comes closer to the vehicle. The avoidance strategy is also paired with a bounded trajectory tracking control and the existence of local minima, as well as a heuristic solution to avoid them, is discussed. We finally illustrate the performance of the control policies via two examples.

Original languageEnglish (US)
Title of host publicationAdvances in Intelligent Vehicles
PublisherElsevier Inc.
Pages89-110
Number of pages22
ISBN (Print)9780123971999
DOIs
StatePublished - Dec 1 2013

Keywords

  • Avoidance strategy
  • Collision avoidance
  • Cooperative
  • Discrete observations
  • Noncooperative
  • Trajectory tracking

ASJC Scopus subject areas

  • Computer Science(all)

Fingerprint Dive into the research topics of 'Guaranteed Collision Avoidance with Discrete Observations and Limited Actuation'. Together they form a unique fingerprint.

  • Cite this