Stereo vision–based navigation for autonomous surface vessels

TLDR

This stereo vision–based system is the first fielded solution tailored for high‑speed autonomous maritime operation on small boats. The study presents a description and experimental analysis of the Hammerhead vision system and key elements of the R4SA control system for autonomous navigation in maritime environments. The Hammerhead vision system detects geometric hazards above the waterline, generating grid‑based hazard maps and discrete contact lists, while the R4SA architecture uses these inputs to execute sensor‑based navigation behaviors such as static obstacle avoidance and dynamic target following. Experimental results demonstrate the integrated vision‑based navigation system operating on high‑speed unmanned surface vessels. © 2010 Wiley Periodicals, Inc.

Abstract

Abstract This paper describes a stereo vision–based system for autonomous navigation in maritime environments. The system consists of two key components. The Hammerhead vision system detects geometric hazards (i.e., objects above the waterline) and generates both grid‐based hazard maps and discrete contact lists (objects with position and velocity). The R4SA (robust, real‐time, reconfigurable, robotic system architecture) control system uses these inputs to implement sensor‐based navigation behaviors, including static obstacle avoidance and dynamic target following. As far as the published literature is concerned, this stereo vision–based system is the first fielded system that is tailored for high‐speed, autonomous maritime operation on smaller boats. In this paper, we present a description and experimental analysis of the Hammerhead vision system, along with key elements of the R4SA control system. We describe the integration of these systems onto a number of high‐speed unmanned surface vessels and present experimental results for the combined vision‐based navigation system. © 2010 Wiley Periodicals, Inc.