Carnegie Mellon University
file.pdf (1.63 MB)

Visual Obstacle Avoidance for Autonomous Watercraft using Smartphones

Download (1.63 MB)
journal contribution
posted on 2013-05-01, 00:00 authored by Tarek El-Gaaly, Christopher Tomaszewski, Abhinav Valada, Prasanna Velagapudi, Balajee Kannan, Paul Scerri

This paper presents a visual obstacle avoidance system for low-cost autonomous watercraft in riverine environments, such as lakes and rivers. Each watercraft is equipped with a smartphone which offers a single source of perceptual sensing, via a monocular camera. To achieve autonomous navigation in riverine environments, watercraft must overcome the challenges of limited sensing, low computational resources and visually noisy dynamic environments. We present an optical-flow based system that is robust to visual noise, predominantly in the form of water reflections, and provides local reactive visual obstacle avoidance. Through extensive field testing, we show that this system achieves high performance visual obstacle avoidance.




Usage metrics