Autonomous Vision-Controlled Micro Rotorcraft

Subject: Autonomous Vision-Controlled Micro Rotorcraft
Faculties: Faculty of Aerospace Engineering
  Faculty of Computer Science
Researchers: Associate Professor Pini Gurfil and Professor Ehud Rivlin

Abstract:

We are designing a quadrotor for autonomous navigation based on computer vision.

The quadrotor prototype is capable of vision-based navigation using an original vision sensor and inertial sensors.

The platform supports tele-operation where a communication link exists, and is able to automatically switch to an autonomous mode in case of communication breaks.

We are pursuing research in two directions: (i) autonomous vision-based control, and (ii) outdoor autonomous navigation.

As a first step, the remotely-operated quadrotor will be controlled by a laboratory ground station, which processes the video stream from dispersed laboratory cameras as well as airborne cameras and closes the flight control loops.

The proposed system will computationally-efficient simultaneous localization and mapping (SLAM) using online mosaicking, and will ultimately enable straightforward incorporation and validation of beyond the state-of-the-art vision-based navigation and control algorithms.

One application is airborne surveillance and path planning for other autonomous platforms, with which the quadrotor can be integrated.