Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Vision Systems for Aerial Robots within Tree Canopy
The main goal is to design a drone which can exploit vision to identify regions of the canopy that are relatively empty from vegetation, therefore move through them to cross the canopy from bottom to top.
Forests are crucial, high-impact regions for the welfare of human society due to their importance for biodiversity, climate regulation, and ecological balance. However, their canopies are restraining from exploration because of their physical complexity: cluttered structure, dense vegetation, dynamic scene and changing light condition pose significant challenges for robots to safely access and navigate. In this project we want to develop a small drone to safely fly inside tree canopies. The main goal is to design a drone which can exploit vision to identify regions of the canopy that are relatively empty from vegetation, therefore move through them to cross the canopy from bottom to top. First, we will study the design of the drone considering two main requirements: (i) protective structures to shield the propellers from the vegetation, (ii) the integration of vision sensors (e.g., RGB-D camera) and an autopilot with imaging and signal processing capabilities. Second, we will evaluate the performance of state-of-the-art vision algorithms in the harsh canopy environment. Third, we will integrate the most effective vision algorithms on the aerial platform and test it while flying through the canopy.
Forests are crucial, high-impact regions for the welfare of human society due to their importance for biodiversity, climate regulation, and ecological balance. However, their canopies are restraining from exploration because of their physical complexity: cluttered structure, dense vegetation, dynamic scene and changing light condition pose significant challenges for robots to safely access and navigate. In this project we want to develop a small drone to safely fly inside tree canopies. The main goal is to design a drone which can exploit vision to identify regions of the canopy that are relatively empty from vegetation, therefore move through them to cross the canopy from bottom to top. First, we will study the design of the drone considering two main requirements: (i) protective structures to shield the propellers from the vegetation, (ii) the integration of vision sensors (e.g., RGB-D camera) and an autopilot with imaging and signal processing capabilities. Second, we will evaluate the performance of state-of-the-art vision algorithms in the harsh canopy environment. Third, we will integrate the most effective vision algorithms on the aerial platform and test it while flying through the canopy.
**Work Packages**:
- Familiarization with state-of-the-art literature on vision systems on UAV (HW/SW)
- Formalization of the platform requirements
- Assembling of the aerial system
- Development of vision techniques for teleoperated/autonomous flight
- Experimental testing on the field
- Detailed reporting of the project
**Requirements**:
- Highly motivated and interested in vision and hands-on hardware interfacing
- Methodological and goal-oriented working behavior
- Strong background in two of the following areas: UAV, computer vision, localization, machine learning
- Excellent programming skills (Python/C++ recommended)
- Experience with working under the ROS is advantageous
**Work Packages**:
- Familiarization with state-of-the-art literature on vision systems on UAV (HW/SW) - Formalization of the platform requirements - Assembling of the aerial system - Development of vision techniques for teleoperated/autonomous flight - Experimental testing on the field - Detailed reporting of the project
**Requirements**:
- Highly motivated and interested in vision and hands-on hardware interfacing - Methodological and goal-oriented working behavior - Strong background in two of the following areas: UAV, computer vision, localization, machine learning - Excellent programming skills (Python/C++ recommended) - Experience with working under the ROS is advantageous
Emanuele Aucone: emanuele.aucone@usys.ethz.ch
, Prof. Dr. Stefano Mintchev: stefano.mintchev@usys.ethz.ch
Emanuele Aucone: emanuele.aucone@usys.ethz.ch , Prof. Dr. Stefano Mintchev: stefano.mintchev@usys.ethz.ch