How to Track a Flying Object. Detection and Tracking in 3D Space. (SP, MT)

4ba93299 c6b9 4d4b 90d7 479cafe43d6d.jpg 500 500

Starting Date: earliest start: (2018-04-26) latest end: (2019-07-31)

Organization: Autonomous Systems Lab

Involved Host(s): Bähnemann Rik , Grinvald Margarita

Abstract: The main task of this project is to find a suitable sensor setup and implement a detection and tracking algorithm to catch flying objects with a UAV. [image]

Description: The Mohamed Bin Zayed International Robotic Challenge (MBZIRC) is a biennial competition aiming to demonstrate the state-of-the-art in applied robotics and inspire its future. The participants need to develop complex autonomous multi-agent flying robotic systems. 2019’s first challenge requires a team of up to 3 UAVs to autonomously locate, track, and interact with a set of objects moving in space. One target is attached to a UAV, following a 3D trajectory. The other targets are balloons, tethered to bases, and randomly placed inside the arena. The targets need to be collected and delivered to a pre-specified landing location. In this project the student will develop a detection and tracking algorithm to predict the position and velocity of a partially known object in 3D space. The main challenge is to find an algorithm that is robust to illumination changes and outliers, generalizes to a predefined set of objects, works both on long and short ranges, and outputs a high frequency accurate prediction running on a regular onboard CPU. The student will find a suitable sensor setup, summarize state of the art tracking algorithms, implement a base line solution, and test and improve the setup in simulation and on real hardware datasets. Related work: 1. 2. 3. Bähnemann, Rik, et al. "The ETH-MAV Team in the MBZ International Robotics Challenge." arXiv preprint arXiv:1710.08275 (2017). [PDF] 4. Loianno, Giuseppe, et al. "Localization, Grasping, and Transportation of Magnetic Objects by a team of MAVs in Challenging Desert like Environments." IEEE Robotics and Automation Letters 3.3 (2018): 1576-1583. [PDF] 6. Li, Rui, et al. "Monocular long-term target following on UAVs." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops. 2016. [PDF] 7.

Work Packages: - Sensor choice - Literature review vision-based and depth-based tracking - Development of a simulation plugin in RotorS Gazebo - Implementation of a detection and 3D tracking algorithm - Real world tests - Sensor and algorithm evaluation

Requirements: - Highly motivated and independent student - Computer vision background (beneficial) - C++ - ROS knowledge (optional)

Contact Details: If you are interested in this project, please send your transcripts and CV to - Margarita Grinvald - Rik Bähnemann

robotics UAV MAV competition balloons computer vision sensors detection tracking ROS MBZIRC challenge 1

Labels: Semester Project Master Thesis CLS Student Project (MPG ETH CLS) ETH Organization's Labels (ETHZ)
Topics: Information, Computing and Communication Sciences Engineering and Technology