Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Aerial Benchmark: Evaluation of Publicly Available Visual-Inertial Odometry and SLAM Frameworks based on Aerial Groundtruth Datasets
The goal of this project is to evaluate publicly available visual-inertial odometry (VIO) and simultaneous localization and mapping (SLAM) frameworks applied to fixed-wing unmanned aerial vehicles (UAVs) based on ground-truth datasets.
Keywords: Visual-Inertial Odometry (VIO), Simultaneous Localization And Mapping (SLAM), Fixed-Wing Unmanned Aerial Vehicles (UAVs), State Estimation, Computer Vision, Benchmark
In the last years, numerous visual-inertial odometry (VIO) and simultaneous localization and mapping (SLAM) frameworks have been developed and are now publicly available. However, it is often unclear how well they perform in real-world applications & on the same dataset in terms of pose error and computational cost.
The goal of this project is to evaluate publicly available VIO and SLAM frameworks based on simulated and real-world ground-truth datasets using camera and IMU sensor data recorded by a fixed-wing unmanned aerial vehicle (UAV).
Together with the supervisors, the student can choose from the following VIO & SLAM frameworks:
- SVO: https://www.ifi.uzh.ch/dam/jcr:e9b12a61-5dc8-48d2-a5f6-bd8ab49d1986/ICRA14_Forster.pdf
- LSD SLAM: https://vision.in.tum.de/_media/spezial/bib/engel14eccv.pdf
- ORB SLAM: http://webdiis.unizar.es/~raulmur/orbslam/
- ROVIO: https://www.youtube.com/watch?v=ZMAISVy-6ao
- MSC-EKF: http://www.ee.ucr.edu/~mourikis/tech_reports/TR_MSCKF.pdf
- PTAM-SSF: https://github.com/ethz-asl/ethzasl_sensor_fusion
- OKVIS: https://www.youtube.com/watch?v=TbKEPA2_-m4
In the last years, numerous visual-inertial odometry (VIO) and simultaneous localization and mapping (SLAM) frameworks have been developed and are now publicly available. However, it is often unclear how well they perform in real-world applications & on the same dataset in terms of pose error and computational cost. The goal of this project is to evaluate publicly available VIO and SLAM frameworks based on simulated and real-world ground-truth datasets using camera and IMU sensor data recorded by a fixed-wing unmanned aerial vehicle (UAV). Together with the supervisors, the student can choose from the following VIO & SLAM frameworks:
- Make yourself familiar with the existing evaluation framework
- Choose from a list of publicly available VIO & SLAM frameworks
- Implement adapters to run the individual frameworks
- Generate simulated ground-truth datasets using our existing Gazebo simulation framework for fixed-wing UAVs: https://github.com/ethz-asl/rotors_simulator
- Evaluate the frameworks based on simulated and real-world data and highlight why they perform worse/better than others
- Make yourself familiar with the existing evaluation framework - Choose from a list of publicly available VIO & SLAM frameworks - Implement adapters to run the individual frameworks - Generate simulated ground-truth datasets using our existing Gazebo simulation framework for fixed-wing UAVs: https://github.com/ethz-asl/rotors_simulator - Evaluate the frameworks based on simulated and real-world data and highlight why they perform worse/better than others
- Student enrolled at ETH Zurich
- Programming experience required (C++)
- Experience with Ubuntu and ROS beneficial. We are mainly using Ubuntu 14.04 and ROS indigo.
- Student enrolled at ETH Zurich - Programming experience required (C++) - Experience with Ubuntu and ROS beneficial. We are mainly using Ubuntu 14.04 and ROS indigo.