Modern visual-inertial odometry (VIO) systems can provide accurate and robust pose and velocity estimate using the combination of cameras and IMUs. However, state-of-the-art VIO methods rely on strong assumptions about the environment and sensors to work properly. If any of the assumptions is violated in practice, the performance will degrade rapidly, and VIO systems may fail completely. This greatly damages the robustness of VIO systems and causes problems for many real-world applications.
Modern visual-inertial odometry (VIO) systems can provide accurate and robust pose and velocity estimate using the combination of cameras and IMUs. However, state-of-the-art VIO methods rely on strong assumptions about the environment and sensors to work properly. If any of the assumptions is violated in practice, the performance will degrade rapidly, and VIO systems may fail completely. This greatly damages the robustness of VIO systems and causes problems for many real-world applications.
The goal of this project is to use machine learning methods to improve the robustness and performance of VIO in situations where conventional methods would fail.
The goal of this project is to use machine learning methods to improve the robustness and performance of VIO in situations where conventional methods would fail.
Zichao Zhang (zzhang at ifi.uzh.ch) Required skills: Linux, ROS and C++. Experience with VO/VIO/machine learning is a plus.
Zichao Zhang (zzhang at ifi.uzh.ch) Required skills: Linux, ROS and C++. Experience with VO/VIO/machine learning is a plus.