Technische Universität München Robotics and Embedded Systems
 

Fusion of IMU and Vision Based Information

 
Type SEP, IDP, BA, DA, MA
Supervisor Prof. Dr.-Ing. Darius Burschka
Advisor Elmar Mair
Research Area Sensor Fusion, Visual Navigation, Embedded Systems
Associated Project DLR 3dMo
Programming Language C, C++
Required Skills  
Useful Knowledge Kalman Filters
Language English or German

Description

Do you like modeling? Do you like to work with sensors of different domains?

Visual Navigation has become an engaging field in the last years, as cameras are not only noninvasive and need low-current, but they became more and more compact, accurate and cheaper. However, the high data rate of cameras seems to be the bottleneck for embedded systems, which are also often highly dynamic. Thus, the fusion of a camera system and an inertial measurement unit seems to cover those lack and allow robust tracking also at low frame rates. imu_small.jpg

merlin.jpg The fusion of IMU data and camera data turned out to be quite tricky and the right parametrization for Kalman filtering is difficult to find. Thus, the sensor data and its uncertainty has to be evaluated in various tests and different settings to find similarities and rules for a proper parametrization. A first application of the results would be on the DLR 3D modeler and flying robots.

For more information to this scientific highly interesting work, please contact Elmar Mair.

Literature