News
The course is cancelled this term due to too less participants.
The registration is now open.
The number of participants is limited to 10.
Registration
Please register by email to
ramirezd@in.tum.de .
Please provide following information:
- Surname, Name
- Date of birth
- Matriculation number
- Branch of study
- Number of semesters in SS09
Overview
Two main problems of the mobile robot navigation are global localization and environment perception: without a map the robot cannot localize itself and not knowing its own pose prevents it to build a proper map. Hence, these two problems have to be solve simultaneously, which leads to the so called SLAM approaches.
Cameras are more and more popular in robotics because they are passive (non-invasive), compact, cheap, and their power consumption is low. Nowadays the algorithms and the processing capabilities of modern computers are powerful enough to deal with the cameras' drawbacks, namely the huge amount of data and the difficulties to extract the desired data out of them. That is why the vision based SLAM (VSLAM) permanentely grows.
In the seminar we are going to deal with the state-of-the-art of following topics:
- Extraction and tracking of natural landmarks in images
- Stability and robustnes of VSLAM
- Loop closure techniques
- 3D sensor-data fusion
- Map building of outdoor enviroments
- fusion of sensor data
Topics
