Bachelor, Diploma or Masters Thesis
Computing Ego-Motion and Structure from Point Features
When viewing video footage from a non-stationary source, it is possible to infer the motion of the camera, as well as the structure of the viewable scene, simply by tracking point features in the image. Such calculations are commonly referred to as solving the "ego-motion and structure" problem, which has attracted much attention over the years. We have recently developed a very simple and efficient algorithm that is designed to be practical for real-time applications with limited hardware resources that require a continually-maintained estimate of position and/or spatial structure. Good examples of such systems are small and inexpensive mobile robots.
The algorithm performs visual point tracking in a circular 360deg field of view, and updates its feature position estimates with every new observation. In a second pass it relies on the feature position estimates to update its ego motion estimate. So far we have only implemented the algorithm in a simulated environment; however, we believe that the algorithm is sufficiently simple to ultimately run on a standard microprocessor (32bit, 60MHz) on board of small mobile robots.
In this thesis the student will get a mobile robot with an omni-directional circular camera that initially connects to the local computer network through WLAN. In a first stage, the student will implement the algorithm in C on a desktop computer and evaluate its real-world performance. During a second stage the student will convert the algorithm to run on a microcontroller on-board of the mobile robot (still in C). Such an algorithm running on tiny on-board hardware with low communication latencies will ultimately allow very fast robot motion for small robots equipped with cheap sensors, such as autonomously navigating micro air vehicles.
- Mathematical investigation and formalization of algorithm.
- Setup of infrastructure for communication with omni-directional robot.
- Implementation of algorithm to run in real-time on PC (communication through WLAN).
- Reshaping of algorithm to run in real-time on microcontroller on-board of robot.
This thesis requires some mathematical understanding and a large portion of programming in C (on PC computer and/or microcontroller). The work will be closely linked into a new student-project started at NST that investigates in swarms of miniaturized autonomous flying vehicles. This thesis will help in self-stabilizing autonomously flying miniature quadro-copters.
Student:
Martin Medler
Advisor:
Prof. Dr. Jörg Conradt
If you are interested in this project please send an email outlining in about five to ten lines why you are interested, state your skills and knowledge, and explain why we should choose you among those who apply. Thank you!