Masterarbeit--Solving Disconvergence Problem Existing in Coooperative Visual Servoing Control

Visual servoing control is the process of using computer vision data to control the motion of a robot. In this topic, the computer vision data are images taken by the camera mounted on the robotino. In multi-robot-system, relative pose between two robots is essential both for cooperation and the formation control. During the task-implementing process, sometimes it is necessary to move two(or more) robots to certain configurations all together at the same period of time.

While two robots are moving towards a specific configuration, problems might happen during the process: they can not reach the desired relative pose because of the disconvergence. To make it more clear, disconvergence here means the two robots can not successfully reach the final relative pose while instead they get trapped in a local movement.

So in this topic, how to design the path of the two robots and control the robots so as to avoid the disconvergence is the important issue which must be taken into consideration.

Masterarbeit--Particle-Filter-based robot localization using active vision

In robotics and computer vision when using camera as the sensor, most programs will choose features extracted from images to represent the data as well as to build the map of the environment, thus features are of utmost importance to the whole process.

This topic includes two main parts: First define good feature from experiment; Second, based on the definition of good-feature, use good-feature-driven methods to control the movement of robot to verify the definition of good-feature.

For the first part, the following needs to be implemented:

1.Set the ground truth environment using marker.

2. Let robot move along different path. During this stage, more and different features will be placed into the testing environment to explore the influence of the number and type of the features on the accuracy of the localization.

3. Recover the path from using Particle-Filter-based method, compare the recovered path with the ground truth.

Based on the comparison results, the optimal path are designed which can help capture more better features so as to improve the accuracy of the robot's localization.

Masterarbeit--EKF-based robot localization using active vision

This topic is similar to Particle-Filter-based robot localization using active vision, except that we are going to use EKF-based method to process the image data. By comparing the performance of different Filter under the same testing environment, we get the idea of their limitations on the numbers of the features, their accuracy of dealing image data in constant time, their strength and weakness in representing the uncertainty…..

So the student doing this topic can cooperate with the student who chooses the above topic.

zurück zur Übersicht der am Fachgebiet angebotenen Arbeiten