Selected Projects

Bending sheets is an important step in the production process of metal parts. In order to increase economic efficiency, an increasing number of companies seek to produce around the clock, thus requiring automated bending. TRUMPF offers the Bendmaster as an high quality robot manipulator, which combines high accuracy and short machining times. Both the manipulator and the press brake can be programmed offline with the TRUMPF offline programming system TruTops. My work includes optimizing the robot movement as well as the integration of new hardware components into the offline simulation.

Recent technological advancements increase the commercial availability of autonomous mobile systems, which can operate in a household environment. At present, they are still lacking the ability to recognize and grasp a priori unknown objects, which restricts their field of application.

The reason for this deficiency is the inability of these systems to adequately determine the state of the environment on basis of sensor data. Recent approaches for image based object recognition use a priori knowledge in order to identify objects. This knowledge is usually based on specific image features like texture, although geometric data is required to determine a suitable grasp. Even if the exact geometry of the objects is known, the available methods for determining anthropomorphic grasps require time-consuming calculations, thus restricting real time application.

In order to grasp a priori unknown objects, this thesis presents a new object representation, which describes the object shape using a topology of geometrical solids. Contrary to the exact quantitative nature of recent feature representations, this approach provides a qualitative description of the object shape without depending on specific color data. Thus, geometrically similar objects with varying dimensions can be encoded with a single object representation. In addition, grasping strategies are linked to the solids, thus removing the necessity to calculate the end effector configuration during the grasp process.

The proposed object representation does not contain any specific image features, which would allow for a direct comparison between a priori knowledge and examined object. Instead, the geometrical bodies have to be identified in the image data. Their scale and pose can vary and some bodies can be partially or totally occluded by others. The recognition of an object is based on a novel concept, which identifies the topology of the geometrical bodies using the object silhouette as image feature. Point distribution models encode all possible geometric variations for the silhouettes of a solid. During object recognition, the object silhouette is separated into contour parts, which are matched to the known point distribution models using active shape models. A recursive algorithm separates the silhouette until all contour segments were successfully matched. In order to cover all solids of the topology, the system performs this analysis on image data acquired from three different viewpoints. The collected shape information is finally used to deduce the object type, its pose and a suitable grasp strategy.

The results of the thesis can be found in my dissertation, which I have published at Dr.Hut. It is available at the publishers website (link below) under ISBN 978-3-89963-932-2.

The aging societies continuously increase the demands on respectful health care for the elderly and physically challenged people. The assistive wheelchair was designed to provide these people with technological assistance and thus making their life more independent from nursing staff. A touch screen can be used to access several functions like navigation supervision or autonomous driving skills. The system perceives its surroundings with several color cameras and a time-of-flight range measurement device. The 7-degree-of-freedom manipulator can autonomously grasp small objects and place them on the wheelchair tray.

Scientifically, the advantages of this system over similar prototypes are manifold. The renunciation of bulky laser, sonar or infrared scanners for range measurement requires an efficient image processing subsystem. The time-of-flight range camera is a cutting-edge development, which yet has found its way only into few research projects of mobile robotics. Unlike error prone stereo cameras it provides accurate 3D data of the surrounding under varying lighting conditions. The visually guided manipulator arm is able to grasp objects with complex shape.

Although robots playing soccer is often perceived as a fun diversion for after-work engagements, creating an autonomous platform which can master this sport is in fact a great technical challenge. The RWTH Aachen RoboCup players (AllemaniACs) are equipped with a pan-tilt color camera and a 360 degrees laser scanner. A wireless network connection provides the communication link to the other team players. The notebook, which can be seen on the picture at the right is just used a terminal to access the two compact personal computers resting inside the belly of the 60 kg robot, which is robust enough to support the weight of an adult person during navigation.

The research group at the RWTH consists of the Knowledge-based Systems Group (providing artificial intelligence and general software architecture) and the Chair of Technical Computer Science (providing image processing algorithms). Our team successfully participated in several German and international tournaments like the annual world championship, but our game performance has not yet brought us among the top three teams. Recently, the robot soccer research work at my chair has been shut down to shift resources to the assistive wheelchair.

Graphical visualization of multi-dimensional data sets has evolved to be a significant technique for providing the user with an intuitive interface to display results from scientific simulations. While various methods of graphical rendering have been successfully developed and implemented, the other human senses remained mostly unused. In my master thesis, I developed and implemented techniques to render abstract information with a haptic interface. These information were provided by dynamic fluid simulations and were both visually and haptically rendered in a virtual environment.

The work was carried out at the Virtual Reality Center Aachen (VRCA). The library to interface with the graphic environment can be used on Windows PCs, Unix workstations as well as Sun workstations. The computer handling the haptic device, a PHANToM device with 3 translational degrees of freedom for force output, has to be a Windows PC. The two machines handling graphic and haptic rendering are communicating with each other via UDP, thus synchronizing both rendering modes. Real-time support is a critical element of the complete system, since it has to be ensured that the users visual and haptical impression correspond with each other. This way the user can touch the streamlines, the cursor is forcefully drawn to vortices or repelled by areas of high pressure. In total I developed and implemented 13 different methods to feel vector or scalar data in fluid simulations.

As student assistant I worked for nearly 4 years in the area of virtual reality, both for the Virtual Reality Center Aachen and the Chair of Technical Computer Science. After a steep learning curve I decided to summarize my knowledge in two articles, one dealing with virtual reality hardware and the other one dealing with 3D computer graphics. You can find the links to these articles below.

My last and most interesting task was to develop a gimbal device from scratch. A gimbal is like a computer mouse with the exception that it has 6 degrees of freedom (3 translational and 3 rotational). That means, you can hold the device in your hand, move and rotate it anywhere you like while the cursor attached to it in the simulation moves and rotates likewise. I used the knob of a standard joystick as device, embedded a sensor of a tracking device (a so called Flock of Birds) into the knob and connected the buttons of the joystick to a controller board. This board converted the button signals to send them via an RS232 interface to the computer. In addition I wrote a software driver for the VR software to make the integration of the gimbal device as easy as possible for the user.

During my time at the Sense8 group of Engineering Animation Inc. I did not only enjoy the beautiful bay area, but also enjoyed working on a shuttle simulation which NASA engineers developed together with my employer. The NASA engineers had already developed a physics-based modeling which ensured a realistic behavior of the shuttle when activating its thrusters at nearly zero gravity. Moving the shuttle manually into its docking clamps without crashing was a tedious task since it required continuous course corrections at a very low speed.

I was given the task to develop an autopilot which was able to activate the thrusters autonomously and thus moving the shuttle appropriately without user intervention. It required a lot of lines of sophisticated code before the autopilot finally managed to dock without collision.

A space glove is another input device for virtual reality simulations. The movements of the users hand are detected by sensors and transmitted to the computer. By interpreting the sensor data, a virtual representation of the users hand can be rendered, imitating the users movements.

In order to easily incorporate the device into your own VR simulation, drivers are needed which can be interfaced with the VR software. My task was to develop a device driver for two different versions of the 5DT space glove for the World Up toolkit. World Up can be used to create complex VR simulations via drag&drop with a graphical user interface, additionally providing easy access to a wide range of VR hardware. At the end of this project, one were able to add the glove to the simulation with a few mouse clicks and the driver was immediately released by EAI as a plug-in for World Up.