One of our key research interests is haptic rendering. The word "haptic" is derived from the greek work "haptesthai" which means "to touch". Haptic rendering means to create an artificial tactile sensation by using a computer and a special haptic device. By using haptic rendering we are able to not only to see virtual objects but also touch them.

A good overview over the research area of haptic rendering is given in [1].

The human tactile sensoric system employs several different kinds of sensors. They all provide signals that are transmitted to our brain where they are combined to a full perception of a haptic sensation.

One can group the sensors into categories, according to the different kinds of haptic sensations we can experience. The group one immediatly thinks of is the group located in our skin. This group is recognizing pressure or temprature while we are touching the surface of an object. But there are other sensations which we experience when touching an object. Kinesthetic feedback means that we are able to recognize a force that is applied to our body by the help of sensoric cells that are located at the end of the tendons or between the muscle strands. They are also able to measure the force we apply to objects.

It is important to see that all the sensoric information is combined in our brain to a complete haptic experience. Furthermore our brain is able to recognize surface features, even if not all sensoric channels provide information. Vendors of haptic devices make use of this feature as they don't sell devices that actually create a sensation for all haptic channels. Most of them only provide a kinesthetic sensation, also known as force feedback. These force feedback devices are nowadays the most popular haptic devices, although there are also other kinds of devices available (e.g. the braille readers for visually impaired people).

The virtual reality research over the past 50 years was very successful. We are today able to let the user experience a virtual world in an almost photorealistic way. But after all the progress that was made in computer graphics, we nowadays clearly discover that virtual reality still does not come close to the actual reality. Especially the interaction with the virtual objects is still very artificial.

This artificial interaction makes interactive tasks in virtual worlds more complex than in the real world. Another related field - teleoperation/telerobotics - found that out relative early and uses haptic rendering intensively.

Several psychological studies discovered that haptic perception is critical for the human performance in interactive tasks. Not only the speed but also the precision increases significantly if the user is able to get a force feedback while he is fulfilling his task. E.g. Hannaford and Wood describe in [2] an experimental study in which the provision of haptic rendering lead to a reduction of the completion time by approximately 30%, to a reduction of the applied force by a factor of 7 and to a reduction of errors in performing the task by 63%.

Although the results of these studies are very promising and should lead to an increasing acceptance and widespread use of haptic devices, there is one major drawback of haptic rendering. The force that is outputted to the user has to get calculated in realtime, which results in most of the non-trivial applications in a quite complex calculation. In addition to that, the force-update-rate that has to get achieved in order to provide a realistic haptic impression, is very high - usually a rate between 500Hz to 1kHz is cited in literature (see [3]). As a result the force calculation is the most challenging part if you would like to add haptic rendering to a virtual reality application.

We are using haptic rendering to simplify the complex task of interactively fitting a high resolution molecular structure into a low resolution electron density map. The user is guided by a soft force towards a better fitting location, whereby the cross correlation coefficient is used as criterion - which is also used by most of the algorithmic docking programs.

Interactive programs are in this field still very popular as the full-automatic algorithms still have problems with noisy experimental data. Our approach combines ideas from both interactive and algorithmic programs, as it provides the user with the information how an algorithm would rate the current situation, although the user is still in charge of searching for the perfect fit. This makes our approach robust and intergrates the users expertise.