Augmented Reality (AR) applications are highly dependent from accurate and precise tracking data. Since current tracking technologies do not always provide such information everywhere in real-time application developer must combine certain trackers together to minimize the disadvantage of one tracker by another. These sensor networks can then be used to deliver positional relation information of objects to the application which then can be evaluated.
Currently most AR application bring along their own customized solution of this problem. However these solutions are hardly reusable in other systems. This inhibits the development of large-scale sensor networks because there are no standard interfaces between these technologies. By introducing the Ubitrack framework it is possible to form ubiquitous tracking environments which may consist of several sensor networks.
One of the main problem in the Ubitrack framework is the question how to set up the system for new participating users when they are using Ubitrack themselves. Or in other words: how do you identify new objects in the spatial relationship graph in Ubitrack?
Consider the following situation: Bob is a computer science student at the TU München. He is currently working in the AR lab at the software engineering chair of Prof. Brügge. His friend Alice is at the bus station and wants to know where he is at the moment. So she asks this to her mobile Ubitrack system.
The problem now is how we can integrate the Ubitrack system of Alice into the whole system. Her client does not know anything about the existing tracking environment at the bus station. This is the same problem which all peer-to-peer systems have: where to find potential partners without a central coordination facility and how to form a network with these partners.
Another problem is the identification of objects in the combined graph. Alice is tracked by her mobile setup and possibly by the tracker setup at the bus station. When both are combined the system must know that the two tracked objects are the same person.
The first step is the evaluation of literatur concerning P2P architectures, especially efficient ways to build ad-hoc P2P networks. The results of the literature search is then used to modify DWARF to suit the needs mentioned in the project's description. This can be done either by finding attributes which can be used to model the bootstrapping protocol or by introducing new components to the current middleware.
The purpose of the new model in DWARF is to detect new edges which combine several subgraphs together. This includes the fusion of all available context sources to defer the spatial relationship between two graph nodes. The problem thereby is the identification of newly created objects (see above). Possible solutions are to find similarities in the movement of current objects by motion estimation.
-- FranzStrasser - 19 Jul 2004