El desarrollo del mercado inmobiliario chileno y las medidas

The expert review was performed with six experts for automotive virtual reality applications. An even number of experts was chosen to allow for paired reviews that enable productive dis- cussions. The experts were recruited from various departments of the company. They were ergonomists, simulation experts and automotive VR system developers.

Apparatus

The expert review was conducted in a three-sided CAVE. This CAVE is driven by a PC-cluster generating images for 10 full-HD projectors. A pair of two projectors provides a passive stereo image separated by the INFITEC technology. Two pairs of projectors are vertically arranged for

Chapter 5. Evaluation

Figure 5.1: The interaction framework architecture.

each of the side walls with horizontal edge blending, resulting in a total resolution of 1920x1920. The floor image is diagonally projected by one projector-pair from above.

Each projector pair is connected to a workstation responsible for rendering the respective im- age. Therefore an instance of the VR framework is running on each of the rendering clients holding a copy of the virtual environment that is accessible from a common network resource. The actual application, managing in- and output devices, the interaction with the virtual environ- ment and all other issues, is running on a master workstation. This master application distributes the inter-frame differences of the camera and the input devices to the clients via network. This cluster support is originally provided by the VR framework and is used as a basis for the devel- opments presented within this work.

The interaction framework that was conceptually described in the preceding chapters was re- alized based on the VR system Virtual Design 2. This particular system allows for extensions by loading Dynamic Shared Modules (DSO) that provide well defined interfaces (see Section 2.5). For the realization of all aspects of direct finger-based interaction, a couple of modules and soft- wares were developed that interact with each other (Figure 5.1). The grasp manager collects information from the system-inherent collision detection to detect and manage grasp intentions of the users. It gathers the user input during manipulation and provides this information via

Chapter 5. Evaluation

Figure 5.2: Prototype application with a user manipulating the interior mirror in the CAVE

UDP to an object reaction server running independently from the VR system. This server is provided with knowledge about the constraints of each manipulatable object via a config file and calculates the pseudophysical object reaction on user input.

This client server design was chosen for providing cluster support which is necessary for the desired display systems. Since the original VR system is not capable of managing object ma- nipulations within a cluster framework, the resulting object matrices have to be distributed to all rendering clients (and the master) to guarantee consistent object states on all partial screens. Because of the loose coupling and the missing synchronization, small differences and delays happen to occur between the screens. Since the interaction server always provides the latest object matrix and due to the relatively high update rate, the differences are small and nego- tiable. Moreover, these small differences have to be preferred over an overall system latency that would have been caused by a more sophisticated synchronization method. Each of the client-applications and the master-application run a DSO that is responsible for applying the latest manipulated matrices to the respective virtual objects.

Precise finger tracking is realized through the wireless finger tracking gloves from Advanced Realtime Tracking. Optical hand and head tracking is supported by the same camera system. As a test scenario, the interior of a Volkswagen Touareg was chosen consisting of approximately two million triangles (Figure 5.2). The whole application in the CAVE was running at around 12-15f ps. Within this test scenario the following objects — each representing a group of objects with special characteristics — could be manipulated:

• Steering wheel • Sun shields • Glove box lid • Interior mirror • Light switch

Chapter 5. Evaluation

• A freely movable soda bottle • Driver’s door

Design and Procedure

The expert review was performed in terms of semi-structured interviews with three pairs of experts. This design was chosen to allow for a free discussion among the interviewees. Each session started with a testing period of approximately 20 minutes for each of the experts. They were asked to interact with all available objects during this period. While they interacted with the virtual car interior, the experts already discussed the prototype. After each subject experimented with the interaction metaphors, they were interviewed together for another 20 minutes. As a guide throughout the interview, a number of questions were prepared and casually asked during the discussion. These questions addressed the main challenges for the usability of the interaction metaphor:

• Are you able to grasp the virtual objects easily? • Are you able to release the virtual objects easily?

• Do you identify differences concerning the virtual objects? • Do the objects behave as you expected?

• Are you able to judge the position of your hands and fingers with respect to the virtual objects?

• Are you able to evaluate the accessibility of the virtual objects?

• For which aspects of virtual car validation do you consider this kind of interaction to be appropriate?

• For which particular aspects could this metaphor be used in contrast to an indirect controller- based method?

• Which improvement suggestions do you encounter and which application areas would open up as a result?