Connectables

ConnecTables: Dynamic Coupling of Displays for the Flexible Creation of Shared Workspaces In: Proceedings of the 14. Annual ACM Symposium on User Interface Software and Technolo (UIST'01), ACM Press (CHI Letters 3 (2)), 2001. pp. 11-20.

ConnecTables: Dynamic Coupling of Displays for the Flexible Creation of Shared Workspaces Peter Tandler, Thorsten Prante, Christian Müller-Tomfelde, Norbert Streitz, Ralf Steinmetz GMD – German National Research Center for Information Technology IPSI – Integrated Publication and Information Systems Institute AMBIENTE – Workspaces of the Future Dolivostr. 15, D-64293 Darmstadt, Germany +49-6151-869-{863, 924, 867, 919, 869} {tandler, prante, tomfelde, streitz, steinmetz}@darmstadt.gmd.de http://www.darmstadt.gmd.de/ambiente/ ABSTRACT

We present the ConnecTable, a new mobile, networked and context-aware information appliance that provides affordances for pen-based individual and cooperative work as well as for the seamless transition between the two. In order to dynamically enlarge an interaction area for the purpose of shared use, a flexible coupling of displays has been realized that overcomes the restrictions of display sizes and borders. Two ConnecTable displays dynamically form a homogeneous display area when moved close to each other. The appropriate triggering signal comes from built-in sensors allowing users to temporally combine their individual displays to a larger shared one by a simple physical movement in space. Connected ConnecTables allow their users to work in parallel on an ad-hoc created shared workspace as well as exchanging information by simply shuffling objects from one display to the other. We discuss the user interface and related issues as well as the software architecture. We also present the physical realization of the ConnecTables.

change of different activity modes results in better support of work processes. They will become more fluent because of less interrupts and less media breaks. Another starting point is that future work situations will be more and more determined by “beyond-the-desktop” scenarios [23, 25]. This requires to support a wide range of different devices and information appliances in an integrated fashion. In the context of this paper, we focus on the subset of co-located work situations. The goal of the presented research is to contribute to the facilitation of interaction with information as well as interaction between people by providing better affordances for the type of activities described.

KEYWORDS: interactive tables, context-awareness, roomware, dynamic coupling, shared workspaces, colocated groupwork. INTRODUCTION

Current and future office work is determined to a large degree by information processing work. A significant part of this type of work requires the cooperation of people in groups. Thus, people have to move between different modes of operation: individual work, small group activities as well as plenary situations. Facilitating the initiation and

Figure 1: (a) Three working modes of the ConnecTable (b) Two connected ConnecTables. Supporting Co-located Group Work

For interaction-intensive tasks, e.g., working together on highly interdependent complex problems, face-to-face meetings are still the premium collaboration mode [16]. A workshop-style face-to-face meeting might start in a plenary situation. As the meeting proceeds, the participants will probably break up into subgroups and individuals working in parallel. Later, there is a re-gathering of subgroups possibly leading to another plenary situation, etc. It has been shown, that having cycles of individual and cooperative work within a single meeting, generating ideas and discussing them with partners, can improve meeting

productivity and quality of output compared to having several meetings in a row to accomplish the same tasks. This flexibility in work situations needs to be facilitated by intuitively subdividing, exchanging, and aggregating new and old information. Consequently, a really interactive and flexible medium and corresponding devices are required in order to support the emergent group work process rather than hindering it. Such a medium, being flexible, fluent in use, and group-perceptible has been called interaction medium in [12]. It is to be used during a meeting to support collaborative problem solving, e.g., through facilitating visualization and progressive structuring of ideas. When engaging in a tight cooperation, e.g., elaborating together proposed material, people quite frequently use “displays” and workspaces that are jointly accessible. People then often gather around a whiteboard or a table to visualize important points of the discussion in parallel. This is usually done manually, listing items, or sketching diagrams accompanying the discussion, which is known to facilitate and catalyze discussion. In contrast to this, today, computers are mainly used either as presentation tools or for taking private notes as well as capturing the results of a meeting. When people come together to explore and elaborate on project ideas or having a design session, the focus is on human-human interaction rather than on human-computer interaction [25]. Therefore, the explicit human-computer interaction should be kept to a minimum and should be constrained to dealing with the subject matter of the meeting. Other necessary operations should be performed as implicit interactions, i.e. users should not have to primarily interact with the computer system but the system should be able to interpret actions in the environment as input for triggering system operations [19]. This allows that users can focus on their primary individual and collaborative tasks related to the subject matter.

coupling of previously single-user displays to form a larger homogeneous display area providing a multi-user shared workspace for synchronous work. This mechanism addresses what can be called the “secondary interaction problem”, i.e. the initiation and handling of cooperation modes of devices vs. the “primary interaction problem”, i.e. the handling of input and output of content. RELATED WORK

In the past, several systems have been developed to support meetings with digital media in order to avoid media breaks and to foster efficiency of the meeting process. Electronic meeting systems [15] provide every meeting participant with a traditional desktop- or laptop-computer allowing for quick information access and parallel input reducing the problem of air time fragmentation. In supporting workshopstyle, dynamic meetings, these systems have not proven useful because they require pre-structuring the meeting process. According to [10], it is rather common and not an exception that the agenda is often changed in the course of such meetings and there is the danger that such meetings are overly confined. One of the first computer-setups to support face-to-face meetings, Colab [4], provided the meeting participants with a collection of so called meeting tools, each for a different purpose. No strict process was prescribed. Much later, one of Colab’s successors, the Ocean Lab hosted the groupware system DOLPHIN [24] providing a modeless pen-based user interface to hypermedia structures allowing to support synchronous cooperation through parallel input to the same document. All of the so far mentioned labs and approaches respectively used a static setup of a large public display and several table-mounted personal computers. This does not allow for enough flexibility to support various emerging work situations.

One approach to the realization of our goal is considering and designing computers or information technology components in general as secondary artifacts and not as primary artifacts as the current desktop computer still is. By integrating them into the prime objects of people’s activities, we augment the usage of those primary artifacts. This enables us to make use of the affordances provided by the primary artifact but at the same time augment it by computer-supported functionality. This results in an integrated design of physical and virtual worlds into hybrid worlds [25].

Numerous single display groupware applications [2, 13, 22] have enriched the one-computer setup by providing multiple input channels to it using different devices. But still, the collaborating users are not freed to move around. They share one single display that cannot be split. For example, Tivoli [11] offers a pen-based user interface with elaborated features to organize various materials on an electronic whiteboard enabling at most three users to work in parallel using different pens. The users have to switch between modes to enter gestures or scribbles. Also, the meeting setup is restricted to the whiteboard only and consequently the user interface has not been adapted to other devices.

In this paper, we present a “connection mechanism” that supports gradual, cue-rich, fluent, and intuitive transitions between individual and cooperative work as they appear in the type of meetings described above. The connection mechanism is based on explicit physical movements in the real space resulting in actions in the virtual space, i.e. the

In the i-LAND project [23], the first generation of roomware components, i.e. room elements with integrated information technology, were built. The InteracTable and the DynaWall are in some way “ancestors” of the ConnecTable. The InteracTable has a single horizontal

touch-screen integrated, but there is only one mouse cursor. As a consequence, when people are using the InteracTable for discussion they have to work together in a turn-taking style that harms the fluidity and naturalness of communication and interaction. The DynaWall results from the tight integration of three interactive SMART Boards [21] into a wall: Up to three persons can either work individually in parallel or can share the entire display space. Thus, a single ConnecTable is similar to the InteracTable with respect to the way of interacting with a horizontal display. Now, two ConnecTables constituting a homogeneous tabletop and interaction area are similar to the DynaWall with respect to the provision of combined and shared workspaces but they are different because they can be put together dynamically on demand.

positions, whereby users are enabled to work while sitting on a chair or standing in front of a ConnecTable. This provides for an ergonomic change between the two positions allowing the user to adopt the position he or she currently likes best. In addition, the display can be tilted in different angles to provide the optimal view position (see fig. 1a and the corresponding video submitted along with this paper).

Regarding the easy exchange of information between devices, different mechanisms have been proposed including Passage [8]. Other examples were presented by Rekimoto. Pick’n’Drop [17] uses a pen to transfer information, while hyperdragging [18] allows to drag objects across the boundary of a laptop screen onto an interactive table or wall using the laptop’s mouse. However, these techniques are bound to the available displays and do not allow combining displays dynamically to a homogeneous interaction area, thus offering less flexibility.

When two ConnecTables are moved close to each other as shown in figure 2b, a shared workspace is temporally created combining the two previously personal ones. This workspace expands across the borders of the two displays, which now form a homogeneous physical display area. This way, the logical interaction area is dynamically enlarged for the purpose of shared use overcoming restricted display sizes. Once the two ConnecTables are connected, users are able to exchange information objects just by shuffling them over to the other display (fig. 2c). The common workspace exists until the ConnecTables are pulled apart from each other.

Context-aware systems exploit context to provide relevant information and/or services to the user, where relevancy depends on the user’s task [5]. The software used with the ConnecTables knows on which roomware component it is running and reflects state changes indicated by integrated sensors. CONNECTABLES

The ConnecTable is a new mobile, networked, and contextaware roomware component that is designed to support interaction-intensive, workshop-style face-to-face meetings. The ConnecTable is equipped with a pen-sensitive display constituting its tabletop and a translucent chassis as a container for information technology components. Two small wheels at its very base provide the ConnecTable’s mobility. The height of the ConnecTable’s display can be quickly adapted in order to accommodate different working

(a)

(b)

Regarding the explicit user interaction, the ConnecTable is a pen-based information appliance. With respect to implicit user interaction, there are sensors integrated into the tabletop allowing the interpretation of physical actions in the real world, which are then used to trigger appropriate actions implemented by software.

BEACH USER INTERFACE

BEACH (“Basic Environment for Active Collaboration with Hypermedia”) [26] is a synchronous groupware tailored to support face-to-face group work with roomware components. Due to a modeless gesture-based interaction style, interaction is fluent, continuous, and without interrupts providing a high speed of interaction. The concept of an explicit selection mode has been discarded from the user interface. This avoids disturbing interference of input, when two users work in parallel on the same document, as afforded by two connected ConnecTables. We experienced parallel input to occur quite often in the course of tight collaboration. The ConnecTable, in its upright position, has a horizontal information display, which has no predefined orientations as, e.g., top and bottom. Therefore, the rotation of objects allows adapting the alignment of documents to the positions

(c)

(d)

Figure 2: To switch between individual work (a) and tightly coupled work, two ConnecTables are placed next to each other (b). This results in a homogeneous display area enabling to exchange information objects (c). Moreover, users can have their own but shared views of the same information object for tight collaboration (d).

of the people standing around the table. In addition, several views on the same document can be created and simultaneously used on various ConnecTables. This allows face-to-face communication at the tables, while working on the same document. The rotation of objects is only provided and visualized at roomware components with a horizontal display, thereby realizing an example of device sensitive adaptation of functionality. USER EXPERIENCE

When building compound computational artifacts as, e.g., the ConnecTable, the design of the user experience has to be extended beyond the design of the user interface on the display. In most cases, these artifacts will be dedicated information appliances and not general-purpose devices and tools as desktop or laptop computers are. Thus, the design of the user experience can no longer assume the standard usage context of PCs or laptops. The ConnecTable provides an excellent example of the integrated design of both the physical and the virtual world [23]. The primary artifact is a small mobile table that has been enhanced with information technology. Thus, the interaction with the computational device (secondary artifact) has moved out of the focus into the background and the affordances mediated by the primary artifact are used and augmented. When two ConnecTable displays form a larger tabletop, (fig. 1b), this is reflected in the virtual world. Gestalt laws of perception then suggest the perception of the displays as a coherent area. Conforming to the expectations evoked by this, the software provides a logical interaction area encompassing both displays. The temporarily created common workspace contains all objects from the two previously separated workspaces keeping their positions and sizes constant. As an indicator for the users, the background color of the common workspace is changed. The user interface of the ConnecTable encompasses not only the user interface and the content displayed on the screen, but also the physical components. The latter are good candidates to mediate information guiding or providing cues for the user interacting with the devices in an implicit and peripheral way. When the device is in use, there are no cables for network or power-supply indicating the technical nature of the artifact. In addition, there are blue LEDs integrated, shimmering through the translucent material, whereby the user gets a peripheral cue of the enhancement and the proper functioning. The display of the device in turn is very visibly forming the tabletop of the ConnecTable and consequently attracting the user’s attention to the pen-based user interface.

Initiating Cooperation

Being located in one room, it is quite a natural behavior to move closer to each other when engaging into a discussion, especially when multiple parallel discussions are ongoing. Bringing one’s ConnecTable along, then moving close to each other and subsequently connecting them provides an appropriate medium for visualizing ideas and results while discussing. Natural “body language” can be used to control the situation when someone comes closer to engage into a tight collaboration and vice versa. This way it is easy for both parties to signalize whether tight collaboration is now appreciated or not while proceeding to work. For example, showing no reaction when somebody else is approaching (“communication request”) is a good sign for the approaching person that he or she should not disturb. When ready, the approached person might look up to show that the situation has changed (“communication accept”), having already noticed the communication request peripherally. It turns out that movement in real space and engagement in discussions also provides good awareness cues for the other participants of, e.g., a workshop when working in parallel. As the whole process of initiating and conducting cooperation in real space is rather a continuous flow than a sequence of sudden state changes, it affords to be observed peripherally. This led to the design of the connection mechanism. Interaction Possibilities

The initiated cooperation is reflected by the fact that the two connected ConnecTables create a multi-user component ad-hoc. The homogeneous display area supports a discussion of two users: They can visualize important points in parallel and exchange them from time to time. Another option is to work together more tightly coupled using simultaneous views enabling the users to concurrently work on the same information object (fig. 2d). Using a simple gesture creates a simultaneous view of an information object. This second view of the same object can then be moved to the other display and rotated, just as one would do with a sheet of paper. In addition, each user is able to work on “one sheet of paper” using two different displays each showing the sheet in the proper orientation for the respective user. Both options allow the users to proceed working on the exchanged information objects when their ConnecTables are not connected anymore. PHYSICAL REALIZATION OF THE CONNECTABLES

The current realization of the ConnecTable uses a translucent chassis as a container for all the IT components and the power supply. The pen-interactive display forming the tabletop allows a user to interact with information objects by pen on a display area of 13-inch diagonal with

1024x768 pixels. The decision to use a WACOM PL-400 graphic tablet [27] was motivated by the tight coupling of display pixels and pen-technology and the high quality of the pen-interaction. Another property of the pen-based display is that one can lean on the display with the hand, e.g., to support the writing hand without initiating irrelevant touch signals.

(a) coil (b) tag (c) sensor

approximately 3 cm and takes 1.5 sec. The sensor connected to the main coil communicates with the embedded PC using a serial RS232 interface. SOFTWARE INFRASTRUCTURE FOR THE CONNECTABLES

The functionality needed to support the ConnecTables is implemented as a new module of BEACH. This section first gives an overview over the software architecture of BEACH and the design of the model of the roomware components. Based on this, it is first described how this model was extended to allow combined roomware components. Using this extended model, it is explained what happens in the software when two ConnecTables are placed next to each other.

Figure 3: Sensor technology integrated in the ConnecTable. Coil and tag are placed left and right at the top of the display to detect other tables.

The ConnecTable’s computer unit is based on an embedded PC board in a 5.25" form factor with a 266 MHz Pentium mobile CPU. Two PCMCIA slots mounted on the embedded PC provide the system with the possibility of flexible peripheral extensions. The independent power supply for at least 3 hours is based on custom-built Nickel Metal Hydride (NiM) accumulators integrated into the upper part of the chassis. The accumulators are managed by a special electronics unit that charges with a pulsating current to reduce the memory effect of the accumulators and to allow to “charge while operate”. The mobility of the component is achieved by employing a state-of-the-art wireless network connection (WaveLAN, Lucent Technologies) with a maximum transfer rate of 11 Mbps. The use of rotating elements or inductive electric power consumers has been avoided and instead a passive cooling, flash-disk, etc. were employed. Hence, the ConnecTable is completely noise free, its computer equipment has not only visibly but also acoustically disappeared. Sensing Technology

The sensing technology integrated into the ConnecTable’s tabletop to detect others provides contact-less identification of passive tags based on radio frequency transponder technology. The right part of figure 3 shows the utilized components and the left part shows their location in a ConnecTable. A coil (see fig. 3a) is establishing an energy field by emitting electromagnetic waves on a certain frequency. When a passive transponder tag (see fig. 3b), which consists of a small coil and electronics enters the energy field emitted by the main coil, it uses the induced energy to send back its unique 32 bit identification to the main coil using another frequency. Signals from the main coil, which also acts as an antenna, are forwarded to the sensor (see fig. 3c) where they are processed. The identification mechanism starts at a distance of

Station

Display

1..*

1..*

Roomware Component

shared model objects

DisplayArea

Workspace

Figure 4: A roomware component consists of one to many stations. The stations’ displays are combined to a display area where a workspace is displayed. Architecture of BEACH

BEACH has a layered software architecture consisting of four layers: core layer, model layer, generic layer and modules layer. The core layer uses the COAST framework to provide a shared object space that can be distributed over multiple machines [3, 20]. To be able to access shared data a BEACH client must be running on every PC of each roomware component. The BEACH clients connect to a server that is responsible to control the replication and synchronization of the clients. To keep the displayed information always up-to-date – especially because the shared objects can be changed by different clients – BEACH couples the local1 view objects to the shared model objects using automatically tracked dependencies. (For model-view-controller see [9].) This is a technique similar to one-way-constraints used in many other systems (e.g. [14]). The model layer provides the basic structure for the higher layers by defining interfaces for documents, user interface, tools, the physical environment, and interaction styles as the abstractions common for all devices. On top of these 1

The term “local” is used for objects local to one computer, in contrast to “shared” objects accessible by multiple computers.

models, a set of generic components is defined that provides the basic functionality necessary in most teamwork and meeting situations. This includes for example standard data types like text, graphics, and informal hand-written input (scribbles), as well as private and public workspaces for generic collaboration support. Based on the models and the generic components, modules for specific support can be added, which define tailored functionality for distinct tasks. The architecture of BEACH is presented in more detail in [26]. Model of Roomware Components

In order to include the physical setting and configuration of the roomware components in the interaction with the software the physical environment is modeled with shared objects. Sensors watching for changes in the physical setting keep this model up-to-date. BEACH provides support for the handling of sensors as part of the core layer (see below).

If one roomware component consists of multiple stations their displays are combined to a single display area (as in the example of the DynaWall above). Every display’s view object should then show the part of the display area belonging to its local display. Therefore, every display has the attribute offset defining the position relative to the other displays. The attribute rotation specifies the orientation of the display relative to the display area, in case not every display has the same orientation relative to the other displays of a display area (an example is given in figure 9). A DisplayLayouter is responsible to assign the right coordinates to each display. shared model objects

rotation = 180 offset = (0, 0)

aStation1

aDisplay1

aDA1

model

aDisplayView1

model

aDAView1



aWorkspace1

aDisplayView1



aTranslation Wrapper1 model

aDisplay Layouter1

local display

s tran

n latio

com

d pute

by >

aDAView1

local view objects



aCT1

model

> by

aDisplay1

d ute mp co

aRotation Wrapper1

aDA1 shared model objects

local view objects

n tio

The generic layer defines the workspace as the part of documents storing the information a user is working with. A user can write or draw directly on a workspace or place images or text in it. Every display area displays the contents of a different workspace by default.

Composite Roomware Components

ta ro

Figure 4 shows the basic components: A station is a PC running a BEACH client. A roomware component consists of one to several stations. Each station can have one display. The displays of all stations belonging to the same roomware component are combined to one display area, which represents the complete visual interaction area of the roomware component. This is necessary, as e.g. a DynaWall [23] consists of several displays that are coupled via software.

that is currently attached to this display as a sub-view (aDAView1). The display area view will open a view for the workspace currently attached to its display area (aWSView1). Due to the dependency-mechanism of COAST, the views will automatically be re-computed as soon as the display is moved to a different display area or a different workspace is attached to the display area. This recomputation is triggered independently of the client actually changed the shared model.

model

aWSView1

Figure 5: Object diagram showing the shared model objects of a single ConnecTable (fig. 4) together with the dependencies to the local view objects.

When a BEACH client is started, local view objects are created to display the information stored in the shared model objects. Figure 5 explains which objects are used to display information on a ConnecTable (aCT1): The display view (aDisplayView1) is opened on the display of the local station (aStation1). It will open a view for the display area

Figure 6: Assigning the position within a display area to the local display view. In addition to figure 5, the two transformation wrappers are shown that adapt the subview’s local coordinate system. Transforming the Views. In the view hierarchy BEACH inserts two transformation wrappers. A wrapper is an implementation of the decorator pattern [7], used to modify the functionality of another object without having to change existing code. Transformation wrappers are capable of changing the local coordinate system of a view to e.g. a different position, orientation, or scale relative to the surrounding coordinate system. The wrapped view does not notice any difference – it is still using its local coordinate system. This is similar to the internal cameras used in Jazz [1].

A rotation wrapper is inserted on the display view, a translation wrapper on the display area view (figure 6). The rotation wrapper on the display view asks the

DisplayLayouter for the correct rotation of the local display. The translation wrapper on the display area view asks for the right translation to show the appropriate part of the display area.

the currently connected roomware components (see below). 0..*

Station



Device

Combining Roomware Components. In order to combine

the displays of two ConnecTables to a homogeneous surface when they are connected, the model described so far was extended by a new abstraction for the dynamic combination of several roomware components. A CompositeRoomwareComponent consists of multiple other roomware components, but has its own display area (see fig. 8 below). If other roomware components are connected to a composite roomware component their displays will belong to the common display area – which adjusts the local view of all connected roomware components automatically.

0..*

SensorClient

If a new value is sensed, the ConnectionMonitor checks if the ID belongs to another station (fig. 7). In this case, the model of the roomware components is updated representing



Connection Monitor

comPort : Integer value : ID

Tag id : ID

TagSensor

Figure 7: Sensors attached to a station. The ConnectionMonitor handles the connection if a sensor recognizes another ConnecTable.

Sensor Management and Concurrency

As stated before, sensors watching for changes in the physical setting keep the model of the physical environment up-to-date. For each sensor connected to a station (i.e. PC), a process responsible for the communication with this sensor is started when a BEACH client is initialized. This process will trigger the actions associated with this sensor every time the sensed value changes. Depending on the sensor, both push and polling to retrieve information from the sensor can be used. This part of BEACH was originally developed to implement the passage mechanism [8] and was reused without any modification.

Sensor



When two ConnecTables are connecting, concurrency has to be taken into account: Every ConnecTable has one sensor attached resulting in redundant information about the connection. To avoid that both tables try to change the same attributes of shared objects at the same time only one station will change the model. Otherwise, the conflicting accesses as soon as they are recorded by the BEACH server would result in rollbacks. Therefore, only the client running on the station with the smaller station ID executes the connection. Connecting two ConnecTables

The client triggered by the connection monitor to execute the connection has the task of reflecting the new configuration of the roomware components in the physical model. This includes the adjustment of the displays and

aComposite Roomware Component ConnecTable 1

ConnecTable 2

aStation1

aCT1

aCT2

aStation2

aDisplay1

aDA1

aDA2

aDisplay2

shared model objects

aDA3

model

aDisplayView2 so nd pe

so > n>

de

nd



aDisplayView1 local view objects

model

model

aDAView2

Figure 8: Two connected ConnecTables. Both displays are temporarily assigned to the common display area (aDA3) of the composite roomware component.

merging all workspaces to one common workspace. Connecting the Roomware Components. To connect two ConnecTables a new composite roomware component with a new display area is created. The two ConnecTables are added as components to it. This also moves their displays from the ConnecTables original display areas to the new common one (fig. 8). As a result, the view objects will be updated to show the correct part of the common workspace: Per definition the station changing the model will be the lower display using a zero rotation and an offset of (0, 1). The other ConnecTable is assigned to the upper display. As both ConnecTables have the sensors and tags built in on the same side of the display, this display is rotated by 180 degrees relative to the display area (fig. 9). Therefore its rotation is 180 and its offset is set to (0, 0). Connecting the Workspaces. Every display area shows originally an own workspace. When two ConnecTables are connected, the contents of their workspaces must be moved to the new common workspace of the composite roomware component’s display area (fig. 9). All objects shown in the upper display’s workspace (workspace1) must be rotated, as the new workspace’s orientation is upside-down compared to the original workspace’s orientation. The objects in the lower display’s workspace (workspace2) must be moved down by the size of the upper display.

Finally, the background color of the workspace is darkened to give feedback to the user that the ConnecTables are connected.

workspace3 workspace1 display1

display1

display2

display2

workspace2

(a) disconnected

(b) connected

Figure 9: Combining the contents of the two individual workspaces to a larger common one. To keep the contents at the same physical position they have to be rearranged with respect to the coordinate system of the common workspace. Disconnecting. When the sensors recognize that the

ConnecTables are separated again, the reverse operations are performed. The objects in the common workspace are moved back to the private workspaces of the ConnecTables depending on their position inside the workspace. Objects

overlapping the border between the displays are moved to the display’s workspace where most of them is visible. Then, the ConnecTables are removed from the composite roomware component and the displays are attached to the previous display areas. Consequently, the views are updated to show these display areas and workspaces. CONCLUSIONS, EXPERIENCES, AND FUTURE WORK

In this paper, we have presented the ConnecTable as an example of how to integrate software tightly with its physical environment – using actions in physical space as triggers of software functionality. Single-user pen-based displays are coupled causing an ad-hoc multi-user setup for shared work to be created and vice versa. We claim to thereby provide a fluent and coherent interaction experience to the users, when they gradually engage in and leave tight face-to-face cooperation, where the appropriate triggering signals come from built-in sensors. We have presented the user interface of ConnecTables arguing that user-interface design for such composite computational artifacts has to be understood in a broader sense. We also discussed aspects of the user experience as well as the physical realization of and the software architecture for ConnecTables. The ConnecTable originated in the research & development consortium “Future Office Dynamics” (FOD) where the non-IT related parts of the roomware components of the second generation have been built by the German office furniture company Wilkhahn in cooperation with GMD. Four fully functioning ConnecTables have been used along with other roomware components in an Expo 2000 showroom, operated by Wilkhahn employees for half a year to demonstrate the concepts to the public, where we received very positive feedback. The implementation of the connection as a new module shows that the architecture of the BEACH groupware offers the flexibility needed to support new forms of interaction. Most components like transformation wrappers and sensors could directly be reused, only some needed additional functionality. The models for composite roomware components and the “tag” for the tables had to be added. The first software tests on the ConnecTables showed that the values measured by the sensors were not as reliable when integrated in the tables as if they had been used standalone, possibly due to some interference. Therefore, the sensor module had to be extended to cope with small dropouts of the sensed values. This change could be handled internally in the sensor module, not affecting its interface. This proves that the proposed design defines flexible interfaces being able to hide small changes of the implementation, thus increasing its maintainability.

The first version of the ConnecTables presented in this paper allows two displays to form a composite one. Further work will go into the coupling of more ConnecTables to form a larger homogeneous area. Adding sensors and tags also at the left and right side of the ConnecTable’s tabletop would enable to form a public display consisting of more than two ConnecTables. Further, we will also equip mobile computing units with large interactive vertical displays with the introduced sensor technology to dynamically couple segments to form a large interactive wall. ACKNOWLEDGMENTS

We would like to thank all our students and colleagues for helpful discussions and for the support with the time consuming implementation work, especially Sebastian (spax) Pape who implemented the connection module. We also thank our cooperation partners Wilkhahn and Wiege in the research & development consortium “Future Office Dynamics” (FOD) for the constructive cooperation and for building the furniture part of the ConnecTables. REFERENCES

1.

2.

Bederson, B. B., Meyer, J., Good, L. Jazz: an extensible zoomable user interface graphics toolkit in Java. In Proceedings of User Interface and Software Technology (UIST’00), ACM Press, pp. 171–180, 2000. Bier, E., Freeman, S. MMM: A user interface architecture for shared editors on a single screen. In ACM SIGGRAPH Symposium on User Interface Software and Technology, Proceedings UIST’91, pp. 79–86, 1991.

3.

COAST Homepage. (2001) http://www.opencoast.org

4.

Stefik, M. et al. Beyond the chalkboard: computer support for collaboration and problem solving in meetings. CACM 30(1), pp. 32–47, Jan. 1987.

5.

Dey, A. K. Understanding and using context. Personal and Ubiquitous Computing 5(1), pp. 4–7, 2001.

6.

Dey, A. K., Ljundstrand, P., and Schmidt, A. Distributed and disappearing user interfaces in ubiquitous computing (workshop abstract). Extended Abstracts of CHI 2001, Seattle, WA, pp. 487–488, March 31 – April 5, 2001.

7.

Gamma, E., Helm, R., Johnson, R. Vlissides, J. Design Patterns. Addison-Wesley, 1995.

8.

Konomi, S., Müller-Tomfelde, C., Streitz., N. A. Passage: Physical Transportation of Digital Information in Cooperative Buildings. In Cooperative Buildings – Integrating Information, Organizations, and Architecture. Proceedings of the Second International Workshop (CoBuild'99). LNCS 1670,

Heidelberg, Germany, Springer, ISBN 3-540-66596-X, pp. 45–54, 1999. 9.

Krasner, G. E., Pope, S. T. A Cookbook for Using the Model-View-Controller User Interface Paradigm in Smalltalk. Journal of Object Oriented Programming. 1(3), pp. 26–49, 1988.

10. Moran, T. et al. Evolutionary engagement in an ongoing collaborative work process: a case study. In Proceedings of the ACM 1996 Conference on Computer Supported Cooperative Work (CSCW'96), ACM Press, pp. 150–159, 1996. 11. Moran, T., van Melle, W. Tailorable Domain Objects as Meeting Tools for an Electronic Whiteboard. In Proceedings of the ACM 1998 Conference on Computer Supported Cooperative Work (CSCW'98), ACM Press, pp. 295—304, 1998. 12. Moran, T.P., Chiu, P., van Melle, W. Pen-Based Interaction Techniques For Organizing Material on an Electronic Whiteboard. In Proceedings of the ACM Symposium on User Interface Software and Technology (UIST’97), pp. 45–54, 1997. 13. Myers, B. A. et al. Collaboration using multiple PDAs connected to a PC. In Proceedings of the ACM 1998 Conference on Computer Supported Cooperative Work (CSCW'98), ACM Press, pp. 285–294, 1998. 14. Myers, B. A., et al. The Amulet Environment: New Models for Effective User Interface Software Development. IEEE Transactions on Software Engineering, 23(6), pp. 347–365, 1997. 15. Nunamaker. Electronic Meeting Systems. CACM, 34(7), pp. 40–61, 1991. 16. Olson, G. M., Olson, J. S. Distance Matters. HumanComputer Interaction Journal. 15(2&3), pp. 139–178, 2000. 17. Rekimoto, J. Pick and drop: a direct manipulation technique for multiple computer environments. In Proceedings of UIST’97, pp. 31–39, 1997. 18. Rekimoto, J. Saitoh, M. Augmented Surfaces: A spatially continuous work space for hybrid computing environments. In Proceedings of the ACM Conference on Human Factors in Computing Systems (CHI'99), Pittsburgh, Pennsylvania, USA, pp. 378–385, 1999. 19. Schmidt, A. Implicit Human Computer Interaction Through Context. Personal Technologies, 4(2&3), Springer-Verlag, pp. 191–199, June 2000. 20. Schuckmann, C., Kirchner, L., Schümmer, J., Haake, J. M. Designing object-oriented synchronous groupware

with COAST. In Proceedings of the ACM 1996 Conference on Computer Supported Cooperative Work (CSCW'96), ACM Press, pp. 30–38, 1996. 21. SMART Technologies Homepage. (2001). http://www.smarttech.com 22. Stewart, J., Bederson, B. B., Druin, A. Single Display Groupware: A Model for Co-Present Collaboration. In Proceedings of Human Factors in Computing Systems (CHI 99), ACM Press, pp. 286–293, 1999.

LiveBoards, local and desktop environments. In Proceedings of CSCW '94, ACM Press, pp. 345–358, 1994. 25. Streitz, N., Tandler, P., Müller-Tomfelde, C., Konomi, S. Roomware: Towards the Next Generation of Human-Computer Interaction based on an Integrated Design of Real and Virtual Worlds. In J. A. Carroll (Ed.): Human-Computer Interaction in the New Millennium, Addison Wesley, pp. 553–578, 2001.

23. Streitz, N. et al. i-LAND: An interactive landscape for creativity and innovation. In Proceedings of the ACM Conference on Human Factors in Computing Systems (CHI'99), Pittsburgh, Pennsylvania, USA, pp. 120-127, 1999.

26. Tandler, P. Software Infrastructure for a UbiquitousComputing Environment Supporting Collaboartion with Multiple Single- and Multi-User Devices. To appear in Proceedings of UbiComp’01 (Atlanta, USA). Lecture Notes in Computer Science, Springer, Heidelberg, 2001.

24. Streitz, N., Geißler, J., Haake, J., and Hol, J. DOLPHIN: Integrated meeting support across

27. Wacom Homepage (2001). http://www.wacom.com/