Human-human Collaboration Formalism for Groupware

Tailorability in Collaborative Augmented Environments

Samir Benbelkacem

, Nadia Zenati-Henda

, Hayet Belghit

, Samir Otmane

and Abdelkader Bellarbi

1,2

Centre de Development Des Technolgies Avancées, Cité du 20 août 1956, Algiers, Algeria

University of Evry, Rue du Pelvoux, 91020 Evry, France

Keywords: Augmented Reality, Collaborative Environment, Multi-Agent System, Groupware, Tailorability.

Abstract: In this paper, human-human collaboration formalism has been proposed to support groupware tailorability for

Collaborative Augmented Environments (CAE). Our work is based on the 3C functional model proposed by

Ellis. This model decomposes the collaboration into communication, coordination and production. This

decomposition has been adapted to design tailorable groupware for CAE. A new concept called “distribution”

has been introduced to consider the properties of collaborative and distributed 3D environment. A new

formalism integrating this concept is proposed in order to adapt a groupware system to the real need of users

evolving in 3D shared scene. Multi-agent technology is, therefore, used to determine the collaboration

protocol between humans, through machines, over the network for implementing the desired tailorability.

1 INTRODUCTION

Computer-based collaborative tools support the

transition from simple human-computer interfacing

to more human-to-human interfacing mediated by

computers. This emphasis on the mediation role of

computers adds new technical challenges to the

development of IT tools. Augmented Reality (AR)

technologies are suited for mediating human-to-

human interactions over the engineered facility life

cycle because the combination of images and

information from the real (field conditions) and

virtual (plans and other engineering information)

sources and the attendant interaction metaphors can

be tailored to enhance group decision-making

processes. AR technology benefits can be maximized

for various situations if the concepts of Computer-

Supported Collaborative Work (CSCW) and

groupware tailorability are incorporated into the

design of AR systems envisaged to mediate human-

human collaborations for shared production tasks.

This paper presents an investigation into how

groupware principles and concepts should be applied

in designing collaborative AR systems in order to

support human-human collaboration.

Recently, researchers have explored how

Collaborative Augmented Environments (CAE) can

provide spatial cues to support group interactions.

Works on this field often adopt approaches based on

desktop computers, HMDs, backpack laptops and

handhled devices. In this sense, several projects have

been realized to develop AR for CSCW applications

such as Shared Space (Billinghurst, 1998),

TransVision (Rekimoto, 1996) and AR pad (Mogilev,

2002). The drawback of the developed applications is

that often use AR libraries in their development and

not based on a software architecture design. Other

applications are based on components-based

architecture design but could not support adding or

modifying functionalities and services within the

application. Computer Supported Collaborative

Work in CAE requires the construction of tailorable

groupware that supports interaction by multiple users.

Tailorable groupware concepts could, then, be

applied in a study for AR mediated human-to-human

collaboration. The research results presented in this

paper could be useful to assist in designing tailorable

groupware for CAE. In this paper, principles and

formalism have been proposed to design tailorable

groupware for AR collaborative applications. Agents-

based approach is proposed to support the human-

human collaboration formalism over Internet.

498

Benbelkacem S., Zenati-Henda N., Belghit H., Otmane S. and Bellarbi A..

Human-human Collaboration Formalism for Groupware Tailorability in Collaborative Augmented Environments.

DOI: 10.5220/0005358604980503

In Proceedings of the 10th International Conference on Computer Graphics Theory and Applications (GRAPP-2015), pages 498-503

ISBN: 978-989-758-087-1

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

2 GROUPWARE

TAILORABILITY

Several researches in CSCW domain show that

tailorability is a fundamental property that should be

taken into consideration when developing

collaborative systems. The authors in (Stiemerling,

1999) define a tailorable application as a system that

can be properly adapted to the changes and diversity

of needs. The authors in (Biemans, 1999) argue that

tailorability is the capacity of an information system

to enable a user to adjust the application to his/her

personal needs, or the task that is being done. The

authors in (Bourguin, 2004) emphasizes that a

tailorable application is both usable and modifiable

by its users. One of the reasons that software should

be tailorable is the complexity of establishing users'

needs before using the application or having a task at

hand. The authors in (Kahler, 2001) provide three

essential reasons for software to be tailorable (1)

Multidimensional diversities that tailorability must

take into consideration in order to implement a

software able to support different uses, (2) the

dynamism of individual and organizational work that

matches the changing nature of work, forces the

software itself to change over time, (3) the uncertainty

and ambiguity due to work practices require the use

of alternative methods to achieve tasks.

For Collaborative Augmented Environment

(CAE) other reasons make that the tailorability is

necessary such as (4) the evolution of AR interfaces

(changing states of 3D visualization data and 3D

interactions, distribution of 3D data and tasks, etc.)

and (5) the constraints due to the AR environment

(tracking, scene recognition, occlusion, brightness,

etc.).

In our work, multi-agent systems will be used to

build tailorable groupware for CAE. In fact, software

agents have been successfully used for implementing

collaborative architectures. They increase the

capacity of systems to become autonomous and

intelligent while exchanging and distributing 3D data

and tasks within Internet. An important benefit of

software agents is their ability to provide flexibility in

human-human collaboration.

2.1 The 3C Model

Our approach is based on the 3C model proposed by

Ellis, shown in Figure 1 (Ellis, 1994). According to

this functional model, a groupware is described by

three specific functions: communication,

coordination and production.

Figure 1: 3C's Ellis model.

The communication space allows actors

exchanging a set of information. The coordination

space defines the tasks to be achieved in order to

produce objects in the production space. The latter

represents the objects resulting from the activity of

the group.

There exists several works that adopt the 3C

model for constructing collaborative applications

(Laurillau, 2002), (Fuks, 2007), (Oliveira, 2007). One

of the advantages of this decomposition is to help

evaluators focus their attention on the

communication, coordination and productions

aspects of the application for identifying usability

problems (Fuks, 2007). In this paper, the

decomposition of Ellis (Ellis, 1994) is adopted in

order to design tailorable groupware for CAE, this,

based on the advantages of software agents concepts.

With this model the three main aspects of the

collaborative work will be preserved.

2.2 Multi-Agent Systems

The authors in (Khezami, 2005) have identified an

agent as a computing object (in the sense of object-

oriented languages) whose behaviour can be

described by a script with its own means of

calculation, and can move from a place to another in

order to communicate with other agents. According

to (Maamar, 2003), an agent is a piece of software that

acts on an autonomous basis. The agent shows a

number of features that differentiate it from other

traditional components, including self-direction,

collaboration, continuity, character, communication,

adaptation, mobility and temporal continuity.

3 HUMAN-MACHINE-HUMAN

COLLABORATION

Software engineering methods give way to new

development paradigms, including component based

and agent-based approaches. These approaches gain a

Human-humanCollaborationFormalismforGroupwareTailorabilityinCollaborativeAugmentedEnvironments

499

lot of attention, where passive software components

are remedied by the dynamics and social character of

software agents. Indeed, agents-based technologies

provide new mechanisms for components in order to

engage in tasks as well as cooperate and process the

requirements of dynamic and heterogeneous

environments.

One of the multi-agent systems developed for

groupware we can cite the C4 model (Khezami, 2005)

dedicated to the collaborative teleoperation through

Internet. This model is based on the PAC* model

(Khezami, 2005) that proposes three agents dedicated

to the three spaces of the 3C model, this, to ensure the

modularity of the system. In addition, the C4 model

proposes a fourth agent: Collaboration agent. The

combination of these four agents constitutes the

"Collaborator Agent". Despite the advantages of this

approach, this model does not consider the

distribution of information, data and tasks, especially,

in AR collaborative environment. It focuses only on

collaboration aspects between communication and

coordination agents. Thus, we propose to enrich this

model by adding “Distributor agent” responsible for

integrating distributed aspects of AR collaborative

environment. We obtain in our configuration the C5

model (Figure 2).

In the other hand, the collaboration using C4

model is limited to communication and coordination

of tasks; the result produced by a collaborator agent

is not communicated to its counterpart, whereas it’s

necessary in VR & AR applications. Indeed,

collaboration in VR/AR concerns usually the

visualization and manipulation of virtual objects

(translation, rotation, zoom, etc). In most of VR/AR

applications, it is necessary for each collaborator to

see the result performed by its counterpart. For some

applications, it is necessary for each collaborator to

have the viewpoint of its counterpart.

Even if (Cheaib, 2011) have proposed a

collaboration protocol based on web services between

machines over the network in order to exchange

common services, this protocol does not allow the

agents, not concerned by the collaboration, to have a

copy of the exchanged services and data. Also, the

exchanged services concern only communication and

coordination results (the production result are not

exchanged). On the contrary, in our case, the

distributor agent play the role of distributing

production results issued form collaboration in AR

distributed system. For example, a user, in

collaborative AR space, may view the tasks

performed by other users and the state of virtual

objects manipulated.

In our case, the C5 model integrates the

relationship between production agents of different

collaborator agents. Therefore, the collaboration is

not restricted to communication and coordination as

presented in the C4 model, but it can be extended to

the production and distribution sides.

Figure 2: Internal interaction in a collaborator agent i.

3.1 Collaboration Formalism

We propose to enrich the C4 model by integrating

“Distributor agent” in the collaboration agent

proposed in (Khezami, 2005) and enriched by

(Cheaib, 2011). Also, we integrate a relationship

between production agents. We, therefore, obtain,

new formal model of collaboration (C5 model) which

is based on multi-agent systems. This model

integrates the properties of software agents and the

characteristics of the production and distribution in

addition to communication and coordination

considered in the C4 model. Moreover, we make the

C5 model more dynamic and proactive.

In the C5 model, a software agent can be

described by a pair of dynamic system < i, w > where

the agent has only one global state in relation to the

collaboration task. A world w of an agent i is dynamic

and changing at every action or reaction of the

collaborator agent. This world w is modelled by:

- E represents the environment in which a

collaborator agent evolves. It’s represented by other

collaborator agents of the environment that

collaborates with this agent.

- Γ is the set of actions produced by an agent that

modifies the world’s evolution. In our configuration,

Γ is composed of four subsets {Inf

}, {Action

{Result

}, {Dist

} (equation 1).

Γ = {{Inf

}, {Action

}, {Result

}, {Dist

}} (1)

Inf

: a set of information sent by the agent i to

other agents j of the environment that collaborates.

Action

: a set of actions executed by the agent i.

Result

: a set of results issued from the execution

of actions of the agent i.

GRAPP2015-InternationalConferenceonComputerGraphicsTheoryandApplications

500

Dist

: a set of information, actions and results of

the agent i distributed to other collaborator agents of

the environment.

The distributor agent is designed such that its

response time and latency should be reduced to a

minimum when it distributes results of production

space.

- Σ is a set of agent states. For us, a collaborator

agent has four states: communicate, coordinate,

produce and distribute (comm, coor, prod, dist).

Equation 2 describes the content of Σ.

Σ = {comm, coor, prod, dist} (2)

- Percept

is a set of stimuli and sensations that

composed of four subsets {Inf

}, {Action

}, {Result

}

and {Dist

} (equation 3). It represents a behaviour

function depending on agent data Fi.

∀

j, Percept

= {{Inf

}, {Action

}, {Result

{Dist

}}

(3)

Inf

: a set of information received by the

collaborator agent i.

Action

: a set of actions executed by the

collaborator agent j of the environment, other than the

collaborator agent i.

Result

: a set of results produced by the

collaborator agent j, other than the collaborator agent

Dist

: a set of information, actions and results of

the collaborator agent j distributed to other

collaborator agents k of the environment which aren’t

in collaboration with the agent j.

- Pi is an agent perception function (equation 4):

set of perceptions that the agent receives.

Pi: Σ  Percept

∀

j, {comm

, coor

, prod

, dist

}  {{Inf

{Action

}, {Result

}, {Dist

}}

(4)

- Fi is an agent behaviour function that determines

the agent's state from its perceptions and its previous

state (equation 5).

Fi: Σ × Percept

 Σ

{comm

, coor

, prod

, dist

} × {{Inf

{Action

}, {Result

}, {Dist

}}  {comm

coor

, prod

, dist

}

(5)

- Infl

is the agent action function, which modifies

the evolution of the world by producing influences

(equation 6). A collaborator agent, being in a given

state, and following this state, could produce

information that changes its evolution and the

evolution of other collaborator agents in the system.

In the other hand, this parameter represents a

production function of the influences depending on

the agent's behavior.

Infl

: Σ  Γ

{comm

, coor

, prod

, dist

}  {{Inf

{Action

}, {Result

}, {Dist

}}

(6)

- R represents a law of evolution of the

collaboration (equation 7).

R: Σ × Γ



{comm

, coor

, prod

, dist

} × {{Inf

{Action

}, {Result

}, {Dist

}}  {comm

coor

, prod

, dist

}

(7)

3.2 Collaboration Process

Figure 3 shows the collaboration process of three

collaborator agents (collaborator agent i, collaborator

agent j and collaborator agent k.) The collaborator

agent i is the agent that starts collaboration process

with the collaborator agent j. All information between

two agents are represented by “Mission

)” that

represents the mission chosen by the communication

agent i and the communication agent j. “Actions

)”

and “Actions

)” are actions defined by

coordination agent i and coordination agent j.

“Resluts

)” and “Results

)” are the results of

execution of tasks of production agents i and j.

A collaborator agent i has four states:

communicate, coordinate, produce and distribute

(comm

, coor

, prod

, dist

). The transition from a state

to another depends on the perceptions that the agent

receives. When receiving a request for collaboration

{Inf

} from an agent j, or the perception of a need to

collaborate that the collaboration agent triggers, the

collaborator agent switches to the communicate state,

where he communicate with its counterpart executing

the communication function (f

icom

∈ Fi) through the

communication agent, this function returns the new

state of the collaborator agent which is either comm

(communicate state) if the communication process is

not completed or coor

if it is. On the coordinate state,

the collaborator agent execute the f

icoor

function ∈ Fi

that allows the collaborator i to coordinate the

operations with its counterpart by the coordinate

agent. Once the coordination ended, the collaborator

agent switches to the produce state (prod

) where the

production agent executes the actions

arising from

the previous state and produces a set of results that are

communicated to the collaboration agent that

instructs the distribution agent, to share the relevant

data and tasks with the “team work”; during this

procedure the collaborator agent passes to the

distribute state (dist

Human-humanCollaborationFormalismforGroupwareTailorabilityinCollaborativeAugmentedEnvironments

501

Figure 3: External interaction between multiple agents in

C5 model. The agent k is not involved in the collaboration

process between agents i and j but it belong to the same

working group of these two agents.

To facilitate the collaboration process and avoid

that the distributor agent produce an explosion of

messages in the system, our collaboration network

was decomposed into collaborators groups. Each

collaborators group share the same augmented

working environment. The corresponding

collaborator agents in the same group communicate

information and coordinate their actions together, and

produce a set of results. Once the collaboration

completed, relevant data and tasks are shared through

the distribution agents of the concerned collaborators

group without affecting the other groups. Thus, each

agent in the group could have access to the results of

the agents that collaborate.

4 GROUPWARE

ARCHITECTURE

In this section we adopt the same approach as

presented in (Cheaib, 2011) to design our software

architecture (Figure 4) for building a tailorable

groupware. The latter is based on the integration of

software agents where the formalism is presented in

the previous section.

Figure 4: Groupware architecture.

We based on the Arch model (Bass, 1992) to

distinct the functional core and the physical interfaces

of the system. Thus, our functional core is connected

to Internet to engage the collaboration protocol with

other users using machines over the Network.

Furthermore, we use Dewan’s model (Dewan, 1999)

in order to construct our groupware system in the

form of shared and replicated layers. The shared

layers are represented by two highest layers (Layers

N and N-1) corresponding to the FC of the system.

The replicated layers are the lowest layer of the

system. They represent the material (physical) level

that implements the interactions with users. Figure 4

shows our groupware architecture based on Dewan’s

model.

Our groupware architecture is composed of a root

representing shared layers, meaning that it is shared

among all the users in the system, and several

replicated layers for every user. The layers

communicate vertically using interaction events, and

use collaboration events (formalism presented) for

the interaction human-machines-human over the

network. The FC in our model is represented by two

layers that are both shared and constitute the root of

the system, in contrast to the clover model (Laurillau,

2002) where the FC is split into two layers: one

GRAPP2015-InternationalConferenceonComputerGraphicsTheoryandApplications

502

private and shared, while the other is replicated and

public.

5 CONCLUSIONS

In this paper, we have proposed a human-machines-

human collaboration formalism to implement

groupware tailorability in collaborative context of

augmented reality. Also, we have suggested software

architecture for groupware based on the proposed

formalism. The originality of our model is the use of

multi-agent approach in order to generate tailorable

and interoperable groupware architecture for

collaborative augmented reality environments. In

fact, the functional breakdown in the software

architecture proposed will result in a greater

modularity which reduces the complexity of

groupware’s implementation.

We believe that the work presented in this paper

is a first step towards shifting the agent technologies’

into tailoring CSCW systems. For our future work,

we aim to extend the collaboration formalism

discussed in this paper to the machine-machine

collaboration over Internet this by using Web services

technology.

REFERENCES

Bass, L., 1992. A metamodel for the runtime architecture of

an interactive system. User Interface Developers’

Workshop, SIGCHI Bulletin, 24(1), 1992.

Biemans, M., Ter Hofte, GH. 1999. Tailorability: state-of-

the- art, Gigaport project deliverable, Telematica

Institut, Netherlands.

Billinghurst, M., Bowskill, J., Jessop, M., Morphett, J.

1998. A wearable spatial conferencing space. In Second

IEEE International Symposium on Wearable

Computers, 76-83, Pittsburgh, PA, USA.

Bourguin, G. 2004. Proposition pour une gestion

dynamique de l'inter-activités dans le TCAO. In

Proceedings of the 16th conference on Association

Francophone d' Interaction Homme-Machine, 191-

194, ACM New York, USA.

Cheaib, N., Otmane, S., Mallem, M. 2011. Tailorable

Groupware Design Based on the 3C Model.

International Journal of Cooperative Information

Systems. 20(4), 405-439.

Dewan, P. 1999. Architectures for collaborative

applications. CSCW Journal, Trends in Software, John

Wiley & Sons, 169–194.

Ellis, C., A Conceptual model of groupware. 1994.

Proceedings of the ACM conference on Computer

supported cooperative work, CSCW’94, New York,

NY, USA, 79-88.

Fuks, H., Raposo, A. B., Gerosa, M. A., Pimentel, M., De

Lucena, C. J. P. 2007. The 3C collaboration model.

Journal of the Encyclopedia of E-Collaboration, Ned

Kock (org), 637 – 644.

Kahler, H. 2001. Supporting collaborative tailoring. PhD

thesis. Department of Communication, Journalism and

Computer Science. Roskilde University, Denmark.

Khezami, N., Otmane, S., Mallem, M. 2005. A new formal

model of collaboration by multi-agent systems.

Proceedings of the IEEE International Conference on

Integration of Knowledge Intensive Multi-Agent

Systems, Massachusetts, USA, 32-37.

Laurillau, Y., Nigay, L. 2002. Clover architecture for

groupware. Proceedings of the ACM conference on

Computer supported cooperative work, NY, USA, 236-

245.

Maamar, Z., Sheng QZ., Benatallah, B. 2003. Interleaving

Web Services Composition and Execution Using

Software Agents and Delegation. Workshop on Web

Services and Agent-Based Engineering.

Mogilev, D., Kiyokawa, K., Billinghurst, M., Pair. J. 2002.

AR Pad: An interface for face-to-face AR

collaboration. In CHI'02, extended abstracts on Human

factors in computing systems, 654-655.

Oliveira, F. F., Antunes, J.C.P, Guizzardi, R.S.S. 2007.

Towards a collaboration ontology. Proceedings of the

2nd Workshop on Ontologies and Metamodels in

Software and Data Engineering.

Rekimoto, J. 1996. Transvision: A hand-held augmented

reality system for collaborative design. In Proceedings

of the International Conference on Virtual Systems and

Multimedia, 85-90.

Stiemerling, O., Hinken, R., Cremers. AB. 1999. The

EVOLVE tailoring platform: supporting the evolution

of component-based groupware. Proceedings of the

IEEE Third International Enterprise Distributed Object

Computing Conference, EDOC'99, Mannheim,

Germany, 106-115.

Human-humanCollaborationFormalismforGroupwareTailorabilityinCollaborativeAugmentedEnvironments

503