Coordination in a Human-Agent Teamwork using Natural Language Communication 

Get Complete Project Material File(s) Now! »

Characteristics of collaborative work

A number of research activities in recent years have offered insights into the collaborative work. Some of the important properties of collaborative works are as follows:
• Transition between Shared and Individual goals: Collaboration needs the active participation of both individuals and team members and thus, requires considerable amount of information exchange between collaborators to be successful. Individuals need to communicate about a shared understanding of goals [Redfern and Naughton, 2002], goal decomposition, sub-goal allocation and of goal/sub-goal progress. It is important for each collaborator to understand what has been done and what is currently being done in the context of goals.
• Sharing Context: This is one of the crucial components for collaborative tasks. This can be the shared knowledge of each other’s current activity, shared knowledge of other’s past activities, shared artefacts and shared environment. Together, this leads to shared understanding [Johnson et al., 2011].
• Awareness of Others: Awareness can be considered as the sense of presence, to feel, to under- stand or to be conscious of the other’s activities, which helps to provide context to your own activities. Awareness, as the sense of presence, involves consideration of peripherals as well as focused attention and more accurately characteristics what occurs when team members are engaged ongoing shared activities [Churchill, 1998]. Awareness can also relate to the activities outside of current task context where one is interested in the activities of a collaborator who is not currently present and who may not be working for the shared tasks.
• Conversation: Conversation is a critical component in collaborative tasks. Conversation can be considered as the process of exchanging messages between team members and have long been recognized as important in collaborative tasks [Mirolli and Nolfi, 2010]. Both non-verbal and verbal behaviors are crucial for the conversation. Studies of non-verbal behaviors take into account facial expression, body posture and gestures, whereas, the verbal conversation, including both text and voice, takes into account the theory of speech act [Searle, 1975], progression of conversation, and natural language processing.
• Interaction: Team members influence the environment which react to those influences resulting in a new state. In the same way, changes in the environment correlate to changes in the team members internal state. How interactions are possible and how to derive the effects of interaction is determined by interaction laws or rules [Parunak et al., 2003].

3C Relationship in human collaboration

In [Falzon, 1994], Falzon mentioned that the collaboration can be seen as an intersection of cooperation, coordination and communication (Figure 2.1, page 23). The author advocated that the role of dialogue in collaborative activities can contribute to realize the cooperation and can also be considered as a way of coordinating behavior. In the following paragraph, we describe the relationship of cooperation and coordination with communication dialogues.
• Cooperation for Communication: The cooperative principle is defended by considering that the participants recognize a common purpose or set of purposes and at each stage what con- versational moves are appropriate and are limited by the purpose or direction. The cooperation principles require that the contributions of each participant should be clear, unambiguous, concise, and relevant in the dialogue context [Grice, 1975]. The dialogue is a collaborative activity in which each participant cooperates to ensure the better functioning of communication.
• Communication for Cooperation: In the context of a task, dialogue is a mean of cooperation necessary for the resolution of a problem. In this vision, the dialogue is an activity in itself. For example, request for information is an speech act which considers not only the context of shared knowledge but also the application of a cooperation principle.

Human-Human Teamwork

The human team can be defined as a distinguishable set of two or more people who interact dynamically, independently, and adaptively towards a common goal [Cannon-Bowers et al., 2001]. The coverage of the collaborative behavior of human teamwork is very vast, which include, collaborative planning, decision-making, conflict handling, collaborative situation awareness, communication, collaborative learning etc. In this section we discuss some of the important aspects of human teamwork. Dimensions: Human teamwork can be divided into three dimensions: cognition, skills, and attitudes [Cannon-Bowers et al., 2001]. Cognition or the knowledge category includes information about the collective activity, goal, specific characteristics of team members, resources and protocols. Human teamwork skills include behaviors such as, communication patterns, performance monitoring, decision-making, interpersonal coordination adaptability, and conflict resolution. Attitude measures the participants’ mutual trust, collective orientation, importance of the team, and team cohesion. Shared Mental model: Research in human team performance suggests that experienced teams develop a shared understanding or shared mental model utilized to coordinate behaviors by anticipating and predicting each other’s needs and adapting to task demands [Fiore et al., 2001,Fiore and Schooler, 2004]. Thus, the shared mental model allows team members to coordinate their activities and better communicate. It includes the knowledge about team objectives, team members’ roles, and along with the procedures of the task. Communication: In human teamwork, communication is a crucial component for developing team- work skills such as group decision-making, listening, information seeking, eliciting, referring, and questioning to clarify uncertainties. Human team members exhibit both reactive and proactive con- versational behavior. The reactive communication behavior also known as ask/reply mode. The proactive communication behavior refers to the mode where people provide information without being asked by others. Another important characteristic of the human teamwork is that the members proactively provide information to other team members based on the anticipation of their needs [Yen et al., 2004b]. Being able to anticipate other team member’s information need is the key aspect of proactive communication behavior. The natural language communication is an important feature of the human teamwork, where team members have the ability to interpret, process, and to produce natural language conversation among them.

Human-Agent Teamwork

Human can collaborate with agents to achieve goals. An effective collaboration can help humans to build and maintain situation-awareness [Yen et al., 2004b], and can make better decisions using information at a greater accuracy and to reduce both individual and team errors. Human-agent teams have been used in a variety of applications e.g., disaster rescue simulations [Schurr et al., 2005], team training in virtual environments [Traum et al., 2003a], personal information management [YorkeSmith et al., 2009], and risk management training [Barot et al., 2013]. Creating shared understanding is one of the important challenge of mixed-initiative human-agent organizations. The limiting factor in most human-agent interactions is the user’s ability and willing- ness to participate in a collective activity [Sierhuis et al., 2007]. Authors formulate the problem of mixed-initiative user interaction as a process of managing uncertainties: (1) managing uncertainties that agents may have about user’s goals and focus of attention, (2) uncertainty that users’ have about agent plans and status. The agents can play different roles in the team while collaborating with human team members, which can be characterized as follow.
Agents supporting team members: In this category, agents aid a single human in completing his/her tasks and do not directly interact with other human team members. The two organizational structures most commonly found in these types of human-agent teams are: 1) each human is sup- ported by a single agent proxy. Agent proxies interact with other agents to accomplish human tasks; 2) each human is supported by a team of agents that work to accomplish the single human directives. Often there are no other humans involved in the task, and the only teamwork involved is between software agents. Examples of such systems include agents assisting humans in allocating disaster rescue resources [Schurr et al., 2005], and multi-robot control systems in which teams of robot perform tasks under the guidance of a human operator, or assisting human users to manage their e-mails.


Formal models of Teamwork Collaboration

It is interesting to analyze what it takes for an agent to cooperate. Number of works have been pro- posed in the literature for the formalization of teamwork. Each of these theories answers this by fo- cusing different aspects of collaboration. [Kinny et al., 1994] states that the team of agents must have mutual belief, joint goal, joint plan and joint intention. [Wooldridge and Jennings, 1999] have proposed the four stages for the cooperation behavior, which includes identifying possibility of collaboration, team formation, plan formation and team execution. Searl defines the philosophical description on team formation [Searle, 1969]. Others including [Jennings, 1996,Tambe, 1997, Singh, 2000] consider practical approaches and provide techniques to implement team on the top of these formalization. In the following sections, we will analyze different approaches of teamwork collaboration.

Joint Intention

The theory of joint intention [Cohen and Levesque, 1990b] focuses on the persistence of joint inten- tions and when agents will drop the intentions. Their notion of joint intention is viewed not only as a persistent commitment of the team to a shared goal, but also implies a commitment on the part of all its members to a mutual belief about the state of the goal. Team members are committed to inform the team when they reach the conclusion that a goal is achievable, impossible, or irrelevant. This obligation to inform others on abandoning the team goal captures one of the basic computational aspect of teamwork. In a collaboration, team members account for the commitments of other, first to the goal, and second to the mutual belief about the state of the goal. This theory predicts that communication is required for an efficient and robust collaboration between team members. Sharing information through communication acts is critical given that each team members often have only partial knowledge relevant to solving the problem, different capabilities, and possibly diverging beliefs about the state of the task. They use constructs of dynamic logic to describe sequences of actions and modal operators to express time associated propositions.

Table of contents :

Table of Contents
List of Figures
List of Tables
1 Motivation
2 Research Issues
3 Aim and objective
4 Approach
5 Contributions of this Thesis
6 Thesis Scope
7 Thesis Outline
I Context & Related Work 
1 Context 
1.1 The CORVETTE Project
1.2 Issues
1.3 Human-Agent teamwork
1.4 Model of Propositional contents
1.5 Multiparty natural language conversation
1.6 Collective Decision Making
1.7 Conclusion
2 Collaboration between Team members 
2.1 Collaboration
2.1.1 Characteristics of collaborative work
2.1.2 3C Relationship in human collaboration
2.2 Team Work
2.2.1 Human-Human Teamwork
2.2.2 Agent-Agent Teamwork
2.2.3 Human-Agent Teamwork
2.3 Taxonomy of Coordination Strategies for Team
2.3.1 Coordination Strategies
2.3.2 Overview of Extended Taxonomy
2.4 Formal models of Teamwork Collaboration
2.4.1 Joint Intention
2.4.2 Shared Plan
2.4.3 Team Plans
2.4.4 Cooperative Problem Solving
2.4.5 Collective Intentions
2.4.6 Cooperative Subcontracting
2.4.7 Discussion
2.5 Conclusion
3 Dialogue Management 
3.1 Introduction
3.2 Coverage of Conversational Acts
3.2.1 Speech Act Theory
3.2.2 From Speech Acts to Dialogue Acts
3.2.3 Typology of dialogues
3.3 Dialogue act Taxonomy
3.3.1 Conversational Act theory
3.3.2 DIT++
3.3.3 Discussion
3.4 Dialogue Management approaches
3.4.1 Finite state based and Frame based approaches
3.4.2 Plan-based approaches
3.4.3 Information state based approach
3.4.4 Discussion
3.5 Dialogue Modeling for Collaboration
3.5.1 Collaborative Dialogue Models
3.5.2 Multiparty Dialogue
3.6 Other components of Dialogue Modeling
3.6.1 Natural Language Generation
3.6.2 Natural Language Understanding
3.6.3 Co-verbal Communicative Behaviors
3.7 Conclusion
4 Cognitive Agent Architectures for Collaborative Virtual Humans 
4.1 Introduction
4.2 Collaborative Virtual Environment for Training
4.3 Cognitive Agents Architecture
4.3.1 Decision Centric Agent architectures
4.3.2 Embodied Virtual Humans
4.3.3 Summary
4.4 Knowledge Representation about Virtual Environment and Human Activities
4.4.1 Characteristics of Knowledge Representation and Organization
4.4.2 Representation of virtual Environment
4.4.3 Representation of Human Activities
4.4.4 Summary
4.5 Semantic Modeling of Virtual Environment: The Mascaret Approach
4.5.1 Model-based Architecture
4.5.2 Current Limitations
4.6 Conclusion
II Contributions 
5 Coordination in a Human-Agent Teamwork using Natural Language Communication 
5.1 Introduction
5.2 Commitment and Shared Plan based Integrated Formal model of team coordination .
5.2.1 Reasoning with Shared Goal Tree and Group Activity Plan
5.2.2 Collective Decision-Making and Intention Update
5.3 Preliminaries and Formal Background for Team Coordination
5.4 Five level Mechanism for Team Coordination
5.4.1 Level 1 : Potential for collaboration
5.4.2 Level2: Team decision for Goal selection
5.4.3 Level 3: Team decision for Plan selection
5.4.4 Level 4 : From Intentions to Joint Commitment
5.4.5 Level5 : Plan Execution
5.5 Effects of Communication on Shared Mental Model
5.5.1 Two-Party Conversation in Multiparty Settings
5.5.2 Multiparty Information Exchange
5.5.3 Proactive information exchange
5.5.4 Proactive and Multi-Party-Proactive Request
5.6 Cultivating collaboration through Dialogues
5.6.1 Dialogues during Recognition of Potential for Cooperation
5.6.2 Dialogues during Team Decision for Goal selection
5.6.3 Dialogues during Team Decision for Plan selection
5.6.4 Dialogues during Constructing Joint Commitment
5.6.5 Dialogues during Plan execution
5.6.6 Receiver’s reaction during Cultivation of collaboration
5.7 Discussion
6 Knowledge Representation 
6.1 Metamodel Based Approach: Motivation and Positioning
6.2 Knowledge Organization
6.3 Modeling Human Activities using Have metamodel
6.3.1 Organization Model
6.3.2 Shared Goal Tree
6.3.3 Group Activity Plan
6.3.4 Discussion
6.4 Semantic Modeling of Virtual Environment using Veha
6.5 Information State based Extended Context Model
6.6 Task-Oriented Dialogue Acts
6.6.1 Identifying Interaction Patterns for Natural Language Dialogues
6.6.2 Extended Task-oriented Information-Transfer Dialogue Acts
6.6.3 Discussion
6.7 Conclusion
7 C2BDI Agent: Conversational and Decision Making behavior 
7.1 Introduction
7.2 Collaborative-conversational BDI Agent architecture
7.3 Conversational Behavior of C2BDI agent
7.3.1 Natural Language Understanding
7.3.2 Dialogue Act Interpretation
7.3.3 Select and update
7.3.4 Proactive conversational behavior
7.4 Model Driven approach for Natural language generation
7.4.1 Generating Natural Language Utterance from Class
7.4.2 Generating response from SGT and GAPs
7.5 Collaborative Conversational Protocols: CCPs
7.5.1 CCP-1
7.5.2 CCP-2
7.5.3 CCP-3
7.5.4 Discussion
7.6 Decision-Making Mechanism
7.7 Identifying Cooperative situations
7.8 Resource Allocation between Team Members
7.9 Conclusion
8 Applications 
8.1 Introduction
8.2 Implementation
8.2.1 Technical Architecture
8.2.2 Belief-Revision
8.2.3 Natural Language Processing
8.2.4 Context Management
8.2.5 Summary
8.3 BrestCoz: An Interactive Virtual Tour of Brest Harbour
8.3.1 Introduction
8.3.2 The Brest’Coz Application
8.3.3 Summary
8.4 AFPA: Knowledge Exchange between Users and Autonomous Agents in a CVET .
8.4.1 Behavioral Architecture to Exchange Knowledge
8.4.2 Shell
8.4.3 Role Exchange Protocol
8.4.4 The Educational Scenario
8.4.5 Implementation
8.4.6 Summary
8.5 Experimental Validation: Furniture Assembly Scenario
8.5.1 Characteristics of the scenario
8.5.2 Semantic Modeling of Virtual Environment
8.5.3 Modeling of Human Activities
8.5.4 Instantiation of the environment and simulation
8.5.5 Welcome to participants
8.5.6 Installation of furniture
8.5.7 Evaluation
8.5.8 Summary
8.6 Conclusion
1 Contributions
2 Perspectives
A Publications


Related Posts