Understanding Agent Systems

by D'Inverno, Mark; Luck, Michael

• ISBN13: 9783540407003
• ISBN10: 3540407006
• Edition: 2nd
• Format: Hardcover
• Copyright: 2004-01-01
• Publisher: Springer Verlag

Summary

This book helps to organise the diverse landscape of agent-based systems by applying formal methods to provide a defining and encompassing agent framework. The Z specification language is used to provide an accessible and unified formal account of agent systems and inter-agent relationships. In particular, the framework precisely and unambiguously provides meanings for common concepts and terms for agent systems, enables alternative agent models and architectures to be described within it, and provides a foundation for subsequent development of increasingly more refined agent concepts. It describes agents, the relationships between them and the requisite capabilities for effective functioning in multi-agent systems, and is applied in different case studies. In the second edition the authors have revised and updated the existing chapters of the book to respond to advice from readers of the first edition, to add references to recent work in agent systems, and generally to bring the content up to date. They have extended the introduction and conclusions chapters to include a better review of the field and the current state-of-the-art. This new edition features chapters on agent interaction and norms, and outlines an implementation framework. The book will appeal equally to researchers, students and technologists interested in intelligent agents and multi-agent systems. Comments from experts in the field: An excellent book that lays out a clear conceptual framework for studying and analysing agent-based systems. Nick Jennings Mark d'Inverno and Michael Luck have, over the last six or seven years, been at the forefront of European research in agent systems. This book poses some important foundational questions about agents and their interactions in multi-agent systems and answers them in a coherent and convincing way. It's an extremely valuable contribution to the field. Michael Georgeff It is undoubtedly a clear and most comprehensive attempt to describe agent-based systems in a unified manner. Simon Parsons

Author Biography

 Mark d'Inverno has been working in the field of agent-based systems for well over 10 years and is currently Professor of Computer Science in the Cavendish School of Computer Science at the University of Westminster. He gained a BA in Mathematics in 1986 and an MSc in Computation in 1988 both from Oxford University, and in 1998 was awarded a PhD from University College London. He has published numerous papers in the area and is an expert in formal sepcification of agent systems. Prof d'Inverno is a founder of the UKMAS workshops, has chaired two UKMAS wortkshops and is a member of the UKMAS Steering Committee. He sits on several conference and workshop programme committees.Michael Luck is a Senior Lecturer in the Intelligence, Agents and Multimedia Group in the School of Electronics and Computer Science at the University of Southampton, UK. He has worked in the field of agent technology and multi-agent systems for over ten years, having previously led the Agent-Based Systems Group at the University of Warwick for seven years, and having gained his PhD from University College London in 1993 for work on agent-based discovery. Dr Luck is a co-founder of the UK Special Interest Group on Multi-Agent Systems, and currently Chair of the UKMAS Steering Committee. He is a member of the Advisory Boards of FIPA (the agent standards body), MAAMAW (the European agent conference) and CEEMAS (the Central and Eastern European agent conference). He has served on numerous programme committees for agent conferences and workshops, and has organised and chaired several international conferences in the area of agents, including those for industry. Dr Luck has contributed to policy making forums for national and European agencies, has reviewed proposals for many national and international funding agencies, and has published extensively in this area (with over 75 papers and 5 books). Since 2000, he has been Director of AgentLink, the European Network of Excellence for Agent-Based Computing.

Table of Contents

1. The Agent Landscape	1	(14)
1.1 Introduction 1
1.2 Agents	3	(4)
1.2.1 Terminology 3
1.2.2 Problems with Definition 5
1.3 Multi-Agent systems 6
1.4 Desiderata for a Conceptual View of Agents	7	(1)
1.5 A Formal Framework for Agent Definition and Development	8	(7)
1.5.1 Formal Frameworks	8	(1)
1.5.2 Notation	9	(1)
1.5.3 Specification Structure Diagrams	10	(5)
2. The SMART Agent Framework	15	(26)
2.1 Introduction	15	(1)
2.2 Initial Concepts	15	(4)
2.3 Entities	19	(4)
2.3.1 Entity State	20	(3)
2.3.2 Entity Operations 21
2.4 Objects 21
2.4.1 Object Behaviour 22
2.4.2 Object State 22
2.4.3 Object Operations	23	(6)
2.5 Agents 24
2.5.1 Introduction 24
2.5.2 Agent Specification 24
2.5.3 Agent Perception 26
2.5.4 Agent Action 27
2.5.5 Agent State 27
2.5.6 Agent Operations 28
2.6 Autonomy	29	(4)
2.6.1 Introduction	29	(1)
2.6.2 Autonomous Agent specification	30	(1)
2.6.3 Autonomous Agent Perception	31	(1)
2.6.4 Autonomous Agent Action	31	(1)
2.6.5 Autonomous Agent state	32	(1)
2.6.6 Autonomous Agent Operations	32	(1)
2.7 Applying SMART: Tropistic Agents	33	(2)
2.7.1 Tropistic Agents	33	(1)
2.7.2 Reformulating Perception	33	(1)
2.7.3 Reformulating Action	34	(1)
2.7.4 Discussion	35	(1)
2.8 Specification Structure SMART	35	(3)
2.9 Related Work	38	(1)
2.10 Summary	39	(2)
3. Agent Relationships	41	(30)
3.1 Introduction	41	(1)
3.2 Multi-Agent Systems	42	(2)
3.2.1 Multi-Agent System Definition	42	(1)
3.2.2 Server-Agents and Neutral-Objects	42	(1)
3.2.3 Multi-Agent System Specification	43	(1)
3.3 Goal Generation	44	(6)
3.3.1 Discussion 44
3.3.2 Goal Generation Specification	45	(5)
3.4 Goal Adoption 47
3.4.1 Goal Adoption by Neutral-Objects 48
3.4.2 Goal Adoption by Server-Agents	50	(1)
3.4.3 Autonomous Goal Adoption	51	(1)
3.4.4 Autonomous Goal Destruction	52	(1)
3.5 Engagement	53	(8)
3.5.1 Direct Engagement	53	(2)
3.5.2 Direct Engagements in a Multi-Agent System	55	(1)
3.5.3 Engagement Chains	55	(2)
3.5.4 Engagement Chains in a Multi-Agent System	57	(4)
3.6 Cooperation 58
3.6.1 Cooperations in a Multi-Agent System 59
3.6.2 Discussion and Example 60
3.7 The Agent Society	61	(2)
3.8 Agent Relationships Taxonomy	63	(5)
3.8.1 Direct Engagement Relation	63	(1)
3.8.2 Generic Engagement Relation	63	(1)
3.8.3 Indirect Engagement Relation	64	(1)
3.8.4 Generic Ownership Relation	65	(1)
3.8.5 Direct Ownership Relation	65	(1)
3.8.6 Unique Ownership Relation	66	(1)
3.8.7 Specific Ownership Relation	66	(1)
3.8.8 Generic Cooperation Relation	67	(1)
3.9 Summary	68	(3)
4. An Operational Analysis of Agent Relationships	71	(22)
4.1 Introduction	71	(1)
4.2 Initial Concepts	72	(2)
4.3 Making Engagements	74	(5)
4.4 Breaking Engagements	79	(2)
4.5 Joining Cooperations	81	(2)
4.6 Leaving Cooperations	83	(3)
4.7 An Illustrative Example	86	(5)
4.8 Summary	91	(2)
5. Sociological Agents	93	(34)
5.1 Introduction	93	(1)
5.2 Agent store	94	(7)
5.2.1 Applying sMART: Hysteretic Agents	98	(1)
5.2.2 Applying sMART: Knowledge-Based Agents	99	(2)
5.3 Agent Models	101	(9)
5.3.1 Entity Models 101
5.3.2 Sociological Agents	102	(4)
5.3.3 Modelling the Motivations of Others	106	(2)
5.3.4 Modelling the Models of Others	108	(2)
5.4 AgentPlans	110	(14)
5.4.1 Introduction	110	(1)
5.4.2 Plan-Agents	110	(2)
5.4.3 Multi-Agent Plans	112	(4)
5.4.4 Multi-Agent Plan-Agents	116	(1)
5.4.5 Sociological Plan-Agents	117	(3)
5.4.6 An Illustrative Example	120	(4)
5.4.7 Modelling the Plans of Others	124	(1)
5.5 Summary	124	(3)
6. Autonomous Interaction	127	(12)
6.1 Introduction	127	(1)
6.1.1 Speech Acts	127	(1)
6.2 Problems with Autonomous Interaction	128	(2)
6.2.1 Pre-determined Agenda 128
6.2.2 Benevolence 129
6.2.3 Guaranteed Effects 129
6.2.4 Automatic Intention Recognition 129
6.2.5 Multi-Agent Modelling 129
6.2.6 Summary	130	(1)
6.3 A Model of Autonomous Interaction	130	(6)
6.3.1 Sociological Goal Generation	131	(1)
6.3.2 Agent Interaction	131	(2)
6.3.3 Prediction	133	(1)
6.3.4 Experimentation	133	(1)
6.3.5 Observation and Evaluation	134	(2)
6.3.6 Revision	136	(1)
6.4 Summary	136	(3)
7. The Contract Net as a Goal Directed system	139	(16)
7.1 Introduction	139	(1)
7.2 Contract Net Protocol	139	(2)
7.3 Contract Net Components	141	(2)
7.3.1 Nodes	141	(1)
7.3.2 Agents	141	(1)
7.3.3 MonitorAgents	142	(1)
7.3.4 Idle Nodes	142	(1)
7.3.5 Server-Agents	142	(1)
7.4 Contract Net Relationships	143	(3)
7.5 Contract Net State	146	(1)
7.5.1 Task Announcements	146	(1)
7.5.2 Bids	146	(1)
7.5.3 System state	147	(1)
7.6 Contract Net Protocol	147	(5)
7.6.1 Axiomatic Definitions	147	(1)
7.6.2 Making Task Announcements	148	(1)
7.6.3 Making Bids	149	(1)
7.6.4 Awarding Contracts	150	(2)
7.6.5 Terminating Contracts 151
7.7 Summary	152	(3)
8. Computational Architecture for BDI Agents	155	(12)
8.1 Introduction	155	(1)
8.2 AgentSpeak(L)	155	(1)
8.3 Types	156	(3)
8.3.1 Beliefs	157	(1)
8.3.2 Events 157
8.3.3 Plans	158	(1)
8.3.4 Intentions	159	(1)
8.4 AgentSpeak(L) Agents	159	(1)
8.5 AgentSpeak(L) Agent Operation	160	(5)
8.6 Summary	165	(2)
9. Evaluating Social Dependence Networks	167	(16)
9.1 Introduction	167	(1)
9.2 Social Dependence Networks	167	(2)
9.2.1 Action and Resource Autonomy	169	(1)
9.2.2 Dependence Relations	169	(1)
9.2.3 Dependence Situations	169	(1)
9.3 External Descriptions	169	(4)
9.3.1 Introduction.	169	(1)
9.3.2 SDN in SMART	170	(2)
9.3.3 Formalising External Descriptions	172	(1)
9.4 Action and Resource Autonomy	173	(2)
9.5 Dependence Relations	175	(3)
9.6 Dependence situations	178	(2)
9.7 Summary	180	(3)
10. Normative Agents	183	(23)
10.1 Introduction	183	(1)
10.2 Norms	184	(3)
10.2.1 Obligations and Prohibitions	185	(1)
10.2.2 Social Commitments	186	(1)
10.2.3 Social Codes	186	(1)
10.3 Chains of Norms	187	(2)
10.3.1 Norm Instances	187	(1)
10.3.2 Interlocking Norms	188	(1)
10.4 Normative Agents	189	(3)
10.4.1 Normative Multi-Agent Systems	190	(1)
10.4.2 Normative Roles	191	(1)
10.5 Norm Dynamics	192	(3)
10.5.1 Changing Norms	192	(1)
10.5.2 Norm States	193	(2)
10.6 Norm Compliance	195	(11)
10.6.1 Enforcement Mechanisms	195	(11)
10.6.2 Normative Agent State 196
10.6.3 The Norm Compliance Process 197
10.7 Conclusions 198
11. actSMART: Building a SMART System 201
11.1 Introduction 201
11.2 An Example 202
11.3 actsMART Agent Implementation 203
11.3.1 Design Principles 203
11.3.2 Java for Agent Systems 204
11.3.3 Jini 204
11.3.4 eXtensible Markup Language 205
11.4 Agent construction model for actSMART	206	(1)
11.5 Object and Agent Creation in actSMART	207	(4)
11.5.1 Neutral-Object	207	(1)
11.5.2 Autonomous Agents	208	(1)
11.5.3 Example architecture	209	(1)
11.5.4 Engaging Neutral-Objects	210	(1)
11.6 Conclusions	211	(2)
12. Conclusions	213	(6)
12.1 Summary	213	(1)
12.1.1 The SMART Framework	213	(1)
12.1.2 Agent Relationships	213	(1)
12.1.3 Agent Architectures	214	(1)
12.2 Evaluation	214	(5)
12.2.1 Generality	215	(1)
12.2.2 Application	215	(4)
12.3 Concluding Remarks 216
A. The Z Specification Language 219
A.1 Introductionto Z	219	(3)
A.2 Generic Z Definitions	222	(5)
A.2.1 Sets	222	(1)
A.2.2 Relations	223	(1)
A.2.3 Functions	224	(1)
A.2.4 Sequences	225	(2)
References	227	(10)
Index	237