MULTIAGENT SYSTEMS by Munindar P. Singh

Abstract. Multiagent systems are drawing much research attention these days. This book is about some of the key technical aspects of multiagent systems. For a variety of reasons, anthropomorphic cognitive concepts can provide useful high-level abstractions for dealing with complex systems. Indeed, reasoning about knowledge has found its way into mainstream distributed computing. However, to be fruitfully applied in computational settings, anthropomorphic terms must be given rigorous technical definitions.

This book develops a semantics for two primitives -- intentions and know-how -- in a general model of actions and time. Key consequences of this semantics include a characterization of the nontrivial conditions under which an agent's intentions may succeed, and the observation that intentions and know-how are essentially independent. These definitions are used also to provide a semantics for the different modes of communication, such as promises and prohibitions. The proposed framework involves actions, possible and actual, abstract and concrete, that agents perform. This enables one to use intentions, know-how, and communications in stating constraints on system behavior that more naturally capture users' requirements. The proposed framework can thus serve as a foundation on which to develop specific approaches and methodologies for specifying, designing, and implementing complex systems.

This book has been made accessible to a wide audience, including graduate students and researchers in computer science (especially, artificial intelligence and distributed computing) and cognitive science. The focus of the work is in computer science, however. No special background is required beyond a familiarity with logic and some mathematical maturity. A knowledge of temporal and modal logic would help, but is not essential.

1 MULTIAGENT SYSTEMS
1.1 Intentions, Know-How, and Communications
1.2 The State of the Art
1.3 Major Contributions

2 TECHNICAL FRAMEWORK
2.1 The Core Formal Framework
2.1.1 The Formal Language
2.1.2 The Formal Model
2.1.3 Semantics
2.2 Temporal and Action Operators: Discussion
2.3 Coherence Constraints
2.4 Results on Time and Actions
2.5 Strategies
2.6 Belief and Knowledge
2.7 More on Actions and Other Events
2.7.1 Events in Natural Language
2.7.2 Trying to Act
2.7.3 Actions and Events in Artificial intelligence
2.8 Rationale for Qualitative Temporal Logic

3 INTENTIONS
3.1 Dimensions of Variation
3.2 Intentions Formalized
3.2.1 Formal Language and Semantics
3.2.2 Axioms for Intentions
3.3 Properties of Intentions
3.4 Desires
3.5 Other Formal Theories of Intentions
3.6 Philosophical Remarks
3.7 Conclusions

4 KNOW-HOW
4.1 Intuitive Considerations on Know-How
4.1.1 Traditional Theories of Action
4.1.2 The Proposed Definition
4.2 Reactive Ability
4.3 Strategic Ability
4.4 Results on Ability
4.5 Incorporating Action Selection: Reactive Know-How
4.6 Strategic Know-How
4.7 Strategic Know-How Defined
4.8 Results on Know-How
4.9 Conclusions

5 COMBINING INTENTIONS AND KNOW-HOW
5.1 Some Basic Technical Results
5.2 Success at Last
5.3 Normal Models
5.4 Other Theories of Ability and Know-How

6 COMMUNICATIONS
6.1 Protocols Among Agents
6.1.1 Speech Act Theory
6.1.2 Speech Act Theory in Multiagent Systems
6.1.3 The Need for a Semantics
6.2 Formal Model and Language
6.2.1 Speech Acts as Actions
6.2.2 Formal Language
6.2.3 Whole-Hearted Satisfaction
6.2.4 Interrogatives
6.3 Applying the Theory
6.3.1 Normative Constraints on Communication
6.3.2 The Contract Net
6.4 Conclusions

Department of Computer Science
North Carolina State University
Raleigh, NC 27695-8206, USA