After a discussion of the theory of software agents, this book presents IMPACT (Interactive Maryland Platform for Agents Collaborating Together), an experimental agent infrastructure that translates formal theories of agency into a functional multiagent system that can extend legacy software code and application-specific or legacy data structures.Software agents are the latest advance in the trend toward smaller, modular pieces of code, where each module performs a well-defined, focused task or set of tasks. Programmed to interact with and provide services to other agents, including humans, software agents act autonomously with prescribed backgrounds, beliefs, and operations. Systems of agents can access and manipulate heterogeneously stored data such as that found on the Internet.After a discussion of the theory of software agents, this book presents IMPACT (Interactive Maryland Platform for Agents Collaborating Together), an experimental agent infrastructure that translates formal theories of agency into a functional multiagent system that can extend legacy software code and application-specific or legacy data structures. The book describes three sample applications: a store, a self-correcting auto-pilot, and a supply chain. Contents......Page 6 List of Figures......Page 10 List of Tables......Page 12 Preface and Acknowledgments......Page 14 1 Introduction......Page 16 1.1 A Personalized Department Store Application (STORE)......Page 19 1.2 The Controlled Flight into Terrain Application (CFIT)......Page 23 1.3 A Supply Chain Example (CHAIN)......Page 26 1.4 Brief Overview of Related Research on Agents......Page 29 1.5 Ten Desiderata for an Agent Infrastructure......Page 36 1.6 A Birdseye View of This Book......Page 38 1.7 Selected Commercial Systems......Page 40 2.1 Overview of Architecture......Page 44 2.2 Agent Architecture......Page 46 2.3 Server Architecture......Page 51 2.4 Related Work......Page 65 3.1 Agent Service Description Language......Page 68 3.2 Metric of Service Descriptions......Page 72 3.3 Matchmaking as Nearest Neighbor Computations......Page 75 3.4 Range Computations......Page 81 3.5 Simulation Evaluation......Page 83 3.6 Related Work......Page 87 4.1 Software Code Abstractions......Page 90 4.2 Code Call Conditions......Page 95 4.3 The Message Box Domain......Page 106 4.4 Integrity Constraints......Page 108 4.5 Some Syntactic Sugar......Page 111 4.6 Linking Service Descriptions and Code Calls......Page 113 4.7 Example Service Description Programs......Page 114 4.8 Related Work......Page 120 5 IMPACT Server Implementation......Page 132 5.1 Overview of dbImpact Services......Page 134 5.2 TCP/IP String Protocol for IMPACT servers......Page 137 5.3 Sample Client-Server Interactions......Page 148 6.1 Agent Decision Architecture......Page 156 6.2 Action Base......Page 158 6.3 Action Constraints......Page 171 6.4 Agent Programs: Syntax......Page 174 6.5 Agent Programs: Semantics......Page 186 6.6 Relationship with Logic Programming and Nonmonotonic Logic......Page 212 6.7 Related Work......Page 219 7.1 Extending CFIT by Route and Maneuver Planning......Page 224 7.2 Belief Language and Data Structures......Page 227 7.3 Meta-Agent Programs: Semantics......Page 241 7.4 How to Implement Meta-Agent Programs?......Page 257 7.5 Related Work......Page 264 8 Temporal Agent Programs......Page 266 8.1 Actions with Temporal Duration......Page 268 8.2 Syntax of Taps......Page 276 8.3 Semantics of Taps......Page 282 8.4 Compact Representation of Temporal Status Sets......Page 293 8.5 Computing Feasible Temporal Status Sets......Page 294 8.6 An Application of Tap: Strategic Negotiations......Page 299 8.7 An Application of Tap: Delivery Agents in Contract Net Environments......Page 303 8.8 Related Work......Page 307 9 Probabilistic Agent Programs......Page 310 9.1 Probabilistic Code Calls......Page 312 9.2 Probabilistic Agent Programs: Syntax......Page 318 9.3 Probabilistic Agent Programs: Semantics......Page 321 9.4 Computing Probabilistic Status Sets of Positive paps......Page 330 9.5 Agent Programs are Probabilistic Agent Programs......Page 335 9.6 Extensions to Other Causes of Uncertainty......Page 337 9.7 Related Work......Page 340 10 Secure Agent Programs......Page 344 10.1 An Abstract Logical Agent Model......Page 347 10.2 Abstract Secure Request Handling......Page 353 10.3 Safely Approximate Data Security......Page 362 10.4 Undecidability Results......Page 381 10.5 IMPACT Security Implementation Architecture......Page 383 10.6 Related Work......Page 409 11 Complexity Results......Page 414 11.1 Complexity Classes......Page 415 11.2 Decision Making Problems......Page 422 11.3 Overview of Complexity Results......Page 426 11.4 Basic Complexity Results......Page 432 11.5 Effect of Integrity Constraints......Page 455 11.6 Related Work......Page 474 12 Implementing Agents......Page 476 12.1 Weakly Regular Agents......Page 477 12.2 Properties of Weakly Regular Agents......Page 492 12.3 Regular Agent Programs......Page 502 12.4 Compile-Time Algorithms......Page 508 12.5 The Query Maintenance Package......Page 515 12.6 The IMPACT Agent Development Environment (IADE)......Page 520 12.7 Experimental Results......Page 526 12.8 Related Work......Page 530 13.1 The Army War Reserves (AWR) Logistics Problem......Page 532 13.2 AWR Agent Architecture......Page 534 13.3 AWR Agent Implementation......Page 535 14.1 Progress Towards the Ten Desiderata......Page 544 14.2 Agent Desiderata Provided by Other Researchers......Page 548 A.1 Agents in the CFIT Example......Page 552 A.2 Agents in the STORE Example......Page 556 A.3 Agents in the CHAIN Example......Page 559 A.4 Agents in the CFIT* Example......Page 563 References......Page 570 Index......Page 586 A......Page 587 B......Page 588 C......Page 589 D......Page 590 I......Page 591 P......Page 592 S......Page 593 U......Page 594 Y......Page 595
Software agents are the latest advance in the trend toward smaller, modular pieces of code, where each module performs a well-defined, focused task or set of tasks. Programmed to interact with and provide services to other agents, including humans, software agents act autonomously with prescribed backgrounds, beliefs, and operations. Systems of agents can access and manipulate heterogeneously stored data such as that found on the Internet.After a discussion of the theory of software agents, this book presents IMPACT (Interactive Maryland Platform for Agents Collaborating Together), an experimental agent infrastructure that translates formal theories of agency into a functional multiagent system that can extend legacy software code and application-specific or legacy data structures. The book describes three sample applications: a store, a self-correcting auto-pilot, and a supply chain.
Software agents are the latest advance in the trend toward smaller, modular pieces of code, where each module performs a well-defined, focused task or set of tasks. Programmed to interact with and provide services to other agents, including humans, software agents act autonomously with prescribed backgrounds, beliefs, and operations. Systems of agents can access and manipulate heterogeneously stored data such as that found on the Internet.After a discussion of the theory of software agents, this book presents IMPACT (Interactive Maryland Platform for Agents Collaborating Together), ah experimental agent infrastructure that translates formal theories of agency into a functional multiagent system that can extend legacy software code and application-specific or legacy data structures. The book describes three sample applications: a store, a self-correcting auto-pilot, and a supply chain.
