Open Agent Architecture A framework for integrating a

Open Agent Architecture “A framework for integrating a community of software agents in a distributed environment” Ganesh K. Panchanathan 11/1/2020 CS 6704 Design Patterns and Component Frameworks 1

Overview of the Presentation Part I - Ganesh Comparison with distributed technologies and conversational agents The Delegation based computing model of OAA System structure Example showing Interactions between agents Part II - Murali ICL - Inter-Agent Communication Language OAA Triggers Example OAA Application Scalability and Characteristics of OAA 11/1/2020 CS 6704 Design Patterns and Component Frameworks 2

Distributed Computing methodologies Distributed Objects – CORBA, DCOM + Location Transparency - Explicitly coded interactions - RPC style blocking scheme Mobile Agents – Aglets, Voyager + Network Bandwidth + Parallelism - Must specify where to go and what to do - Little support for interactions among agents - Written in language supported by execution environment Blackboard architecture – FLip. Side, LINDA Data Store + Global Data store - Cannot refer to specific process during computation 11/1/2020 CS 6704 Design Patterns and Component Frameworks 3

Conversational Agents Components for effective communication: § § Transport mechanism Interaction protocol Content language Ontology KQML and KIF § § Knowledge Query and Manipulation Language – Interaction Protocol Knowledge Interchange Format – Content Language Limited by reliance on fixed set of atomic performatives Difficulty in arriving at the right set Belief desire and Intention (BDI) § Stronger assumption about the knowledge and processing of agents § Applicability limited by structural requirements imposed on individual agents. § Difficulties in interoperating with legacy systems. 11/1/2020 CS 6704 Design Patterns and Component Frameworks 4

The Delegated Computing model of OAA Similar to above models, in …… § Creation of networked applications like Distributed Objects § Rich and complex interactions like Conversational Agents § dynamic, flexible, and extensible communities like Blackboard, Publish & Subscribe Requests specify ‘what to be done’ not ‘who to do’ or ‘how to do’ Requests delegated to a specialized server agent – FACILITATOR. Coordinates the activities of agents for the purpose of achieving higher-level, often complex problem-solving objectives. In addition to delegation…… § Direct calls to specific agents like distributed objects § Broadcast requests like Blackboard, Publish and Subscribe 11/1/2020 CS 6704 Design Patterns and Component Frameworks 5

Facilitator Agent Facilitator makes use of knowledge in following locations: § REQUESTER which specifies goal and advices how it should be met § PROVIDERS who register capabilities to the facilitator § FACILITATOR which maintains list of provider agents and strategies § META-AGENTS which contain domain or goal specific knowledge Facilitator coordinates with agents to achieve high level problem solving § § MATCHES a request with registered agent(s) DELEGATES the task to the agent(s) COORDINATES the efforts and communication between agents DELIVERS result to the requester 11/1/2020 CS 6704 Design Patterns and Component Frameworks 6

Goals of OAA Interoperation and coordination § § flexibility in assembling communities of autonomous service providers flexibility in structuring cooperative interactions Impose the right amount of structure on individual agents – KQML Vs BDI Include legacy and “owned-elsewhere” applications. Human oriented user interfaces § § More accessible and intuitive user interfaces for non-expert users Speech recognition and pen-based handwriting/gesture recognition Presentation of Systems’ responses using multiple media Collaboration between users and agents Realistic software engineering requirements § Minimize effort to create new agents and Encourage reuse § Support lightweight, mobile platforms § Minimize language and platform barriers 11/1/2020 CS 6704 Design Patterns and Component Frameworks 7

OAA System Structure Facilitator Agents receive ICL requests and coordinate multiagent execution Natural Language Agents produce requests in ICL User Interface Agents accept multimodal i/p and present results Application Agents wrap legacy applications Meta Agents apply domain knowledge to help other agents 11/1/2020 CS 6704 Design Patterns and Component Frameworks 8

Sample Interaction – Automated Office System What is my Schedule? Facilitator 1 8 2 5 3 UI NL 4 11/1/2020 CS 6704 Design Patterns and Component Frameworks 6 7 Cal 9

Another Sample Interaction When mail arrives for me about security, notify me immediately Facilitator 2 3 1 UI NL 4 5 Mail 6 9 7 Notif Cal 8 4. Installing trigger 10 Tel 5. Signal for mail 7. Meta-Agent in action Key Ideas: § § § Dynamic addition of new agents. Distributed interpretation and execution of a task. Single request produces communication across platform and languages 11/1/2020 CS 6704 Design Patterns and Component Frameworks 10

References • Technical White paper of Open Agent Architecture http: //www. ai. sri. com/~oaa/whitepaper. html • D. L. Martin, A. J. Cheyer, and D. B. Moran, "The Open Agent Architecture: A framework for building distributed software systems, " Applied Artificial Intelligence: An International Journal. Volume 13, Number 1 -2, January-March 1999. pp 91 -128. QUESTIONS? ? ? 11/1/2020 CS 6704 Design Patterns and Component Frameworks 11