Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Multi-agent interactions on the Web through Linked Data Notifications


Published on

Linked Data Notifications for multi-agent systems on the Web at EUMAS 2017

Published in: Internet
  • Get HERE to Read eBook === ebookdfsrewsa.justdied.com2263038698-7-vies-de-michael-schumacher.html
    Are you sure you want to  Yes  No
    Your message goes here
  • Hey guys! Who wants to chat with me? More photos with me here 👉
    Are you sure you want to  Yes  No
    Your message goes here

Multi-agent interactions on the Web through Linked Data Notifications

  1. 1. Multi-agent interactions on the Web through Linked Data Notifications Jean-Paul Calbimonte • Davide Calvaresi • Michael Schumacher Institute of Information Systems University of Applied Sciences and Arts Western Switzerland (HES-SO Valais-Wallis) European Conference on Multi-Agent Systems EUMAS Evry, December 2017 @jpcik
  2. 2. 2 HES-SO: University of Applied Sciences and Arts Western Switzerland Institute of Information Systems
  3. 3. 3 Use Case: an example wellbeing monitoring health recommendations weather information point-of-care restauration services point-of-interest
  4. 4. 4 A Web of Agents Agent Agent Agent Agent Agent Agent Agent Web Standards? Rich Semantics? Decentralized? Agent identifiers? Metadata?
  5. 5. 5 Requirements R1: Standard and extensible messaging. R2: Standard metadata. R3. Asynchronous and distributed communication. R4. Standard Web protocols. R5. Web identifiers. R6. Semantic representation. • any type of data • different formats • representations • Web standards • participants, time constraints, performatives, conditions, etc. • send and receive messages • coordination • no central entity • Web Standards: HTTP • not excluding others • no commitments to a particular agent implementation • URL/URI/IRIs • unicity and de-referenceability. • Web standards (e.g. OWL, RDF), • extensibility and high expressiveness.
  6. 6. 6 Linked Data Notifications • Recently endorsed W3C Recommendation • Decentralized data interchange of notifications on the Web • Generic and simple mechanism to send and consume data • Based on the Linked Data principles • Use URIs to name things • Use HTTP URIs • Get relevant data by dereferencing a URI • URIs linked to other URIs • Usage of RDF for data representation. • Virtually any type of notifications
  7. 7. 7 LDN: Basics Notifications are to/about it Sends notifications Exposes notifications through inbox Creates notifications in inbox Sender Target Consumer Receiver Consumes notifications inbox
  8. 8. 8 LDN: interactions GET/HEAD Sender Target Consumerinbox GET/HEAD inbox Receiver inbox POST GET notifications
  9. 9. 9 LDN: example GET/HEAD Sender Agent Jean-Paul Consumer Agentinbox Receiver Agent inbox POST GET notifications {"message": "Your Talk at EUMAS is at 11:00"} Container of received messages Get all messages Get one particular message
  10. 10. 10 HTTP-based communication • LDN is entirely based on HTTP requests and responses • Natural candidate for most types of exchanges • Other protocols could be used in certain circumstances • e.g. WebSockets for push-subscriptions.
  11. 11. 11 Agent Identification • URIs (or IRIs) are used to identify all entities involved. • Agents identified with URIs • de-referenceable to obtain more information • agent visibility • perception of agent’s environment can be enriched GET • Response: • include metadata about the agent, • e.g. name, scope, endpoint, ontologies, etc. • Each agent can provide an endpoint inbox • Not necessary located within the same environment
  12. 12. 12 Endpoint Discovery • An agent may advertise its inbox • e.g. JSON-LD message content: • Indicates inbox location/other useful metadata • Discovery phase: first interaction between two agents that wish to establish a conversation or initiate a negotiation { "@context": "", "@id": "", "inbox": "" }
  13. 13. 13 RDF data representation • Any representation format and/or model. • LDN agents: use RDF as standard representation framework. • use of URIs for identifiers • using extensible vocabularies • possible to attach explicit semantics • Metadata annotations should be expressed in RDF, • i.e. sender, receiver, performative, protocol, date-time, reply information, conversations, etc. • e.g. (JSON-LD serialization) information about an agree message, indicating the sender agent, receiver, conversation information, etc. { "@id": "ex:agree_request1", "ag:permormative": "ag:Agree", "ag:sender": "ex:agent1", "ag:receiver": "ex:agent2", "ag:reply-to": "ex:agent3", "ag:protocol": "ag:RequestWhen", "ag:conversationId": "ex:conv321", "ag:inReplyTo": "ex:conversation1", "ag:ontology": "", "ag:content": "..." }
  14. 14. 14 Sending agent notifications • LDN agent may POST notifications to an agent inbox endpoint • JSON-LD representation of a call for proposals agent message: POST /agents/health-agent/inbox HTTP/1.1 Host: Content-Type: application/ld+json { "prov:generatedAtTime": "2017-09-14T04:00:00.000Z", "@id": "ex:callForProposals1", "@graph": [ { "@id": "ex:cfp1", "ag:permormative": "ag:CallForProposals", "ag:sender": "ex:agent1", "ag:protocol": "ag:ContractNet", "ag:ontology": "", "ag:content": "..." } ], "@context": { "prov": "", "ex": "", "ag": ""} }
  15. 15. 15 Interaction protocols • example: the Request Interaction Protocol (FIPA)
  16. 16. 16 Interaction Protocols • call for proposals made available to an agent inbox. • accessed by the inbox owner / participants • allows participants to respond to it by sending proposals • proposals can be accepted by the initiator agent. • all messages conform to an agreed RDF structure • exchanged preferably through HTTP, with GET and POST POST GET Inbox Initiator Participant call-for-proposals propose proposal accept-proposal GET POST POST cfp
  17. 17. 17 Publishing inbox elements • Consumers may access an agent inbox • Obtain the messages available there through HTTP requests. • GET over an inbox: return notification URIs • URIs for collections of notifications • URIs for individual notifications • Notification URIs referenced with LDP ldp:contains predicate. • Security, privacy and ownership schemes should be enforced
  18. 18. 18 Agent reasoning • LDN adoption do not affect traditional reasoning engines • Guarantee the retro-compatibility with already existing mechanisms • Foster the development of new ones • Increased capabilities due to the semantic expressiveness • Simplified and extended interactions.
  19. 19. 19 LDN for Agents: Challenges • Adoption of semantically rich messaging mechanisms (e.g. RDF-based) among Agents on the Web. • Usage of ontologies and vocabularies that link existing Web protocols (e.g. LDN) and Agent-communication standards (e.g. FIPA ACL). • Definition and agreement of system-agnostic Agent communication primitives, based on existing MAS languages. • Provision of agent discovery, selection and orchestration services, based on existing standards. • Implementation and adoption of best practices of agent-based mechanisms for Web interactions.
  20. 20. gracias! ¿tienes preguntas? Jean-Paul Calbimonte University of Applied Sciences and Arts Western Switzerland HES-SO Valais-Wallis @jpcik