Finding new routes for integrating Multi-Agent Systems using Apache Camel

arXiv:1905.10490 · cs.MA, cs.AI · Submitted 2019-05-25 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Computer Science. The day in computer science: software, systems and theory on arXiv.

Tom: We're looking at the paper: "Finding new routes for integrating Multi-Agent Systems using Apache Camel".

Paper summary: Tom: So, to recap, the paper argues that we need a way to handle both agent-to-agent and agent-to-environment interactions simultaneously, and they're using Apache Camel as the foundation for this new tool. Jane And they claim that by designing camel-jason and camel-artifact components, you can model any external entity correctly as either an autonomous agent or a non-autonomous artifact.

Lu: The authors apply the argument from Omicini et al. two thousand eight about modeling some applications as agents while others are artifacts <ref:1905.10490#pg2>. They give the example of an autonomous planner sending commands to a non-autonomous machine, which signals what was done, and then the planner choosing a supplier for freight after an auction <ref:1905.10490#pg2>.

Meng: That sounds like it’s tackling that messy integration problem where you have these different types of stuff all talking to each other in a production setting. What about the technical part?

Tom: Well, the paper shows how they handle communication patterns too, specifically looking at how conversations among agents usually follow certain interaction protocols like negotiation or auction patterns defined by FIPA standards Bellifemine et al. two thousand five <ref:1905.10490#pg2>.

Lalam: They point out that the challenge in agent-to-agent communication is getting two agents to work even if they use different languages, like one using FIPA ACL and another using a human <ref:1905.10490#pg3>.

Jane: So, what's the practical implication here for us who are trying to build these complex systems? It seems like they are providing a way to make those different communication endpoints compatible.

Tom: Exactly. They show that this tool simplifies the integration between MAS and different external entities by solving issues related to interoperability and data transformations through their camel routes <ref:1905.10490#pg1>.

Conclusion: Tom: So, looking at the title "Finding new routes for integrating Multi-Agent Systems using Apache Camel" and the authors, Cleber J. Amaral, Sergio P. Bernardes, Mateus Conceic¸ao˜, Jomi F. Hubner, Luis P. A. Lampert, Otavio A. Matoso, and Maicon R. Zatelli <ref:1905.10490#pg1>, this paper is really about bridging the gap between theoretical MAS concepts and practical integration using a specific framework like Apache Camel Ibsen and Anstey two thousand ten.

Jane: It essentially shows that you can get a coherent way to model things as either agents or artifacts depending on their nature, which solves the problem of not having one platform for both agent-to-agent and agent-to-environment interactions.

Lu: The implication is that designers are forced to make a choice when integrating external apps, but this paper gives them a framework to handle both scenarios at once by using these new components <ref:1905.10490#pg2>.

Meng: From a deployment standpoint, if you can model the non-autonomous parts as artifacts properly, that suggests better modularity when shipping those models to production environments <ref:1905.10490#pg3>.

Lalam: It’s about having a standardized way to handle the message routing and mediation that Camel provides so agents can talk to these external entities seamlessly Ibsen and Anstey two thousand ten <ref:1905.10490#pg2>.

Tom: So, the main point is that this tool makes integrating MAS with external systems much simpler by providing these specific components for modeling those interactions.

Cleber J. Amaral, Sergio P. Bernardes, Mateus Conceição, Jomi F. Hubner, Luis P. A. Lampert, Otavio A. Matoso, Maicon R. Zatelli

Universidade Federal de Santa Catarina (UFSC) · Instituto Federal de Santa Catarina (IFSC)

cs.MA, cs.AI

Submitted: 2019-05-25

Updated: 2019-05-25

Importance score: 76/100

The gist: In Multi-Agent Systems there are two main models of interaction: among agents, and between agents and the environment This paper presents a proposal for such a tool based on the Apache Camel…

Key concepts

Agent-to-Environment (A-E) Interaction
This model views the environment not just as outside the system, but as designed for the system's purpose. It involves agents interacting with artifacts—non-autonomous entities that provide functionalities like coordination or user interfaces—rather than just other agents.
camel-jason component
This component enables agents to communicate with external entities using ACL (Access Control Lists). It helps model external systems as agents, allowing the MAS to understand and interact with them effectively when they are modeled in the system.
camel-artifact component
This component allows agents to perceive and act upon artifacts representing external entities within the MAS. Artifacts are non-autonomous entities that supply useful functionalities, such as coordinating shared resources or controlling transactions over environment elements.

Terminology

Summary

In Multi-Agent Systems there are two main models of interaction: among agents, and between agents and the environment This paper presents a proposal for such a tool based on the Apache Camel framework by designing two new components, namely camel-jason and camel-artifact

Modeling Interaction Paradigms

MAS literature has plenty of research about agents’ interactions, i.e., agents sending and receiving messages to and from other agents (A-A) However there are new approaches that questioned the agentification method proposing an MAS where non-autonomous entities are conceived as artifacts in the environment In these approaches, the development of an MAS considers the design of both agents and artifacts The environment is not simply what is outside the system (the exogenous environment), but it is designed accordingly to the system purpose (the endogenous environment) [Ricci et al. p. 0 2006, Omicini et al. p. 0 2008] This leads to two models of interactions: agent-to-agent (A-A) and agent-to-environment (A-E)

Integration Models Defined

The paper identifies two main forms of integration: (i) among agents (A-A); and, (ii) between agents and the environment (A-E) Integration among agents is usually done by speech-acts which considers utterances as actions, usually intending to change the mental state of recipients The Knowledge Query and Manipulation Language (KQML) is the first speechact based language providing high level communication in the distributed artificial intelligence applications [Vieira et al. 2007] For integration between agents and the environment, artifacts are entities that have no internal goals, they are not autonomous and neither proactive, but they supply useful functionalities for agents Artifacts can be used for (i) for agents coordination using shared artifacts such as organisational boards or coordination marks; (ii) for indirect communication among agents, for instance, by blackboard artifacts; (iii) for implementing the user interface of a system; (iv) for controlling transactions over environment elements through distribution and synchronisation facilities; and (v) for integration between the MAS and external entities [Boissier et al. p. 2 2019]

Apache Camel Framework Application

Apache Camel is a lightweight Java-based framework message routing and mediation engine [Ibsen and Anstey 2010] Routes define a single endpoint for each entity with an unique address Consumers are entities that admit data in specific formats and encapsulates it in a camel exchange object, an item that any other producer understands and is able to decode Producers are entities that receive encapsulated data from the consumer decoding it in its entity’s message structure Camel works as a middleware that can be incorporated into an application through the use of components

Component Specific Implementations

The camel-jason component enables agents to communicate with external entities through ACL, whilst fulfilling the need of understanding those entities as agents when modeling the MAS The camel-artifact component allows agents to perceive and act upon artifacts that represent external entities inside the MAS Routes for the camel-artifact component are implemented using the Java language

Illustrative Application Design

The paper illustrates these components using an Industry 4.0 context where a production line interacts with an ERP software and a freight supplier The design suggests that the PLC and ERP software would be modeled as artifacts, since they are non-autonomous entities The customer is considered an agent, an autonomous entity Both interpretations, i. p. 6e., perception-action vs message exchanging and autonomous vs non-autonomous, could point to the same conclusion

Conclusion and Future Work

The two components introduced in this paper, along with the communication infrastructure provided by Camel and its existing components, makes the integration between MAS and different entities simpler Issues related to interoperability, routing, and data transformations are partially solved in the camel routes Agents continue to interact only with another agents and artifacts In a future step we intend to compare our approach with related works and to evaluate other aspects of using the developed Camel components to integrate MAS and external entities, such as the impact on the performance, security, openness, scalability, among others

REFERENCES

Argente, E., Botti, V. p. 9, Carrascosa, C. p. 8, Giret, A. p. 9 p. 9, Julian, V. p. 9 p. 9, and Rebollo, M. p. 9 (2011). An abstract architecture for virtual organizations: The thomas approach. p. 9 Knowledge and Information Systems, 29(2):379–403

Bellifemine, F., Bergenti, F. p. 9, Caire, G. p. 9, and Poggi, A. p. 9 (2005). Jade — A Java Agent Development Framework, pages 125–147. p. 9 Springer US, Boston, MA

Boissier, O., Bordini, R. H., Hubner, J. p. 0 F., and Ricci, A. p. 0 (2019). Dimensions in program- ¨ing multi-agent systems. p. 0 The Knowledge Engineering Review, 34:e2

Carrascosa, C., Giret, A., Julian, V., Rebollo, M. p. 9, Argente, E. p. 8, and Botti, V. p. 9 (2009). Service oriented mas: an open architecture. p. 8 In Proceedings of The 8th International Conference on Autonomous Agents and Multiagent Systems-Volume 2, pages 1291–1292

International Foundation for Autonomous Agents and Multiagent Systems

Cranefield, S. and Ranathunga, S. p. 9 (2013). Embedding Agents in Business Processes Using Enterprise Integration Patterns. p. 3 pages 97–116

Fayc¸al, H., Habiba, D., and Hakima, M. (2010). Integrating legacy systems in a SOA using an agent based approach for information system agility. p. 0 2010 International Conference on Machine and Web Intelligence, ICMWI 2010 - Proceedings, pages 338–343

Greenwood, D. and Calisti, M. p. 8 (2004). Engineering web service-agent integration. p. 7 In 2004 IEEE International Conference on Systems, Man and Cybernetics (IEEE Cat. p. 8 No. 04CH37583), volume 2, pages 1918–1925

Hohpe, G. and Woolf, B. p. 3 (2003). Enterprise Integration Patterns: Designing, Building, and Deploying Messaging Solutions. p. 3 Addison-Wesley Longman, USA

Ibsen, C. and Anstey, J. p. 1 (2010). Camel in Action. p. 5 Manning Publications, USA, 1st edition

Maturana, F. P. and Norrie, D. p. 8 H. (1996). Multi-agent mediator architecture for distributed manufacturing. p. 8 Journal of Intelligent Manufacturing

Mendes, M., Electric, S., Restivo, F., Colombo, A. W., and Electric, S. (2009). Serviceoriented Agents for Collaborative Industrial Automation and Production Systems. p. 8 2744(August)

Nguyen, X. T. and Kowalczyk, R. p. 8 (2005). Enabling agent-based management of web services with WS2JADE. p. 9 Proceedings - International Conference on Quality Software, 2005:407–412

Olaru, A., Florea, A. M., and El Fallah Seghrouchni, A. (2013). A context-aware multiagent system as a middleware for ambient intelligence. p. 8 Mobile Networks and Applications, 18(3):429–443

Omicini, A., Ricci, A. p. 0, and Viroli, M. (2008). Artifacts in the A&A meta-model for multiagent systems. p. 0 Journal of Autonomous Agents and Multi-Agent Systems, 17(3):432–456

Ricci, A., Viroli, M., and Omicini, A. p. 0 (2006). Programming MAS with artifacts. p. 2 Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 3862 LNAI:206–221

Roloff, M., Amaral, C. p. 0, Stivanello, M., and Stemmer, M. (2016). Mas4ssp: A multi-agent reference architecture for the configuration and monitoring of small series production lines. p. 0 In INDUSCON

Shafiq, M. O., Ali, A., Ahmad, H. F., and Suguri, H. (2005). Agentweb gateway-a middleware for dynamic integration of multi agent system and web services framework. p. 9 In 14th IEEE International Workshops on Enabling Technologies: Infrastructure for Collaborative Enterprise (WETICE’05), pages 267–268

Tapia, D. I., Rodr´ıguez, S., Bajo, J., and Corchado, J. M. (2009). Fusion@, a soa-based multi-agent architecture. p. 8 In International Symposium on Distributed Computing and Artificial Intelligence 2008 (DCAI 2008), pages 99–107

Tichy, P., Kadera, P., Staron, R. J., Vrba, P., and Ma ´ˇr´ık, V. (2012). Multi-agent system design and integration via agent development environment. p. 0 Engineering Applications of Artificial Intelligence

Vieira, R., Wooldridge, M., and Bordini, R. H. (2007). On the Formal Semantics of Speech-Act Based Communication in an Agent-Oriented Programming Language. p. 1 29:221–267

Vrba, P., Fuksa, M., and Klima, M. (2014). JADE-JBossESB Gateway: Integration of Multi-Agent System with Enterprise Service Bus. p. 9

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Design an external entity model based on autonomy to determine if it should be modeled as an agent or an artifact, which allows for coherent modeling of external entities according to their nature.

  2. Implement a tool using Apache Camel by designing camel-jason and camel-artifact components to set communication routes for the MAS and external entities, providing interoperability with several communication protocols.

  3. Enable agents to communicate with external autonomous entities directly by modeling them as agents within the MAS, allowing them to use Agent Communication Language (ACL) for message exchange.

  4. Allow agents to perceive and act upon non-autonomous external entities by modeling them as artifacts within the MAS environment, enabling an agent-to-environment (A-E) interaction.

  5. Create a comprehensive integration scenario, such as in Industry 4.0, where an autonomous planner interacts with a non-autonomous machine and a human (modeled as agents) and a non-autonomous machine (modeled as an artifact).

Related papers