Engineering World-Wide Multi-Agent Systems with Hypermedia

Abstract

A well studied problem in the engineering of open MASs is to enable uniform interaction among heterogeneous agents. However, AOSE as a field has grown to recognize that a MAS consists of more than only agents and thus should be designed on multiple dimensions (including the environment, organization etc.). The problem of enabling interaction among heterogeneous entities across dimensions is either not considered, or it is addressed in an ad hoc and non-uniform manner. In this chapter, we introduce a novel approach to use hypermedia as a general mechanism to support uniform interaction in MASs. The core idea is that agents use hypermedia to discover at runtime (i) other entities in a MAS (e.g., other agents, tools, organizations) and (ii) the means to interact with those entities (e.g., interaction protocols, APIs). This reduces coupling and enhances the scalability and evolvability of the MAS. We present a demonstrator that supports these claims. We believe that a hypermedia-based mechanism for uniform interaction in MASs could provide a foundation for engineering worldwide MASs.

Long-Lived Agents on the Web: Continuous Acquisition of Behaviors in Hypermedia Environments

“…As a new class of MAS, Hypermedia MAS use hypermedia as the engine of application state (HATEAOS) [8] in order to integrate MAS into the hypermedia fabric of the Web [5,6]: agents are situated in a hypermedia-driven environment in which both autonomous entities (e.g. humans, software agents) as well as nonautonomous entities (e.g.…”

“…Recent work in [5] strongly benefits from the similarity between the A&A meta-model to design artifacts that encapsulate Web Things 5 , carry IRIs and TDs, and expose property, event, and action affordances described in their TDs. Agents can instantiate browser artifacts to retrieve, parse, and map individual interaction affordances of Web Things to actions in their repertoire at run time [5]. However, these abilities are currently limited to low-level interactions with individual APIs and do not yet extend to more complex behaviors.…”

“…Browser artifacts for devices and other artifacts are described along with their interaction affordances using the W3C WoT TD. For describing the agent environment, we use the EVE ontology [5] which we extended to address artifact manuals (eve:Manual) and their relations (eve:explains, eve:hasManual) to artifacts (eve:Artifact).…”

Proactive Digital Companions in Pervasive Hypermedia Environments

“…MAS research could therefore provide the concepts and tooling for designing, programming, and deploying systems of DCs and other agents on the Web-and in particular in pervasive hypermedia environments. The Web has already raised a lot of interest in MAS research with the advent of service-oriented computing in the early 2000s [17], but the use of Linked Data and hypermedia services in MASs is a recent and active topic of research (e.g., [18], [19]). The latter allows for the conceptual integration of MASs with distributed hypermedia environments.…”

Enabling Discoverable Trusted Services for Highly Dynamic Decentralized Workflows

“…• Interfaces (APIs) to microservices are provisioned with SPARQL [5] service descriptions, which offer selfdescribing mechanisms for interactions, exploiting a capability for "Continuous Acquisition of Behaviors" [6], to make key features prominent, based on prior successful usage [7]. RDF triples underlying the SPARQL descriptions can be further augmented by service descriptions provided as Web of Things (WoT) 'Thing Descriptions' (TD) [8] or other JSON-LD [9] format Linked Data descriptions.…”