Now showing 1 - 7 of 7
  • Publication
    Domain-Expert Configuration of Hypermedia Multi-Agent Systems in Industrial Use Cases
    Based on the analysis of two real-world use cases for agriculture and manufacturing, we suggest that Hypermedia Multi-Agent Systems (MAS) are a viable option to interconnect and coordinate devices, services, machine-learning systems, and people in industrial scenarios. We propose and implement an architecture based on three components: an infrastructure that manages Web of Things environments and executes Hypermedia MAS, a visual development environment for programming agents, and a goal specification interface for end-users. While the infrastructure manages information flows between the system components and provides an environment for agents, the visual language enables domain experts to configure the behaviour of the system leveraging agent-oriented programming abstractions both at design time and run time, and the goal specification interface permits users to delegate goals to the running Hypermedia MAS while re-using domain vocabulary.
  • Publication
    Signifiers as a First-class Abstraction in Hypermedia Multi-Agent Systems
    (International Foundation for Autonomous Agents and Multiagent Systems, 2023-05-30) ; ; ;
    Hypermedia APIs enable the design of reusable hypermedia clients that discover and exploit affordances on the Web. However, the reusability of such clients remains limited since they cannot plan and reason about their interactions. This paper provides a conceptual bridge between hypermedia-driven affordance exploitation on the Web and methods for representing and reasoning about actions that have been extensively explored in Multi-Agent Systems (MAS) and, more broadly, Artificial Intelligence. We build on concepts and methods from Affordance Theory and Human-Computer Interaction to introduce signifiers as a first-class abstraction in Web-based MAS: Signifiers are designed with respect to the agent-environment context of their usage and enable agents with heterogeneous abilities to act and to reason about action. We define a formal model for the contextual exposure of signifiers in hypermedia environments that aims to drive affordance exploitation. We demonstrate our approach with a prototypical Web-based MAS where two agents with different reasoning abilities proactively discover how to interact with their environment by perceiving only the signifiers that fit their abilities. We show that signifier exposure based on the dynamic agent-environment context helps to facilitate effective and efficient interactions on the Web.
  • Publication
    HyperBrain: Human-inspired Hypermedia Guidance using a Large Language Model
    We present HyperBrain, a hypermedia client that autonomously navigates hypermedia environments to achieve user goals specified in natural language. To achieve this, the client makes use of a large language model to decide which of the available hypermedia controls should be used within a given application context. In a demonstrative scenario, we show the client's ability to autonomously select and follow simple hyperlinks towards a high-level goal, successfully traversing the hypermedia structure of Wikipedia given only the markup of the respective resources. We show that hypermedia navigation based on language models is effective, and propose that this should be considered as a step to create hypermedia environments that are used by autonomous clients alongside people.
  • Publication
    Agent-Oriented Visual Programming for the Web of Things
    ( 2022)
    Burattini, Samuele
    ;
    Ricci, Alessandro
    ;
    ; ; ; ;
    Croatti, Angelo
    In this paper we introduce and discuss an approach for multiagent-oriented visual programming. This aims at enabling individuals without programming experience but with knowledge in specific target domains to design and (re)configure autonomous software. We argue that, compared to procedural programming, it should be simpler for users to create programs when agent abstractions are employed. The underlying rationale is that these abstractions, and specifically the belief-desire-intention architecture that is aligned with human practical reasoning, match more closely with people’s everyday experience in interacting with other agents and artifacts in the real world. On top of this, we designed and implemented a visual programming system for agents that hides the technicalities of agent-oriented programming using a block-based visual development environment that is built on the JaCaMo platform. To further validate the proposed solution, we integrate the Web of Things (WoT) to let users create autonomous behaviour on top of physical mashups of devices, following the trends in industrial end-user programming. Finally, we report on a pilot user study where we verified that novice users are indeed able to make use of this development environment to create multi-agent systems to solve simple automation tasks.
  • Publication
    Signifiers for Affordance-driven Multi-Agent Systems
    The ecological psychologist James J. Gibson defined the notion of affordances to refer to what action possibilities environments offer to animals. In this paper, we show how (artificial) agents can discover and use affordances in a Multi-Agent System (MAS) environment to achieve their goals. To indicate to agents what affordances are present in their environment and whether it is likely that these may help the agents to achieve their objectives, the environment may expose signifiers while taking into account the current situation of the environment and of the agent. On this basis, we define a Signifier Exposure Mechanism that is used by the environment to compute which signifiers should be exposed to agents in order to permit agents to only perceive signifiers that are likely to be relevant to them, and thereby increase their efficiency. If this is successful, agents can interact with partially observable environments more efficiently because the signifiers indicate the affordances they can use towards which purposes. Signifiers thereby facilitate the exploration and the exploitation of MAS environments. An implementation of signifiers and of a Signifier Exposure Mechanism is presented within the context of a Hypermedia Multi-Agent System and the utility and efficiency of this model is presented through the development of a scenario.
  • Publication
    Signifiers as a First-class Abstraction in Hypermedia Multi-Agent Systems (preprint)
    Hypermedia APIs enable the design of reusable hypermedia clients that discover and exploit affordances on the Web. However, the reusability of such clients remains limited since they cannot plan and reason about interaction. This paper provides a conceptual bridge between hypermedia-driven affordance exploitation on the Web and methods for representing and reasoning about actions that have been extensively explored for Multi-Agent Systems (MAS) and, more broadly, Artificial Intelligence. We build on concepts and methods from Affordance Theory and Human-Computer Interaction that support interaction efficiency in open and evolvable environments to introduce signifiers as a first-class abstraction in Web-based MAS: Signifiers are designed with respect to the agent-environment context of their usage and enable agents with heterogeneous abilities to act and to reason about action. We define a formal model for the contextual exposure of signifiers in hypermedia environments that aims to drive affordance exploitation. We demonstrate our approach with a prototypical Web-based MAS where two agents with different reasoning abilities proactively discover how to interact with their environment by perceiving only the signifiers that fit their abilities. We show that signifier exposure can be inherently managed based on the dynamic agent-environment context towards facilitating effective and efficient interactions on the Web.