As a response to centralised platforms hoarding user data on the web, Personal Data Stores (PDSs) are becoming more prominent. In this emerging space of decentralized PDS platforms, we see a growing need for automated agents to take over tasks that are normally handled by those data platforms. These intelligent agents already have their place in our daily lives, however, they are currently largely missing from the domain of Personal Data Stores. To address this need, we created a demonstrator software web agent that acts on the real world and personal data via condition-action rules. We applied this demonstrator to a smart home environment use case, where smart home appliances can be actuated and monitored via a personal data store. The generic architecture used by the implementation leads to maximal re-use of the agent, enabling multi-agent systems and researching more complex use cases.
who formalised a hybrid web agent architecture, we provide a reference implementation that
operates on behalf of a user utilizing a PDS. We created a demonstrator Solid Agent that is
capable of reasoning over PDSs and actuates in the context of smart home scenarios. It is
designed to be generic and extensible so that the agent is not limited to the concrete use case
presented in this paper. The agent uses Notation3 (N3) [
Käfer et al. [
Schraudner et al.[
Zimmerman et al. [5] introduced pody, which defines a Web Agent embodied in a Solid pod. Through this embodiment, they envision that podies can participate as part of a MAS. No open source implementations are currently available.
Lastly, Kirrane [
Smart home applications such as security cameras, smart speakers and smart lights are becoming
increasingly popular in households. Ideally, this data would be securely stored and managed in
a personal data vault, so that managed access can be granted to external systems. To achieve
this, we created a web agent based on the hybrid agent architecture [
PShmilaiprstbHuulbe PhBilirpidsgHeue openHAB Figure 1: Setup of the demonstration: The Solid Agent synchronises the state of the Smart Bulb, through openHAB with a resource in the Solid Pod. 3.1. Architecture of the Solid Agent Figure 2 shows the architecture of our agent. For external observers, the Solid Agent acts as one agent. However, internally it consists of multiple individual actors working together, identifiable with their own WebID 4. We implemented two interfacing actors that percept and actuate external things on the web: the OpenHAB Actor and the Solid Actor. At the core, the Orchestration Actor routes the inputs from the interfacing actors to the reasoning engine and executes actions on the resulting output. These actions are actuated by calling other actors such as, but not limited to, the above-mentioned actors.
The OpenHAB Actor monitors an OpenHAB platform for changes in the state of items. The following functions were implemented: (i) retrieve the state of an item, (ii) map an item state object to an RDF representation of the said item, and (iii) subscribe to an item. The agent can be actuated directly to change the state of an item. To accomplish this actuation, two additional functions are implemented: (i) map an RDF representation of the state of an item to an item state object, and (ii) store the state of an item.
The Solid Actor works in the environment of Solid Pods. This agent monitors an ldp:resource for read/write updates.
The Orchestration Actor implements an Interface, Reactive and Controller component [
Solid Agent: Rule-based intelligent software agent
Inbox
AS2 messages AS2 messages
Inbox Inbox
Solid Actor MessageClient
Solid ReadWrite
Client MessageClient
OpenHAB Actor
OpenHAB ReadWrite
Client OpeRnDHFABTranslator
Solid Protocol openHAB
API End User Configures
Inbox
Controler Component Interface Component Reactive Component
Orchestrator Actor
ReadWrite
Client Rules State Plugins
Koreografeye
Reasoner exPtoralicctyor exPeocliuctyer video demonstration can be found in the Solid Agent repository6. This Inbox makes possible agent-to-agent communication, thus enabling the implementation of MAS systems.
In this paper, we presented a web agent demonstration that incorporates several components
from the hybrid architecture proposed in [
Supported by SolidLab Vlaanderen (Flemish Government, EWI and RRF project VV023/10).
Computer Science, Springer International Publishing, Cham, 2020, pp. 197–202. doi:10. 1007/978-3-030-62327-2\_34. [5] A. Zimmermann, A. Ciortea, C. Faron, E. O’Neill, M. Poveda-Villalón, Pody: a Solid-based Approach to Embody Agents in Web-based Multi-Agent-Systems, in: 11th International Workshop on Engineering Multi-Agent Systems (EMAS2023), 2023. [6] J. Snell, E. Prodomou, Activity Streams 2.0, 2017. URL: https://www.w3.org/TR/ activitystreams-core/. [7] R. Verborgh, J. De Roo, Drawing Conclusions from Linked Data on the Web: The EYE Reasoner, IEEE Software 32 (2015) 23–27. doi:10.1109/MS.2015.63, conference Name: IEEE Software.