A Formal Framework of Virtual Organisations as Agent Societies

We propose a formal framework that supports a model of agent-based Virtual Organisations (VOs) for service grids and provides an associated operational model for the creation of VOs. The framework is intended to be used for describing different service grid applications based on multiple agents and, as a result, it abstracts away from any realisation choices of the service grid application, the agents involved to support the applications and their interactions. Within the proposed framework VOs are seen as emerging from societies of agents, where agents are abstractly characterised by goals and roles they can play within VOs. In turn, VOs are abstractly characterised by the agents participating in them with specific roles, as well as the workflow of services and corresponding contracts suitable for achieving the goals of the participating agents. We illustrate the proposed framework with an earth observation scenario.

💡 Research Summary

The paper presents a formal framework for modelling and operating Virtual Organisations (VOs) in service‑grid environments using an agent‑centric perspective. Traditional VO research often focuses on organisational structures or static contracts, leaving the dynamic creation of collaborative groups under‑specified. To address this gap, the authors abstract agents as entities characterised by a set of goals (the business or technical objectives they wish to achieve) and a set of roles (functional capabilities they can perform within a VO, such as data provision, processing, or validation). This abstraction deliberately separates the logical description of a VO from any concrete implementation technology (e.g., web services, middleware), thereby enabling reuse across diverse grid applications.

A VO is formally defined as a tuple ⟨A, R, W, C⟩ where:

• A is the set of participating agents,
• R maps each agent to the specific role(s) it assumes in the VO,
• W is the workflow – an ordered composition of services that defines data flow and execution order, and
• C is the set of contracts (service‑level agreements) that bind agents to quality, cost, and responsibility specifications.

The core contribution is an operational model that describes how VOs emerge automatically from an “agent society”. The creation process proceeds through six well‑defined stages:

1. Goal articulation – agents publish their goals, which implicitly indicate the services they need.
2. Role discovery – the framework matches required roles with agents capable of fulfilling them.
3. Partner selection – based on role compatibility, cost, and quality metrics, a subset of agents is chosen to form the VO.
4. Workflow synthesis – the selected agents’ service interfaces are composed into a coherent workflow that satisfies all declared goals.
5. Contract negotiation – for each service in the workflow, an SLA is generated, covering performance, reliability, and pricing.
6. Execution – the VO is instantiated on the grid; agents execute their assigned roles under the agreed contracts.

Each stage is expressed with formal inference rules, allowing the process to be automated by a grid management system. The authors argue that this approach supports dynamic, on‑the‑fly formation of VOs, which is essential for environments where service availability and demand fluctuate rapidly.

To illustrate the framework, the paper describes an earth‑observation scenario. Three types of agents are involved: (a) a satellite‑image provider that aims to disseminate the latest imagery, (b) a preprocessing/analysis agent that seeks to transform raw images into actionable information, and (c) a visualization/distribution agent that wants to deliver results to end users. Each agent declares its goal (e.g., “obtain high‑resolution images within 2 hours”) and the roles it can play (e.g., “image retrieval”, “GPU‑accelerated processing”). The framework automatically matches the provider’s image‑delivery role with the processing agent’s consumption role, constructs a workflow “image acquisition → preprocessing → analysis → visualization”, and negotiates contracts that guarantee latency, accuracy, and cost constraints. The resulting VO satisfies all three agents’ goals without any manual orchestration, demonstrating the practicality of the model.

The discussion acknowledges several strengths: (i) a high level of abstraction that promotes domain‑independent VO design, (ii) a formal rule‑based mechanism that enables automated VO creation, and (iii) explicit representation of goals, roles, and contracts that clarifies responsibility and trust among participants. However, the paper also notes limitations. Performance aspects such as the computational overhead of contract negotiation and workflow optimisation are not empirically evaluated. The framework does not provide a concrete conflict‑resolution mechanism for competing goals or overlapping role claims, which could be problematic in more complex business settings. Moreover, the dynamic adaptation of an already instantiated VO (e.g., handling agent failure or the addition of new services) is left as future work.

In conclusion, the authors deliver a rigorous, goal‑and‑role‑centric formalism for VOs that bridges the gap between abstract agent societies and concrete service‑grid deployments. The proposed framework lays a foundation for future research on scalable contract negotiation algorithms, conflict‑resolution protocols, and prototype implementations that validate the model in real‑world grid infrastructures.