Flexibilization of complex Ecosystems using Democratic AI based Decision Support and Recommendation Systems at Work

Project Information

FAIRWork

Grant agreement ID: 101069499

DOI

10.3030/101069499

Project closed

EC signature date 14 June 2022

Start date 1 September 2022

End date 31 August 2025

Funded under

• Digital, Industry and Space

Total cost

€ 3 496 388,35

EU contribution

€ 2 979 375,00

Coordinated by

BOC PRODUCTS & SERVICES AG
Austria

Periodic Reporting for period 2 - FAIRWork (Flexibilization of complex Ecosystems using Democratic AI based Decision Support and Recommendation Systems at Work)

Reporting period: 2024-03-01 to 2025-08-31

Summary of the context and overall objectives of the project

We foster the “democratization” of decision-making in production processes, hence the participation of all involved stakeholders, by introducing a decentralized AI system. Our Democratic AI-based Decision Support System (DAI-DSS) democratically finds the appropriate decision for a concrete situation during production. Each human or technical actor is represented by an agent who negotiates based on the current status provided by the digital shadows and twins. A multiple optimisation algorithm finds the most appropriate solution considering the needs of all involved human and technical stakeholders.

FAIRWork brings “Human, AI, Data and Robots” together,

(a) by introducing Design Decision Models and assessing if those models are “appropriate / fair”,
(b) decentralizes the decision-making, by representing each involved – human or technical – actor within a Multi Agent System and configuring each agent with previously approved decision models, and
(c) broadens the view to optimize the production process, by also introducing social and operational-related parameters, like energy consumption, in addition to the existing technical and business-related aspects. These will be identified and evaluated in user-centred decision making.

Our contribution aims to:

Improve optimization of the production processes.

Strengthening the perception of trust and fairness.

Improve flexibility of European Industry.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

Speed-Up the Configuration of complex Decision-Making: Models speed up the configuration of resource allocation:

Our KER “Knowledge Provisioning Ecosystem” provides a modelling and configuration environments to flexibly compose user interfaces, orchestration of AI services and corresponding connectivity in form of a Low Code / No Code platform. We developed a connectivity meta model and demonstrated its applicability by connecting management software such as the EAM tool ADOIT with LLMS provided by OpenAI and Mistral to flexibly compose connected solutions. This environment is the basis of all demonstrators and most of the prototypes and was used by the end users.

Enable Compliance in Decision-Making by approved Models: Using methods and tools to approve decision models:

Our KER “Data Provisioning Ecosystem” offers the ISO 10303 STEP standard-based repository utilizing the EDMtruePLMTM application, acting as the central hub for data management, and playing a crucial role as the primary repository for all data. It is designed to interface seamlessly with other components, allowing for efficient data exchange, whether it stores results from decision-making or imports data for analysis. Utilizing the EDMtruePLMTM application, the Knowledge Base supports the creation and management of digital twins and digital shadows, ensuring that lifecycle data of physical assets is readily accessible. It integrates with AI services and multi-agent systems, with machine learning algorithms utilizing stored data for training and predictive analytics, and agents’ digital representations linked to their configurations for ongoing performance optimization.

Improve the Co-creation Capabilities: Using physical laboratories enabling participatory and user-centred design:

Our KER “OMiLAB Digital Innovation Environment ” is characterized by a product-service offering, whereas the Scene2Model platform is considered the product and offered as an open-source implementation and corresponding services are provided as added-value capabilities by the community.
The evolution of the KER is organized via community contributions or the world-wide spanning community consisting of 28 laboratories. Any member of the community (organisation or individual can contribute content, functionality, experience/best practice, which are stored within the network and provided based on specific needs and requirements analysis.

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

Democratic AI-based Decision Support System (DAI-DSS)

Decentralized Decision-Making

We developed and advanced the model-based Microservice Integration Framework (OLIVE) that is capable to orchestrate data access orchestrates the connectors to AI services. This platform is the KER “Knowledge Provisioning Ecosystem” which uses cloud agnostic Kubernetes and Docker deployment for the orchestration of data. More advanced components like workflow engines can extend the orchestration layer via an extension mechanisms. Hence our OLIVE platform is agnostic to extensions in form of third party providers such as e.g. Netflix Conductor or the Process Maestro.

Hybrid Modelling for Decisions-Making

The configuration environment evolved to support the configuration on (a) technical connectivity of AI services and data sources and on (b) domain specific knowledge to be used for domain-specific AI. To realise (a) technical connectivity of AI services an own connectivity meta model and corresponding model representation has been developed and used in most of the use case demonstrations. This proprietary language can be extended by workflows and corresponding workflow models such as BPMN. To realize (b) domain specific knowledge, we exported our concept models on domain level via knowledge graph into vector databases which can be combined with LLMs and corresponding MCP servers to compose a domain-specific AI. This was elaborated in the use case to support machine maintenance. While the (a) technical connectivity is ready for going to market, we need more research for (b) domain-specific AI.


Enrich Decision-Making with AI

The current knowledge base collects data based on Open Standards for integrating, storing, and accessing product data from use case site and provides access to AI algorithms. Further AI algorithms have been implemented, and are available in a service repository which consist of a total of 12 algorithms of different maturity levels, some are on experimental level implemented, others are a functional prototypes that are tested and demonstrated by the end users. We expect that in professional settings AI algorithms will be bought from vendors, or will be adapted by third parties. Hence, we focus on the integration and deployment possibilities and see the AI algorithms as exchangeable third party offerings.

Reliability of Decision-Making to Achieve Trust

Our approach was (a) to work out and provide tables consisting transparency measures that foster trust in AI as well as (b) to design and conduct workshops with industrial partners where decision support can be not only worked-out in a co-creative manner but using experiments the effect of AI can be demonstrated. The current status is a transparency matrix for different types of AI-services and meeting-based findings with the use case partners reflection on democratic AI and the usage of AI in their organisations.

2022-09-14-fairwork-logo.png

2022-09-20-fairwork-intro-graphic-v1-1.png