Periodic Reporting for period 1 - AI4Work (Human-centric Digital Twin Approaches to Trustworthy AI and Robotics for Improved Working Conditions)
Periodo di rendicontazione: 2024-01-01 al 2025-06-30
AI4Work will investigate practical methods and tools for optimal sharing of work between humans and AI/robots. AI and robotics are likely to be most powerful means for radical improvement of working conditions in diverse domains, as they can support human operators in diverse tasks starting from difficult and tedious manual labor tasks up to complex decision - making tasks. The vision of the AI4Work project is to improve communication and collaboration between humans, AI and robots, allowing for an improvement of the working conditions within different processes in organisations in several domains in terms of increased efficiency of work, reduction in stress upon employees, increased confidence in decision-making process etc. Due to the high level of uncertainty in modern organisations an appropriate balance between human and machine activities must be found. The key assumption is that to cope with the required flexibility and dynamics, Sliding Work Sharing (SWS), where this balance varies during the operation depending on the situational context, machine-based confidence levels and human interactions, is likely to be the most appropriate for modern organisations. The key challenge of the project is to develop a set of common methods and tools (methodology framework, digital twin service platform, SW building blocks for SWS) that can be applied in diverse sectors and with different AI/robotics services, allowing for an effective experience exchanges. The project will make use of living digital twins of working systems as a mean to increase efficiency and trustworthy of AI/robotics solutions. By this, the project, aiming at improved quality of jobs and creating more decent work for human operators, will contribute to the acceptance of the AI/robots support of work in diverse domains. The project will be driven by six pilots in different sectors: logistics, manufacturing industry, construction, healthcare, education, agriculture.
In the initial phase of the project, the consortium has focused on defining the Use Cases and prioritising the associated requirements for the AI4Work solution. These Use Case requirements have been translated into detailed and prioritised technical specifications, which are now guiding the research, development, and innovation activities within the respective work packages.
Evaluation plans have been established to assess the extent to which the developed technologies and services meet the Use Case requirements and deliver the expected benefits for human–AI–robot collaboration. In parallel, data management plans have been defined for handling the datasets provided by the pilot partners as well as those generated in laboratory environments for technology development, integration, and validation.
The project has completed the initial AI4Work system architecture, represented through multiple viewpoints: Logical View, Development View, Process View, and Physical View. This includes analyses of representative industrial deployments of the AI4Work building blocks and services. The design of the core tools and components has been completed, and development activities progressed towards the delivery of early prototypes by June 2025. These prototypes will be evaluated in the pilots by end of 2025.
In parallel with technical progress, the consortium has been active in dissemination activities. Multiple scientific and technical papers have been presented and published, targeting the AI, robotics, and human–machine interaction communities, as well as the relevant industrial sectors. These actions have already contributed to raising awareness of the AI4Work approach to collaborative human–AI–robot work environments.
Evaluation plans have been established to assess the extent to which the developed technologies and services meet the Use Case requirements and deliver the expected benefits for human–AI–robot collaboration. In parallel, data management plans have been defined for handling the datasets provided by the pilot partners as well as those generated in laboratory environments for technology development, integration, and validation.
The project has completed the initial AI4Work system architecture, represented through multiple viewpoints: Logical View, Development View, Process View, and Physical View. This includes analyses of representative industrial deployments of the AI4Work building blocks and services. The design of the core tools and components has been completed, and development activities progressed towards the delivery of early prototypes by June 2025. These prototypes will be evaluated in the pilots by end of 2025.
In parallel with technical progress, the consortium has been active in dissemination activities. Multiple scientific and technical papers have been presented and published, targeting the AI, robotics, and human–machine interaction communities, as well as the relevant industrial sectors. These actions have already contributed to raising awareness of the AI4Work approach to collaborative human–AI–robot work environments.
Human-AI-Robot Work Sharing
• Moves from static task allocation to Sliding Work Sharing (SWS), dynamically adjusting collaboration between humans and AI/robots depending on context, confidence, and skills.
• Surpasses current systems that either over-automate or leave too much burden on humans.
Human-Aware Task Planning
• Introduces task planning methods that explicitly account for human capabilities, workload, preferences, and cognitive/physical states.
• Goes beyond SotA robotic task planning, which typically focuses only on efficiency or feasibility, by embedding models of human performance and well-being into the planning loop.
Human-Centred Digital Twins
• Extends Digital Twin concepts by not only modelling machines and processes but also integrating representations of human states (e.g. stress, fatigue, expertise).
• Goes beyond industrial DTs that mainly focus on equipment, by embedding worker well-being and context awareness .
Context Sensitivity & Decision Support
• Introduces an ontology-based Context Awareness framework to ensure AI systems adapt to specific work environments and user roles.
• Advances beyond SotA in decision support, which often lacks fine-grained adaptation to human variability .
Explainable & Trustworthy AI
• Embeds Explainable AI (XAI) within critical workflows, ensuring transparency and trust for end-users.
• Goes beyond current explainability practices by linking explanations directly to human-machine collaboration and safety-critical contexts .
Safe AI Monitoring & Adaptation
• Proposes methods for continuous monitoring of AI systems in the workplace, with long-term adaptation under uncertainty.
• Extends SotA approaches that typically validate AI models only offline or in limited pilots .
• Moves from static task allocation to Sliding Work Sharing (SWS), dynamically adjusting collaboration between humans and AI/robots depending on context, confidence, and skills.
• Surpasses current systems that either over-automate or leave too much burden on humans.
Human-Aware Task Planning
• Introduces task planning methods that explicitly account for human capabilities, workload, preferences, and cognitive/physical states.
• Goes beyond SotA robotic task planning, which typically focuses only on efficiency or feasibility, by embedding models of human performance and well-being into the planning loop.
Human-Centred Digital Twins
• Extends Digital Twin concepts by not only modelling machines and processes but also integrating representations of human states (e.g. stress, fatigue, expertise).
• Goes beyond industrial DTs that mainly focus on equipment, by embedding worker well-being and context awareness .
Context Sensitivity & Decision Support
• Introduces an ontology-based Context Awareness framework to ensure AI systems adapt to specific work environments and user roles.
• Advances beyond SotA in decision support, which often lacks fine-grained adaptation to human variability .
Explainable & Trustworthy AI
• Embeds Explainable AI (XAI) within critical workflows, ensuring transparency and trust for end-users.
• Goes beyond current explainability practices by linking explanations directly to human-machine collaboration and safety-critical contexts .
Safe AI Monitoring & Adaptation
• Proposes methods for continuous monitoring of AI systems in the workplace, with long-term adaptation under uncertainty.
• Extends SotA approaches that typically validate AI models only offline or in limited pilots .