Periodic Reporting for period 1 - INVERSE (INteractive robots that intuitiVely lEarn to inVErt tasks by ReaSoning about their Execution)
Période du rapport: 2024-01-01 au 2025-06-30
Project Context and Objectives
The INVERSE project addresses a key challenge in robotics: enabling robots to work effectively in real-world environments that are unpredictable and constantly changing. While Artificial Intelligence has made remarkable progress, most robots still struggle when asked to perform tasks outside of tightly controlled settings. Unlike humans, who can adapt and transfer knowledge across different situations, robots lack the cognitive flexibility to do so.
INVERSE aims to change this by developing a new learning framework that allows robots to continuously improve their understanding and performance. Inspired by how humans learn through experience, feedback, and imagination, the project equips robots with the ability to refine their skills over time, adapt to new domains, and recover from mistakes. This makes them more capable of handling complex tasks in dynamic environments, such as modern manufacturing.
Human involvement is central to this approach. The project uses human feedback to guide robot learning, ensuring that the technology remains practical and safe for real-world use. Two industrial use cases will demonstrate how these intelligent robots can support workers and improve productivity.
Pathway to Impact
INVERSE is expected to contribute significantly to the future of human-robot collaboration (HRC). By making robots more adaptable and responsive, the project supports a smoother and more sustainable integration of robotics into everyday work. This has the potential to enhance job quality, reduce physical strain, and open new opportunities for workers in evolving roles.
The project also responds to broader European goals around digital transformation, industrial resilience, and inclusive innovation. Its outcomes are designed to be scalable and relevant across sectors, helping Europe maintain leadership in advanced manufacturing and AI.
Role of Social Sciences and Humanities (SSH)
Social sciences and humanities play a vital role in INVERSE by ensuring that technology development is guided by human values and societal needs. The project uses SSH methods to understand how people interact with robots, what they expect from these systems, and how collaboration can be made more meaningful and trustworthy.
This includes designing robots that are not only technically capable but also perceived as safe and reliable by workers, managers, and unions. The project emphasizes user-centered design, involving stakeholders in every step, from early ideas to real-world testing. It also explores how HRC can support career development and upskilling, making sure that technological change benefits people as well as industry.
The INVERSE project addresses a key challenge in robotics: enabling robots to work effectively in real-world environments that are unpredictable and constantly changing. While Artificial Intelligence has made remarkable progress, most robots still struggle when asked to perform tasks outside of tightly controlled settings. Unlike humans, who can adapt and transfer knowledge across different situations, robots lack the cognitive flexibility to do so.
INVERSE aims to change this by developing a new learning framework that allows robots to continuously improve their understanding and performance. Inspired by how humans learn through experience, feedback, and imagination, the project equips robots with the ability to refine their skills over time, adapt to new domains, and recover from mistakes. This makes them more capable of handling complex tasks in dynamic environments, such as modern manufacturing.
Human involvement is central to this approach. The project uses human feedback to guide robot learning, ensuring that the technology remains practical and safe for real-world use. Two industrial use cases will demonstrate how these intelligent robots can support workers and improve productivity.
Pathway to Impact
INVERSE is expected to contribute significantly to the future of human-robot collaboration (HRC). By making robots more adaptable and responsive, the project supports a smoother and more sustainable integration of robotics into everyday work. This has the potential to enhance job quality, reduce physical strain, and open new opportunities for workers in evolving roles.
The project also responds to broader European goals around digital transformation, industrial resilience, and inclusive innovation. Its outcomes are designed to be scalable and relevant across sectors, helping Europe maintain leadership in advanced manufacturing and AI.
Role of Social Sciences and Humanities (SSH)
Social sciences and humanities play a vital role in INVERSE by ensuring that technology development is guided by human values and societal needs. The project uses SSH methods to understand how people interact with robots, what they expect from these systems, and how collaboration can be made more meaningful and trustworthy.
This includes designing robots that are not only technically capable but also perceived as safe and reliable by workers, managers, and unions. The project emphasizes user-centered design, involving stakeholders in every step, from early ideas to real-world testing. It also explores how HRC can support career development and upskilling, making sure that technological change benefits people as well as industry.
The INVERSE project has advanced the scientific understanding of how robots can reverse known tasks, a capability that is essential for flexible automation and circular manufacturing. A central achievement is the development of the INVERSE ontology, a new framework for representing robotic knowledge that supports both direct and inverse task execution.
Existing research in knowledge representation has not sufficiently addressed the needs of task inversion. To overcome this limitation, the consortium extended the Factory of the Future ontology developed by partner DLR. These extensions include representations for task context, affordances, sensorimotor primitives, and the conditions under which tasks or primitives can be reversed. This enriched ontology enables robots to understand and adapt tasks across different domains, supporting intelligent and resilient automation.
The consortium has also delivered two foundational project documents. The Quality Plan outlines procedures for efficient collaboration and helps avoid duplication of effort. The Data Management Plan builds on the Grant Agreement, Description of Action, and Consortium Agreement to define clear rules for data handling and cooperation among partners.
To ensure practical relevance, the project has defined two industrial use cases. The first use case, developed with Konecranes and Demag, involves a collaborative assembly process where a crane, robot, and human operator work together. The robot assists in positioning components with high precision and attaches loads to the crane’s hook. The system is semi-automated, allowing human intervention at any time. Safety, accuracy, and seamless control are key challenges addressed in this scenario.
The second use case, developed with CRF, focuses on automotive production. It models human-performed assembly tasks using fixed and mobile cameras. The captured data is used to build a digital model of the operations, which supports proactive quality control and enables partial or full automation of disassembly processes. This is particularly valuable for circular manufacturing, where damaged or end-of-life components are remanufactured or recycled.
Existing research in knowledge representation has not sufficiently addressed the needs of task inversion. To overcome this limitation, the consortium extended the Factory of the Future ontology developed by partner DLR. These extensions include representations for task context, affordances, sensorimotor primitives, and the conditions under which tasks or primitives can be reversed. This enriched ontology enables robots to understand and adapt tasks across different domains, supporting intelligent and resilient automation.
The consortium has also delivered two foundational project documents. The Quality Plan outlines procedures for efficient collaboration and helps avoid duplication of effort. The Data Management Plan builds on the Grant Agreement, Description of Action, and Consortium Agreement to define clear rules for data handling and cooperation among partners.
To ensure practical relevance, the project has defined two industrial use cases. The first use case, developed with Konecranes and Demag, involves a collaborative assembly process where a crane, robot, and human operator work together. The robot assists in positioning components with high precision and attaches loads to the crane’s hook. The system is semi-automated, allowing human intervention at any time. Safety, accuracy, and seamless control are key challenges addressed in this scenario.
The second use case, developed with CRF, focuses on automotive production. It models human-performed assembly tasks using fixed and mobile cameras. The captured data is used to build a digital model of the operations, which supports proactive quality control and enables partial or full automation of disassembly processes. This is particularly valuable for circular manufacturing, where damaged or end-of-life components are remanufactured or recycled.
The INVERSE project has begun addressing key scientific and technological challenges related to cognitive robotics and human-robot collaboration. These efforts have already led to several high-quality publications in international conferences and journals, contributing to the advancement of knowledge in the field.
A major focus has been on understanding and improving Human-Robot Interaction (HRI) and Collaboration in industrial settings. The consortium conducted an in-depth analysis of current trends, challenges, and research directions at the intersection of HRI and Cognitive Robotics. This work has helped clarify how robots can better support human workers in dynamic environments.
To overcome limitations in existing imitation learning approaches, the project developed new algorithms that go beyond simply replicating human motion with high precision. These algorithms enable robots to interpret and adapt human actions more flexibly, improving their ability to collaborate and respond to changing conditions.
In addition, the project has made progress in task and motion planning. Several strategies have been developed to allow robots to plan and execute complex tasks more effectively. The consortium also explored the use of Large Language Models to translate natural language instructions into executable robotic plans, opening new possibilities for intuitive human-robot communication.
A major focus has been on understanding and improving Human-Robot Interaction (HRI) and Collaboration in industrial settings. The consortium conducted an in-depth analysis of current trends, challenges, and research directions at the intersection of HRI and Cognitive Robotics. This work has helped clarify how robots can better support human workers in dynamic environments.
To overcome limitations in existing imitation learning approaches, the project developed new algorithms that go beyond simply replicating human motion with high precision. These algorithms enable robots to interpret and adapt human actions more flexibly, improving their ability to collaborate and respond to changing conditions.
In addition, the project has made progress in task and motion planning. Several strategies have been developed to allow robots to plan and execute complex tasks more effectively. The consortium also explored the use of Large Language Models to translate natural language instructions into executable robotic plans, opening new possibilities for intuitive human-robot communication.