Know more

With digital transition, the concept of the ‘digital twin’, which first appeared in industry in the 1970s, has gained new traction and has undergone many developments, not only in the field of engineering but also in those of research and innovation. 
As part of its Métaprogramme DIGIT-BIO, INRAE is launching a series of digital twin projects to explore the potential applications of this concept in the Institute's areas of expertise.  

What is a digital twin ?

The definitions of digital twins proposed by different sectors and communities vary according to the particular objects and uses involved.                  
We propose to define a digital twin (DT) [i]as the digital representation of an object or system (physical or biological, for example a cell, tissue, plant, animal, farm), with three main characteristics :  

1. A data flow (automated or semi-automated) between the physical object (real entity) and the digital object, acquired using sensors, imaging, networked objects or expert knowledge, allowing the dynamic updating of the digital entity;
2. One or more algorithms (based on machine learning, mathematical modelling and simulations) to assimilate raw data and calculate and update predictions concerning the state of the system and its evolution, integrating the possible impacts of its environment with a high degree of accuracy;
3. A ‘decision-making’ form of feedback to the real entity, either offline (via a human operator running the system) or in real time, which could involve a fully-automated feedback loop.

In short, the digital twin can be defined as ‘a digital representation of an object or a system, updated by a data flow and with predictive and feedback capacity, allowing dynamic intervention in the trajectory of the real system, via either automatic controls or decision making.’

Digital twin: graphical representation of the principle as proposed by the Metaprogramme DIGIT-BIO working group[ii].

[i] This definition was proposed by Metaprogramme DIGIT-BIO’s ‘Digital Twins’ working group.

  [ii] Members: Carole Caranta (INRAE), Michael Chelle (BRGM), Marjorie Domergue (INRAE), Fabien Jourdan (INRAE), Hervé Monod (INRAE), Masoomeh Taghipoor (INRAE), Irène Vigneron Clémentel (INRIA).

Digital twin vs model ?

Digital twins can be distinguished from a classic mathematical models by the following:

• A change of scale (dictated by the complexity of the object or system and its virtual representation) demanding a major advance in technology, and providing a virtual environment that allows work to be carried out on the digital simulation of an individual real entity from different perspectives and at different scales, offering greater predictive accuracy and data precision than standard models;

• A two-way flow of data and actions between the real system and its virtual representation, tracking the kinetics of the system with a high acquisition frequency and possibly in real time.

• An outcome involving material actions affecting the real system : feedback to the same object or system.

Digital Twin also genuinely transforms research practices, allowing experimental designs to evolve dynamically on the basis of data and modelling results.

Finally, because a digital twin can be either an individual or a system, it allows the variability between individuals to be taken into account, opening the door to a new dimension of modelling possibilities. 

Digital Twins and the life sciences : a new framework for research ?

The transfer of the digital twin concept is still emerging when applied to non-industrial areas and outside engineering activities. It raises numerous questions still to be resolved and a degree of scepticism about its application to living beings and systems.
The very use of the term ‘twin’ is itself controversial in this context – should we really use it to describe the digital simulation of a human, animal, or any other living system? Their complexity and variability preclude the strictly identical virtual reproduction of reality, don’t they?

Beyond the term itself – which has now entered general usage despite the ongoing debate surrounding it – the focus  should be the concept of the digital twin. It offers a fresh and potentially transformative framework to develop future interdisciplinary research around an object or system of interest where improved knowledge of the object or system is applied to a practical goal through its digital representation. 

To allow itself a critical exploration of this concept, INRAE took the step in 2021, of establishing a focus group on the topic. The projects launched in 2024 and funded by the DIGIT-BIO metaprogramme will allow us to progress further. They will refine our vision of the possibilities digital twins have to offer when they are applied to living beings, and of the ambitions we might expect tp hold for their development and applications in this field. 

Challenges in methodology to be tackled in Axis 4

• Develop the representation of a small number of systems selected for their interest and for the maturity of their models by drawing inspiration from the concept of the digital twin;
• Identify the data actually required and provide guidance for experiments;
• Explore the behaviour of the systems modelled using simulation 
• Develop human-model interfaces, explore and visualise prediction spaces;
• Adapt processes and steer systems over time in response to changing environments.

This approach will draw on the research carried out in Axes 1, 2 and and 3.