Christian Duriez is making robots that are even better than the real thing. At work, this Inria researcher regularly finds himself surrounded by specialists in electronics and digital simulation, and software engineers or designers, as well as soft, funny-looking machines made from silicon or elastic polymers. On their desks, prototype soft robots are subjected to rigorous testing in order to explore possible deformations and develop digital models. The researchers working on the team spend their days analysing and anticipating how these robots contract, stretch and twist.
“What we are trying to do is calculate the charges applied to a robot as its shape is changed. The robots are produced using 3D printing, which enables us to test our models and ensure that they are not too far removed from reality.”
The robots Christian Duriez and his multidisciplinary team deal with in their work could be used in a whole host of potential fields in the future. After ten years spent working on surgical stimulation, with a particular focus on the mechanical behaviour of anatomical tissue and how it interacts with a surgeon’s tools, Duriez took the decision to change course in his research.
“For this new subject, we have been able to draw on our expertise in the real-time simulation of deformable bodies while exploring new applications in surgical robotics”.
Modelling robots using software
This new way of designing robots could revolutionise the sector in the next few years.
“Complaints are regularly made in industry that robots are dangerous, but deformable models are much less dangerous. They could be used in sectors seeking robots which are less expensive to manufacture, which have a lot of freedom of movement and which are capable of getting into tight spaces”.
Soft robots are opening up new opportunities for manufacturers and the goal of Defrost’s research moving forward is to continue to model and control them with ever greater levels of precision.
“The main obstacle standing in the way of soft robotics is that current methods used in design and testing aren’t compatible. There are infinitely more variables that you need to analyse than with ‘hard’ robots. That's the challenge. Our team has been working on devising new methods and integrating them into a software platform, the goal being for this platform to become a standard.”
Although there are still serious challenges to be addressed from a research point of view, a number of manufacturers are already starting to realise the potential of soft robots.
“They cushion the majority of collisions. They are made from silicon, and are unbreakable, even in the most hostile environments. Whereas ‘hard’ robots are able to avoid collisions, soft robots use contact with the environment to move forward. The design is totally different.”
As part of their efforts to establish ties with companies, Defrost has been working with the manufacturer TDR on the design of a versatile gripper. “We are looking to work on joint research projects with companies who are not satisfied with the current design of robots and which could benefit from the greater level of flexibility on offer with soft robots”, explains Christian Duriez.
In 2020, the team was awarded the Grand Prize by the panel at hackAtech Lille (an event set up to give a boost to start-ups linked to research into digital science) for Octo+, a robot arm designed to help people with reduced mobility to become more autonomous and more mobile.
Working in partnership with a soft robotics team
Dominique Watier, director of TDR: “The project with Inria came about through a desire to design a flexible, affordable alternative to standard industrial robots, which can be expensive [...] The factory of the future is a vast world just waiting to be explored...We are already thinking about new ideas that we could develop with Defrost!”
Still many obstacles to overcome
Moving towards more complex robots
Like all beings, soft robots must have an awareness of their body and their environment, meaning they must be equipped with sensors. Just like our nerves, these sensors change shape with tissue and send signals (changes in resistance, changes in capacity, etc.), and so you need to be able to simulate the physics of these phenomena in order to select the right sensors and the right locations for them. What’s more, just like muscles, some materials change shape because of electrical potential or when in the presence of a magnetic field. There is a lot of excitement around their use in robotics, the feeling being that they could be used as actuators. But, once again, these need to be modelled in order to test them. With the progress made in manufacturing, we are getting closer and closer to being able to combine fibrous or spongy materials with elastic membranes like the organic tissue found in nature. But how can this complexity be designed and modelled? And what is the best way of using it in a robot? The ideal scenario would be to be able to test these robots in simulations prior to manufacturing them.
Making robots more autonomous through better testing
When it comes to testing robots, you have to be able to drastically reduce the time taken to produce ever more complex models in order to factor in behaviour at high speeds or with vibrations. We are now able to give our robots simple commands, such as getting them to follow a trajectory. But our aim is to develop robots that are more autonomous, and here the field is wide open: if a soft robot is allocated tasks the same way you would with a ‘hard’ robot, you miss out on its capacity to get into tight spaces and to mould itself to its environment. We need to find ways of doing this.
Biography of Christian Duriez
With a degree in engineering (Institut Catholique d’Arts et Métiers de Lille, 2000), a PhD in robotics (University of Évry Val-d'Essonne) and an accreditation to supervise research (University of Lille, 2013), Christian Duriez joined Inria in 2006 as a member of the Alcove project team, working on the interactive simulation of deformable objects and haptic feedback. In 2009, he was made deputy head of the Shacra joint project team, which specialises in surgical simulation. He was one of the founders of InSimo, a start-up launched in January 2013 to promote research that began back in 2006.
He is now head of the Defrost project team, which was set up in January 2015 and specialises in soft robotics. His research is primarily focused on modelling, simulating and testing soft robots using SOFA, software that Christian Duriez jointly developed with a number of Inria teams and which is now one of Inria’s star softwares.
In 2018, Duriez was awarded a Fulbright scholarship to spend 7 months at Stanford University as part of the Inria@SiliconValley programme.