Fifty shades of soft robots!
© Inria / Photo H. Raguet
On the 28th of May, the French blog "Binaire - l'informatique - la science au cœur du numérique" published an article by Christian Duriez, manager of the Inria Lille - Nord Europe Defrost research team. The article focused on the main research interest of this team : deformable robotics. Citing parallels with the newly-released Disney film, Big Hero 6 , Christian Duriez helps us to understand whether the soft robot Baymax is solely a work of fiction, or whether it is based on real soft robot research. The article has been published jointly with )i(nterstices, an online scientific journal, created by researchers with the aim of bringing computer science to a wider public.
In the film, Disney invites us to visit San Fransokyo, a city of the future and capital of the robotic world, taking its inspiration from both sides of the Pacific Ocean. The inhabitants of this city have an overwhelming passion for robots, and particularly for illegal robot fights. Those providing the financial backing for these fights will stop at nothing to acquire the latest inventions. Hiro, a young genius excluded from the education system, spends his time alternately building combat robots with his brother and carrying out laboratory work on Baymax, a healthcare companion robot who turns out to be the hero of the film. Baymax has one unique characteristic – he’s soft! Resembling an inflatable Michelin man, Baymax is programmed to prevent health problems afflicting the humans around him. Disney based the film on the highly innovative emerging technology of soft robots.
I belong to this new community of researchers working on new types of robot that are radically different from traditional robots built around rigid articulated skeletons moved by actuators at the joints. Soft robots are built from soft deformable materials, such as silicone or rubber, and are therefore capable of modifying their shape and flexibility to suit the task in hand. They can also work in fragile or complex environments, and work alongside humans in complete safety. These robots typically have a degree of flexibility comparable with organic materials, and their design is often inspired by models found in nature, such as the elephant’s trunk, the octopus, earthworms or slugs. These robots are usually manufactured by a moulding process or by 3D printing, a technology that is already capable of forming deformable materials.
Baymax - © Disney
Robotics is above all a science of movement. In traditional robots, this movement is achieved by articulation. In the case of soft robots, movement is generated by deformation, in the same way as muscles. This can be done by injecting compressed air into cavities within the deformable structure of the robot, by injecting liquids under pressure, or by using electroactive polymers that deform under the influence of an electric field. In my research team, we take a simpler approach, using motor-driven cables to apply forces to the structure of the robot in a similar manner to the tendons and ligaments in living organisms.
We can foresee many applications for these soft robots. In surgery, they can navigate through fragile anatomical structures without applying excessive force, in medicine, they can provide exoskeletal supports and ortheses that are more comfortable to wear than current models, and in underwater applications, they can operate in groups like jellyfish to explore the depths at minimal cost. Industrial applications include robust low-cost robots that can interact with humans without danger, while games and leisure activities will benefit from more organic types of robots, capable of more natural movements.
However, our interest in this work arises from the new research challenges that it raises. In particular, we must now totally rethink our approaches to the design, modelling and control of these robots. From movements being limited by a maximum of a dozen or so articulations, we are now entering a world where deformable robots can move with a theoretically infinite number of degrees of freedom. This implies that completely new software tools will be needed in order to develop the full potential of soft robots. This is the fascinating and ambitious task that has been given to our joint Inria / University of Lille 1 Defrost, DEFormable RObotic SofTware team.
Christian Duriez - © Inria / Photo G. Maisonneuve
We are addressing the challenge of modelling and controlling deformable robots. Models already exist for the theories underlying the mechanics of deformable objects, also known as the mechanics of continuous media. In general, these models do not have analytical solutions, and can only be solved approximately using numerical methods such as the finite element method. Moreover, when controlling a robot, a solution to the model must be available continuously in real time or, in practical terms, within a few milliseconds. While this has long been possible in the case of rigid articulated models, it is another case entirely for finite element models. This is the first challenge to be faced in our research work. However other, even more complex challenges are just around the corner. Here are just a few of them.
Once the real time model has been obtained, we then have to find the inverse solution: The model only tells us the deformation in the structure, given the forces that have been applied to it. in order to control the robot, we need to know what forces need to be applied via the actuators (motors, pistons, etc.) In order to cause the robot to deform in the way we wish. While it is hard enough to solve a finite element model in real time, it is even more difficult to obtain its inverse!
The environment surrounding the robot constitutes a further challenge. Unlike a traditional robotic approach in which we seek to avoid collisions and prevent the robot from touching the enclosure surrounding it, soft robots can come into contact with their environment without harming it. In some applications, it is even desirable for the robot to contact its environment. However, it is necessary to control the contact forces applied, especially when working in fragile environments; in surgical applications for example. In this respect, it is also essential to take the fact that the environment can also deform the robot into consideration. All these factors must be taken into account when building the model. When the environment consists of biological tissue, a biomechanical model of that environment is also needed.
A final, and more fundamental, example relates to the use of sensors. In nature, the nervous system provides the feedback of information in the form of vision, touch, hearing, etc. We use this information to control our movements and to understand our environment. It was a small step for engineers to decide to adopt similar sensors in order to adjust and adapt the models used to control robots. The use of control theory provides a mathematical framework for these system engineering methods. The complexity of this theory depends on the number of variables needed to fully describe the state of the system, together with the coupling between these variables. With a theoretically infinite number of degrees of freedom coupled by the mechanics of continuous media, it is obvious that some modifications to this theory are going to be needed.
May 28th 2015, "Binaire", Le Monde
These articles could interest you:
To read more (in French)
For more information
Exo-biote - © Jonathan Pêpe / Julien Guillery
Arts and sciences