Future robots will be more “softer”, therefore less dangerous for humans

• Future robots will need to be structurally more flexible (in order to interact with humans without risking injury), according to a study published by our partner The Conversation.
• The design of “deformable” robots is increasingly inspired by nature.
• The analysis of this phenomenon was carried out by Christian Duriez, Inria research director (Defrost team) in co-supervision with the University of Lille, CNRS, IMT and Centrale Lille.

–

For several years now, robot design has been evolving towards more flexibility, because to evolve in a complex or changing environment, rigidity can become a handicap.

Manufactured in silicone or plastic, these new robots are no longer designed from rigid articulated skeletons and operated by motors placed at the joints as we traditionally know them. 3D printing today makes it possible to create complex structures made up of rigid and deformable materials, close to organic materials and tissues found in nature such as blood vessels, ears,
honeycomb materials…

3D tire prototype with honeycomb structure © Michelin / 3dnatives.com

To meet some challenges in the field of robotics for industry, such as direct and safe collaboration between operators and robots, miniaturization, for example manipulation of cells, and for medical applications, such as minimally invasive surgical procedures , some researchers are developing new design methods. These soft robots – “flexible” or “soft” robots – also open up prospects in terms of reduction of manufacturing costs, robustness, and safety. This new approach to design could represent a major breakthrough in robotics in the years to come.

Why rigid robots and flexible robots?

Optimizing the choice of materials used is not new in robotics, but in general the design favors maximum rigidity for minimum mass. In this case, we manufacture robots in hollow aluminum or carbon fiber, for example, which do not “shake”, and which can move from one position to another at high speed. This design approach is well suited to certain industrial problems of rate and absolute positioning, such as painting or welding car bodies.

In flexible robotics, we are looking for exactly the opposite. Why ? On the one hand, even in industry, we are moving towards robots that have to work directly with users – these are the famous cobots, whose market is growing. However, too rigid robots moving at high speed can be dangerous in the event of impact and must be placed in protective cages. A design with flexible materials would enhance the intrinsic safety of robots. On the other hand, the traditional “rigid” approach performs very well in a completely open workspace. But if it is necessary to come into contact with the environment, to take support, to grasp, to slip through… then absolute rigidity becomes a handicap.

Inspiration often coming from nature

As often in robotics, the design of deformable robots is inspired by nature: humanoid robots are obviously inspired by humans, there are many versions of robot dogs (rigid), but also
plant-inspired robots. Thus, researchers from Clemson University, pioneers in the field, sought to reproduce the dexterity of movement of an elephant’s trunk with a robotic system made of tendons and springs. The German company FESTO, in other work, created an artificial proboscis using additive manufacturing and compressed air. These projects aim to show that we can reproduce part of the very great dexterity of elephants with their trunk. These robotic arms can perform complex tasks, especially when there are obstacles in the direct environment of the robot.

Festo, “Bionic Handling Assistant”, a flexible pneumatically actuated robot inspired by elephant trunks.

But other invertebrate creatures, like octopuses are another source of inspiration. In underwater robotics, these animals are a role model – they are able to camouflage themselves, to pass through small openings, but also to cling to, grab and manipulate objects.

Other researchers are also interested in caterpillars, to the
towards and even to plants! The objective is to extract the operating principles (moving, sneaking, catching …) to apply them to robots.

Earthworm robot created at MIT. The robot moves forward by creating a wave of contraction along its body …

For which applications?

If researchers and engineers are inspired by nature, they have very specific applications in mind. For industry, soft robotics has already created devices for gripping objects. The design of flexible grippers makes it possible to avoid damaging the products and to be more tolerant of geometric differences, for catching fruits of different sizes, for example. We can cite the
Versaball d’Empire Robotics where the
gripper of the aptly named company
Soft Robotics.

Another very active field on this subject is surgical robotics. Here the flexible robots can navigate the vessels or help squeeze through the abdomen and interact safely, adapting their rigidity to suit the organs and the surgical procedure. For example, taking inspiration from octopuses, researchers have designed a robot that can remain soft, for safe contact with anatomical structures, and can stiffen some of its segments if necessary, to perform surgical tasks. specific. The change in stiffness is obtained thanks to the granular mechanics.

Finally, we are seeing the emergence of projects around assistance to people with reduced mobility: a robotic shower cubicle to gain autonomy, orthoses or pieces of flexible exoskeletons which assist the movement of certain joints such as the ankle or devices to facilitate the transfer of a person from one bed to another.

The challenge of modeling and control

The main obstacle to the emergence of this flexible robotics is that current design and control methods do not work for the deformable. In flexible robotics, the robot deforms for move. You have to be able to analyze an infinitely greater number of movement possibilities than for a rigid robot, that’s the challenge!

In my team DEFROST, we propose different approaches for mechanical modeling of robots, for example to take into account the behavior of flexible materials used to manufacture the robot. We are also developing special algorithms, with computation times short enough to finish these complex computations before the robot needs to move.

An important characteristic of flexible robots is that they use contacts: they weave their way through their environment, they are able to grasp objects better. In the case of orthotics, they are in direct contact with the patient’s skin. While in rigid robotics we work more in collision avoidance, here we are precisely trying to rely on the environment.

Specific methods are therefore needed to control flexible robots. We first proposed “open-loop control” approaches: we consider the digital model of the robot as perfect and we pilot the real robot without taking into account the disturbances of the real world, for example if the cables which deform the robot are also stretch and deform, or if a person leans and deforms the robot’s arm while it is performing a task. Then we used different sensors to correct the robot control and make it more robust (“closed loop control” which corrects the robot’s positions with respect to the model) and even planned re-planning, for example for the robot to try to sneak past the obstacle.

Recently, we have also shown that the model makes it possible to merge different information from sensors to both correct the command but also to provide additional information, such as the measurement of a force that is applied to the robot.

A flexible robot capable of sneaking around. This video shows the result of an algorithm that allows the robot to automatically reach its goal by leaning on an obstacle …

What follow-up to this work?

If we push the imitation of living beings further, the perception that robots have of their body and their environment is very important. By equipping them with sensors, we acquire information on the deformations of the robot or on certain obstacles – a bit like nerves.

In addition, we work with active materials that deform under the effect of an electric potential or a magnetic field, a bit like muscles. The aim is to allow more local control in the robot’s body to facilitate adaptation to the environment.

By relying on these foundations, we will be able to address the real question ofautonomy. There, the site is very open: if we define the tasks on a flexible robot as we define them on a rigid robot, we will not take advantage of its ability to squeeze, to lean, to come and embrace the environment. How to do ?

This analysis was written by Christian Duriez, Inria research director (Defrost team) in co-supervision with the University of Lille, CNRS, IMT and Centrale Lille. The original article was published on the website of The Conversation.

–