A slippery world in bio-inspired robotics
A roboticist at the Max Planck Institute for Intelligent Systems studies the multifunctional feet of the desert locust and its jumping behavior on different surfaces to extract the traits which contribute to enhancing surface friction and stop slips. The scientist then built a robot inspired by the locust. His findings about the morphological intelligence of the insect contribute to solving the complex locomotion problems seen in even the most advanced robots.
Stuttgart – Animals in natural environments come into contact with many different surfaces, where the spectrum ranges from very smooth to very rough. Whereas humans wear different shoes to avoid slipping on different terrains, an animal or insect foot must be multi-functional for it to be able to quickly escape from any surface it might sit or stand on. This sophisticated adaptation increases its chances of survival.
The desert locust, scientifically known as Schistocerca gregaria, is one such well-adapted insect. Its feet have both wet adhesive pads and spines, which allow it to jump quickly and robustly from any surface. That´s why Matthew Woodward chose it for his research project. He is a scientist in the Physical Intelligence Department at the Max Planck Institute for Intelligent Systems (MPI-IS) in Stuttgart. His publication, co-authored by the head of the department Metin Sitti, is titled “Morphological intelligence counters foot slipping in the desert locust and dynamic robots“. It was accepted to appear in the Proceedings of the National Academy of Sciences or PNAS. Together with Nature and Science this scientific journal is among the top three interdisciplinary journals worldwide.
Woodward, who is a mechanical engineer, roboticist and biologist, studied the foot structure of the desert locust and its jumping behavior on different surfaces to extract the traits which contribute to helping the insect gain a grip (by enhancing friction) after slipping. He let the locust jump from surfaces such as hydrophobic glass, mesh, wood or sandstone. He wanted to understand how the insect can easily jump from both a smooth and a rough surface. To do so, he built a robot (see figure 1) so he could imitate the jumping behavior of the insect. He built a lot of locust inspiration into it, however also made changes to the natural blueprint: to restrict the motion of the legs to that necessary for jumping, Woodward mirrored the segments of the locust leg and installed this doubling.
Woodward then tested this bio-inspired robot in his laboratory at the MPI-IS. From the results of his many tests, he concluded that for the real thing, the locust, slipping is not an anomaly, but part of the dynamic locomotion of the insect. Woodward then drew conclusions for his robot: in the absence of control or computational “brain power”, morphological intelligence encoded into the physical system (the robot´s structure) can reduce the severity of slipping. The robot with its intelligent structure – even though it slipped – can preserve its locomotion performance on diverse surfaces. Or to say it more bluntly: It can better cope with slipping and get on with it.
How much can the structure do for you?
“The reason my paper got selected to appear in PNAS is that up until this point, when you look at dynamic robots, it is rare that the engineers designing it have put much thought into what to do when the robot slips. How to cope? Even with the most advanced robots, they often have relatively simple point-contact feet. That´s because most roboticists focus on the computational power of a system – the software. Now however there is a strong trend for more and more scientists to focus on the hardware in trying to optimize walking, jumping, gliding or flying robots.”
Woodward believes that you can make robots better by adding morphological intelligence into a physical system. This is a new and growing field which is gaining interest by both biologists and roboticists. “Researchers like myself are trying to encourage the scientific community to look at the structure of a system and ask how much the structure can do for you. I see the field of robotics drifting away from the question `can we move´ to `how to best do this particular motion´.”
Woodward is among a new generation of pioneer scientists. Like the few others in this field, he wants to be able to solve the complex locomotion problems seen in even the most advanced robots by studying the strategies employed by animals when interacting dynamically with the natural environment, perfected through the course of evolution. Woodward is researching at the intersection of biology, robotics and mechanical engineering. “I look at the animal to understand how it actually moves. Then I try to extract design principles and apply that knowledge to robots. However, I can´t test everything I want on the animal because I can´t change it, so I instead test it on the robot. This allows me to discover new things about the animal and its locomotion, and to begin to answer the question of why evolution chose particular designs. It is a learning cycle where the animal teaches me about the robot and the robot teaches me about the animal. In the end I gain a deeper understanding of both.”
In his study, Woodward looked closely at the hindfoot of the locust: Its foot is made up of three segments, attached to each of which are wet adhesive pads. On the lower leg (the shank in humans, just above the ankle), there are also four spines, which look like claws (in figure 2 only three are visible). “Having both adhesive pads and spines gives the locust the possibility to interact well with both rough and smooth surfaces, including artificial surfaces like glass”, Woodward explains. He discovered in his hundreds of tests with both female and male locusts more than just that slipping was common. And he observed that the spines on their feet in fact do make contact with the surface when the locust jumps, enhancing grip on rough surfaces. Plus, the spines are not fixed, they move. There is in fact a passive joint that allows the spines – if necessary – to find better grips on the surface. Woodward also found that on occasion the locust will initiate its jump without its feet in contact with the surface. “This was really exciting because it meant that the locust’s feet might also be designed for gripping while in motion, making the design principles applicable to other forms of locomotion, such as walking and running”. That way a locust can jump whether or not its feet begin in contact with the surface and if a slip occurs, its adhesive pads and spines are designed to have a significant chance to regrip the surface. “I found that locusts slip a lot when jumping. But at the same time, it is very good in reattaching, reducing the slips´ impact on the performance: We couldn´t see a significant difference in the energy of a no-slip case to a slip-and-reattach-case. That means they are potentially losing only a very small amount of energy when slipping and reattaching – it´s not counterproductive to their ability to escape a predator.”
Building a system expecting to slip
Woodward found that the spines interact well with rough surfaces. Their sharp pointed needle-like structure allows the locust to stop it from slipping by clawing into the rough surface. “What we discovered is that the spines deal with rough surfaces and the adhesive pad with smooth surfaces like window glass. Also, we discovered when building the robot, that if we change some of the characteristics – which we can´t do with the locust – the two attachment mechanisms adhesive pad and spines help each other.” Woodward for instance took off the adhesive pad and increased the jumping power. On the sandstone surface he used, he observed big scratches on the stone, where the spines were ripping through the material. “If you add the pads you actually add sufficient friction to lower the force on the spines to a level where they can hold on to the weak sandstone surface. So not only are the adhesive pad and spines designed for different surfaces, they help each other out.”
Woodward continued changing characteristics in his robot. He came to the conclusion that a locust slipping is not an anomaly, it´s normal in locomotion. “People when walking or running want to avoid slipping. There´s a whole industry of shoe makers trying to avoid it. With a locust however, nature intended it to slip.” Hence the scientist worked on building a system expecting to slip. Because if slipping is part of the natural motion sequence and these properties are built into the structure of robot feet, then, should it ever come to a slip, the effects on its locomotion are limited.
The full scientific paper can be found here: http://www.pnas.org/content/early/2018/08/21/1804239115
Dr. Matthew Woodward is a postdoctoral researcher in the Physical Intelligence Department at the Max Planck Institute for Intelligent Systems, based in Stuttgart.
Woodward received his BSc degree in mechanical engineering from San Diego State University in San Diego in 2008 and his MSc and PhD degrees in mechanical engineering from Carnegie Mellon University in Pittsburgh in 2010 and 2017, respectively. In 2013 he was a Part-Time Professor in Mechanical Engineering at Robert Morris University in Pittsburgh. Since 2014 he has been a research fellow at the Max Planck Institute for Intelligent Systems in Stuttgart.
Woodward aims to enhance robot mobility through an understanding of how the strategies animals use for combining different locomotion modes (jump, fly, swim, etc.) and encoding morphological intelligence (passive behaviors) into their own bodies, contribute to their ability to move robustly through natural (unstructured) environments. Dynamic locomotion poses and even greater challenge as the time scales can become small enough that there is not sufficient time for “brain” control, and instead only “body” intelligence can be used to control certain aspects of the animal or robot. These concepts, coupled with advance actuation techniques, using smart materials, form the foundation for more advanced robotic platforms capable of search-and-rescue, space and planetary exploration, environmental monitoring, and healthcare.
Dr. Metin Sitti is the Director of the Physical Intelligence Department at the Max Planck Institute for Intelligent Systems, based in Stuttgart, and a Professor in the Department of Mechanical Engineering and Robotics Institute at Carnegie Mellon University in Pittsburgh.
Sitti received his BSc and MSc degrees in electrical and electronics engineering from Boğaziçi University in Istanbul in 1992 and 1994, respectively, and his PhD degree in electrical engineering from the University of Tokyo in 1999. He was a research scientist at UC Berkeley during 1999 and 2002. Since then, he has been a Professor in the Department of Mechanical Engineering and Robotics Institute at Carnegie Mellon University in Pittsburgh. In April 2014 he became one of seven Directors at the Max Planck Institute for Intelligent Systems.
Sitti and his team aim to understand the principles of design, locomotion, perception, learning, and control of small-scale mobile robots made of smart and soft materials. Intelligence of such robots mainly come from their physical design, material, adaptation, and self-organization more than to their computational intelligence. Such physical intelligence methods are essential for small-scale milli- and micro-robots especially due to their inherently limited on-board computation, actuation, powering, perception, and control capabilities. Sitti envisions his novel small-scale robotic systems to be applied in healthcare, bioengineering, manufacturing, or environmental monitoring to name a few.
At the Max Planck Institute for Intelligent Systems we aim to understand the principles of Perception, Action and Learning in Intelligent Systems.
The Max-Planck-Institute for Intelligent Systems is located in two cities: Stuttgart and Tübingen. Research at the Stuttgart site of the Max Planck Institute for Intelligent Systems covers small-scale robotics, self-organization, haptic perception, bio-inspired systems, medical robotics, and physical intelligence. The Tübingen site of the institute concentrates on machine learning, computer vision, robotics, control, and the theory of intelligence.
The MPI-IS is one of 84 Max Planck Institutes and facilities that make up the Max Planck Society, Germany’s most successful research organization. Since its establishment in 1948, no fewer than 18 Nobel laureates have emerged from the ranks of its scientists, putting it on a par with the best and most prestigious research institutions worldwide. All Institutes conduct basic research in the service of the general public in the natural sciences, life sciences, social sciences, and the humanities. Max Planck Institutes focus on research fields that are particularly innovative, or that are especially demanding in terms of funding or time requirements. And their research spectrum is continually evolving: new institutes are established to find answers to seminal, forward-looking scientific questions, while others are closed when, for example, their research field has been widely established at universities. This continuous renewal preserves the scope the Max Planck Society needs to react quickly to pioneering scientific developments.