(Nanowerk News) Since the term “soft robot” was adopted in 2008, engineers in the field have built various representations of flexible machines useful in exploration, propulsion, rehabilitation and even space. One source of inspiration: the way animals move in the wild.
A team of MIT researchers has taken it a step further, evolving SoftZoo, a bio-inspired platform that enables engineers to study design alongside soft robots. The framework optimizes the algorithm that comprises the design, which determines what the robot will look like; and controls, or systems that enable robotic movement, improving how users automatically generate outlines for potential machines.
Taking a stroll on the wild side, the platform features 3-D models of animals such as panda bears, fish, sharks and caterpillars as designs that can simulate soft robotics tasks such as locomotion, agile turns and following paths in different environments. Whether with snow, desert, clay or water, this platform exhibits balanced performance of various designs on different terrains.
“Our framework can help users find the best configuration for the robot’s shape, enabling them to design soft robotics algorithms that can do many different things,” said MIT PhD student Tsun-Hsuan Wang, an affiliate of the Computer Science and Artificial Intelligence Laboratory (CSAIL) which is a principal investigator on the project. “In essence, it helps us understand the best strategy for the robot to interact with its environment.”
SoftZoo is more comprehensive than similar platforms, which already simulate design and control, because it models movement that reacts to the physical features of different biomes. The framework’s versatility stems from its differentiable multiphysics engine, which allows simulating several aspects of a physical system at the same time, such as baby seals spinning across ice or caterpillars shuffling across a wetland environment. Machine differentiability optimizes co-design by reducing the number of often expensive simulations required to solve computational and design control problems. As a result, users can design and drive soft robots with more sophisticated custom algorithms.
The system’s ability to simulate interactions with different terrains illustrates the importance of morphology, the branch of biology that studies the shape, size, and shape of different organisms. Depending on the environment, some biological structures are more optimal than others, such as comparing blueprints for machines that accomplish similar tasks.
These biological outlines could inspire more specialized and terrain-specific artificial life. “The jellyfish’s gently undulating geometry allows it to travel efficiently across vast bodies of water, inspiring researchers to develop a new generation of soft robots and opening up limitless possibilities of the capabilities of artificial beings cultivated entirely in silico,” said Wang. “In addition, dragonflies can perform very agile maneuvers that cannot be completed by other flying creatures because they have a special structure on their wings that changes their center of mass during flight. Our platform optimizes movement in the same way that dragonflies are naturally more adept at working in their environment.”
Robots previously struggled to navigate through messy environments because their bodies didn’t match theirs. However, with SoftZoo, designers can simultaneously develop a robot’s brain and body, optimizing land and water machines simultaneously to be more alert and specialized. With increased behavioral intelligence and morphology, robots will be more useful in completing rescue missions and conducting exploration. If a person is lost during a flood, for example, the robot has the potential to traverse waters more efficiently because it is optimized using methods demonstrated on the SotftZoo platform.
“SoftZoo provides open-source simulation for soft robot designers, helping them build real-world robots much more easily and flexibly while accelerating machine locomotion in diverse environments,” added study co-author Chuang Gan, a research scientist at MIT-IBM Watson AI. Lab who will soon become an assistant professor at the University of Massachusetts at Amherst.
“The computational approach to co-designing soft robot bodies and their brains (i.e., their controllers) opens the door to rapidly creating specialized machines designed for specific tasks,” adds Daniela Rus, director of CSAIL and Andrew and Professor Erna Viterbi in the Department of Electrical Engineering and MIT Computer Science (EECS), who is the other author of the work.
Before all kinds of robots were built, these frames could serve as a substitute for field testing of unnatural sights. For example, assessing how a bear-like robot behaves in the desert might be challenging for a research team working in the urban plains of Boston. In contrast, soft robotics engineers can use 3-D models in SoftZoo to simulate different designs and evaluate how effectively the algorithms controlling their robot are in navigation. In turn, this will save researchers time and resources.
However, limitations of current fabrication techniques prevent bringing these soft robot designs to life. “Transferring from simulation to physical robots is still unsolved and requires further study,” said Wang. “Muscle modeling, spatially varying stiffness, and sensorization in SoftZoo cannot be directly realized with current fabrication techniques, so we are working on this challenge.”
In the future, platform designers are eyeing applications in human mechanics, such as manipulation, given their ability to test robot controls. To demonstrate this potential, Wang’s team designed a 3-D arm that throws a snowball forward. By incorporating more human-like task simulations, soft robot designers can then use the platform to assess soft robotic arms that capture, move and stack objects.