Researchers at the Massachusetts Institute of Technology (MIT) have developed a soft robot co-design platform to test optimal shapes for robotic performance in diverse environments.
Since the term “soft robotics” was coined, engineers in this field have been developing various types of flexible machines that can be used in exploration, locomotion, rehabilitation, and even space applications. These machines have been inspired by the movement of animals in the wild.
Researchers at MIT have developed a soft robot co-design platform that can test optimal shapes and sizes for robotic performance in different environments. SoftZoo is a bio-inspired platform that allows engineers to co-design soft robots. The framework optimizes algorithms for design and control, improving how users generate outlines for potential machines. The platform showcases 3D animal models like pandas, fishes, sharks, and caterpillars that simulate soft robotics tasks such as locomotion, agile turning, and path following in varied terrains, highlighting performance trade-offs. The researchers say their framework assists users in designing soft robotics algorithms by finding the best shape configuration for the robot, enabling efficient interaction with the environment.
SoftZoo surpasses similar platforms by modeling biome-specific movement that responds to physical features. Its versatility stems from a differentiable multiphysics engine that simultaneously simulates multiple aspects of a physical system, like a baby seal turning on ice or a caterpillar in a wetland. The engine’s differentiability cuts costs by reducing the number of simulations needed for computational control and design problems, optimizing co-design. This leads to more sophisticated algorithms for designing and moving soft robots. The system’s terrain interaction simulation highlights the significance of morphology, the biology branch studying organisms’ shapes, sizes, and forms. Optimal biological structures vary per environment, similar to comparing blueprints for machines with similar functions.
SoftZoo allows the co-optimization of terrestrial and aquatic machines, enhancing awareness and specialization by simultaneously developing the robot’s brain and body. More behavioral and morphological intelligence would make robots more valuable in rescue missions and exploration. SoftZoo’s optimization methods could enable a robot to traverse waters more efficiently in a flood search. SoftZoo’s open-source simulation aids soft robot designers in building flexible real-world robots while accelerating their movement in various environments. The computational co-design of soft robot bodies and their controllers allows for rapidly creating customized machines for specific tasks.
The researchers believe the SoftZoo can be used for human mechanics applications, like manipulation, due to its control testing capabilities. The team demonstrated this by designing a 3D arm throwing a snowball. By simulating more human-like tasks, soft robotics designers could evaluate arms that grasp, move, and stack objects.
Reference : SoftZoo: A Soft Robot Co-design Benchmark For Locomotion In Diverse Environments: openreview.net/forum?id=Xyme9p1rpZw