SoftZoo: A bio-inspired platform for soft robot co-design and optimization

Soft robotics has been a rapidly evolving field since the term was first coined in 2008. Engineers have been developing a wide range of flexible machines inspired by the way animals move in the wild, suitable for various applications such as exploration, rehabilitation, locomotion, and even space exploration.

A team of MIT researchers has developed SoftZoo, a co-design platform that allows engineers to study soft robotics in a bio-inspired setting. The framework consists of algorithms for design, which determines the robot’s appearance, and control, which facilitates robotic motion. SoftZoo can optimize these algorithms to generate outlines for potential robots automatically.

SoftZoo features 3D models of animals, such as fishes, sharks, panda bears, and caterpillars, that can simulate soft robotics tasks, such as agile turning, path following, and locomotion, in different environments such as snow, clay, desert, or water. This platform helps demonstrate the performance trade-offs of different designs in various terrains.

Unlike other similar platforms that simulate design and control, SoftZoo is more comprehensive because it models movements that respond to physical characteristics in different biomes. The platform’s versatility stems from its differentiable multiphysics engine, which simulates multiple aspects of a physical system simultaneously, such as a baby seal turning on ice or a caterpillar moving through a wetland.

The differentiability of the engine optimizes co-design by reducing the number of expensive simulations required to solve computational control and design problems. Consequently, users can design and move soft robots with more sophisticated and specific algorithms. SoftZoo helps users determine the best robot configuration for a given shape, enabling them to develop algorithms that can perform a wide range of tasks and interact with their environment effectively.

SoftZoo showcases the importance of morphology, a branch of biology that studies the forms, shapes, and sizes of organisms. Depending on the environment, some biological structures are more optimal than others. This concept applies to machines that perform similar tasks, with blueprints tailored to suit the environment.

Observing biological structures inspires specialized artificial life that’s suitable for specific terrains. For instance, a jellyfish’s undulating geometry allows it to move efficiently across large bodies of water, inspiring researchers to develop soft robots that mimic this motion. Similarly, dragonflies’ wings have specialized structures that allow them to perform agile maneuvers, making them well-suited for working through their surroundings. SoftZoo’s platform optimizes locomotion, resulting in robots that navigate their environments efficiently.

In the past, robots struggled to move through cluttered environments because their bodies didn’t conform to their surroundings. However, SoftZoo allows designers to co-optimize the robot’s body and brain, leading to more specialized terrestrial and aquatic machines that are aware of their surroundings.

With increased behavioral and morphological intelligence, robots could conduct rescue missions and explore environments more effectively. For instance, if someone went missing during a flood, robots optimized using SoftZoo’s methods could traverse the waters more efficiently, increasing the chances of success.

SoftZoo is an open-source simulation platform that enables soft robot designers to build more flexible and efficient robots that can navigate diverse environments with ease. By optimizing both the design and control aspects of soft robots, SoftZoo offers a comprehensive computational approach to co-designing soft robots. SoftZoo models the movement of robots that react to different physical features of biomes, allowing designers to create more specialized, terrain-specific artificial life. This can lead to the development of customized machines that are designed for specific tasks and can be more useful in completing rescue missions and conducting exploration.

SoftZoo can be a substitute for field testing unnatural scenes, which saves researchers time and resources. However, the limitations of current fabrication techniques are a challenge in bringing soft robot designs to life. The platform’s designers are eyeing applications in human mechanics, such as manipulation, given its ability to test robotic control. By simulating more human-like tasks, designers can use SoftZoo to assess soft robotic arms that grasp, move, and stack objects. Nonetheless, researchers are working to overcome challenges such as muscle models, spatially varying stiffness, and sensorization to transfer simulations to physical robots.

Source: Massachusetts Institute of Technology

Leave a Comment