Virginia Tech Researchers Control Soft Robotic Arm, Cutting Power Use by Up to 75 Times
Updated
Updated · Virginia Tech · May 19
Virginia Tech Researchers Control Soft Robotic Arm, Cutting Power Use by Up to 75 Times
2 articles · Updated · Virginia Tech · May 19
A Virginia Tech team used reservoir computing to control a simulated soft robotic arm that can flex, twist, warp and bend—described as the first method able to handle such a highly flexible, fast-moving design.
The approach modeled an arm with an elastic core and paired synthetic muscles, then learned its motion dynamics through virtual trials instead of relying on long command sets used in conventional control systems.
When the reservoir was run on a neuromorphic chip that spikes like the brain, power consumption fell by as much as 75 times while outperforming standard AI and machine-learning methods, the researchers said.
The PNAS study points toward smaller untethered soft robots for medicine, agriculture, salvage and infrastructure inspection, with the next step building physical prototypes from the virtual control data.
This brain-inspired AI mastered a virtual robot, but can it overcome the challenges of a physical body?
How can new octopus-like robots achieve incredible flexibility while using 75 times less energy than current machines?
From Simulation to Reality: Virginia Tech’s Neuromorphic Soft Robots Set New Standards in Dexterity and Efficiency
Overview
Virginia Tech researchers, led by Noel Naughton, have made a major breakthrough in controlling simulated soft robotic arms by developing a neural reservoir system. This system refines the movement dynamics of soft robots by using data from virtual trials, where the team set expected outcomes and analyzed results. Building on their understanding of elastic cores and synthetic muscles, the researchers created virtual models to test how muscle pairs work together. By feeding these results back into the system, they developed a sophisticated model that promises greater dexterity and efficiency for future soft robotics.