Photo by Gabriele Malaspina on Unsplash
MANILA, PHILIPPINES [TAC] – A new Artificial Intelligence (AI) control system modelled after human neurons allows soft robotic arms to learn complex movements and adapt to chaos – like heavy loads or hardware failures – instantly and without retraining.
In a major leap for embodied intelligence, Singapore-based researchers have bridged the gap between biological adaptability and mechanical control, developing a neuron-inspired AI that allows soft robots to master new tasks and recover from unexpected physical disruptions in real time.
While traditional robots are rigid and predictable, soft robots are flexible and notoriously difficult to control. This breakthrough from SMART and NUS solves the problem by using a “neural blueprint”: the robot uses built-in skills for basic tasks but employs “plastic synapses” to adjust its grip or balance the moment it feels a disturbance.
Recently published in Science Advances, the study claims that it’s the first system to successfully combine universal learning with a mathematical safety guarantee, ensuring the robot stays stable even while it’s “thinking” on the fly.
Inspired by the way the human brain learns, this system is one of the first to achieve three aspects needed to deploy soft robots in real-world environments — learning capabilities that can be generalised across tasks, the ability to maintain performance under diverse disturbances, and a metric that enables stability during adaptation.
“This new AI control system is one of the first general soft-robot controllers that can achieve all three key aspects needed for soft robots to be used in society and various industries. It can apply what it learned offline across different tasks, adapt instantly to new conditions and remain stable throughout — all within one control framework,” said Associate Professor Zhiqiang Tang, a corresponding author of the paper, who was a postdoctoral associate at M3S and at NUS when he carried out the research. Tang is now an Associate Professor at the Southern University in China.
This breakthrough brings soft robotics closer to human-like adaptability for real-world applications, such as in assistive robotics, rehabilitation robots, and wearable or medical soft robots, by making them more intelligent, versatile and safe.
“Soft robots hold immense potential to take on tasks that conventional machines simply cannot, but true adoption requires control systems that are both highly capable and reliably safe. By combining structural learning with real-time adaptiveness, we’ve created a system that can handle the complexity of soft materials in unpredictable environments. It’s a step closer to a future where versatile soft robots can operate safely and intelligently alongside people — in clinics, factories, or everyday lives,” said Professor Daniela Rus, co-corresponding author of the paper and co-lead principal investigator at M3S and director at computer science and artificial intelligence laboratory, MIT.
