Robotics News and Headlines

SMART and NUS Pioneer Neural Blueprint for Human-Like Intelligence in Soft Robots

SMART, NUS and MIT researchers unveiled a neuron‑inspired AI controller that lets soft robotic arms learn a full suite of motions once and instantly adapt to new conditions without retraining. The system combines offline‑trained structural synapses with online‑updating plastic synapses and a built‑in stability metric, achieving task generalisation, disturbance resilience and safe adaptation in a single framework. Validation on cable‑driven and shape‑memory‑alloy arms showed 44‑55% lower tracking error, over 92% shape accuracy under payload shifts and continued stability even with half the actuators failed. The breakthrough paves the way for real‑world soft‑robot deployments in healthcare, manufacturing and assistive technologies.