Researchers from the Singapore-MIT Alliance for Research and Technology (SMART), NUS, MIT, and NTU Singapore have developed a brain-inspired AI control system for soft robotic arms that achieves three key capabilities simultaneously: transferring learned skills to new tasks, real-time adaptation to changing conditions, and guaranteed stability. The system uses two types of synapses — structural synapses trained offline on foundational movements, and plastic synapses that update online during operation. Tested on two physical platforms, it achieved 44–55% reduction in tracking error under heavy disturbances and over 92% shape accuracy despite payload changes, airflow disturbances, and actuator failures. The work, published in Science Advances, has potential applications in assistive robotics, rehabilitation, medical devices, and industrial automation.
