Scientists at Nanyang Technological University, Singapore, have developed AI-powered mini-brains, allowing robots to recognize pain and self-repair when damaged. Using AI-enabled sensor nodes, the system is capable of processing and reacting to pain emerging from pressure exerted by a physical force. When a robot has minor damage, the system gives it the ability to repair itself without human activity.
This new technique embeds AI into the network of sensor nodes. These are connected to several tiny, less-powerful processing units, behaving similarly to 'mini-brains' distributed on the robot's skin. Learning takes place in a localized area, and the robot's wiring requirements and response time are lessened by five to ten times compared to traditional robots. The system also consists of a self-healing ion gel material that lets the robot regain mechanical functions without human assistance.
To teach robots pain, the team created memtransistors, which are brain-light electronic devices that have memory and information processing capabilities, to behave like artificial pain receptors and synapses. In laboratory testing, the team showed that the robot could learn to respond to injury in real-time. Additionally, they demonstrated that the robot was still responding to pressure even when damaged.
When the robot gets a cut from a sharp object, it loses mechanical function. However, the self-healing ion gel's molecules start interacting. This causes the robot to repair itself by stitching its wound together, which allows it to regain functionality while maintaining high responsiveness.
"In this work, our team has taken an approach that is off-the-beaten-path by applying new learning materials, devices and fabrication methods for robots to mimic the human neuro-biological functions. While still at a prototype stage, our findings have laid down important frameworks for the field, pointing the way forward for researchers to tackle these challenges," said Associate Professor Nripan Mathews, a co-lead author from the School of Materials Science & Engineering at NTU.
Building on previous work involving neuromorphic electronics such as using light-activated devices to recognize objects, the team is hoping to collaborate with industry partners and government research labs to improve their system for larger-scale applications.