Technology

New application of Intel neural mimicry: the gospel of wheelchair

ileod 7 months ago 3 min read

Intel and Accenture, the world’s largest listed consulting company, jointly announced that they would jointly support a project of Intel neural mimicry research community (INRC), led by the neuro biological mimicry Engineering Laboratory of Open University of Israel, in collaboration with Alyn hospital, the most advanced youth rehabilitation center in Israel. with the financial and technical support of Accenture, Intel’s neural mimicry technology and applied brain research (ABR) algorithm, the Israeli research team will develop a wheelchair mounted robotic arm to help patients with spinal injury complete their daily activities, and will conduct clinical evaluation and testing for pediatric patients in arlin hospital. at present, more than 75 million people in the world need to use wheelchairs to travel. However, even the most basic movements, such as drinking water and eating, are extremely difficult for people with neuromuscular or spinal cord injury leading to upper extremity disability. One of the most important challenges has always been how to take and manipulate objects accurately. auxiliary robot technology has always been a key research direction. After installing the robot arm on the wheelchair, the user’s sense of independence can be enhanced, and the necessary human care time can be saved up to 41%. However, the components that can make the robot arm adapt to the user and the surrounding environment in real time are very expensive, and ordinary people can’t expect it. Intel neural mimicry research chip “Loihi” has real-time learning ability, can perform adaptive control, enhance the functionality of the robot arm, and use cost-effective components, which can reduce the cost by more than 10 times. in addition, the Loihi’s energy efficiency is up to 1000 times that of general-purpose chips, so the device can avoid frequent charging and is more suitable for daily use. at present, researchers from Israel’s Open University and arlin hospital have developed a robot arm for product testing. In the next step, a neural network model of the control arm will be established. Based on the “adaptive control hierarchical algorithm driven by repeated error correction” (reach) developed by ABR company, and combined with neural mimicry calculation, simple machines can be designed The human arm moves along complex paths, such as handwritten words and numbers, with a lower error rate. After the completion of the algorithm work, the research team will deploy the new model to Intel’s neural mimicry hardware, and test the function of the robot arm. After improvement, it will enter the clinical test and evaluation stage, serving patients who need to use electric wheelchairs and patients with upper limb sports injury in arlin hospital. if the project is successful, the research team plans to explore the production of auxiliary robot arms and the application of adaptive control technology in flexible manufacturing and industrial automation. Intel neural mimicry research chip Loihi has made major breakthroughs. In March this year, it has “smell”, which can learn and identify dangerous chemicals in the case of obvious noise and cover, and only a single sample can identify a smell. subsequently, Intel also completed the latest neural mimicry research system “Pohoiki springs”, which has the largest scale and the strongest performance so far, which can provide the computing power of 100 million neurons, equivalent to the size of a small mammalian brain.