CRCNS: Transcortical and spinal circuit contributions to hand shaping in primates - Real-time neuromorphic implementation for robotic demonstration

Engineering of robots primarily relies on prescribed algorithms for centralized control. This results in robots with limited versatility because every function must be preprogrammed. Animals, by contrast, rely on adaptable neuronal networks distributed throughout the body that convert and modulate brain signals into specific and well-coordinated muscle actions and corrections.

It has now become possible to record signals from these large neuronal networks in primates in the part of the spinal cord controlling hand function. Therefore, our goal is to extract the functional features of these neuronal networks, and validate their function by controlling bio-inspired robotic hands, as well as human cadaveric hands. This validation will allow the first physical test of the biological mechanisms for grasp function, and will help understand hand disabilities and treatments in, for example, stroke, spinal cord injury and cerebral palsy.

It will also launch a new generation of versatile robots that use the mechanisms of our nervous system.

The overall goal is to create a synthetic functional analogue of the cervical spine that controls multiple grasp modalities in bio-robotic hands. This is made possible by the advent of specialized massively parallel computer chips that allow the implementation of networks of hundreds of simulated neurons and their spiking dynamics (neuromorphic chips).

Therefore, in this project, we will extract network architectures for the control of the hand from the nervous system of primates (Japan) and implement them as neuromorphic circuits to create a new class of versatile robotic hands (USA). Using specialized recording system, will record neural data from hundreds of spinal interneurons and alpha motoneurons in the cervical spinal cord of awake, behaving monkeys during manipulation—while also recording EMG and hand kinematics.

This will be the most complete data set to date for cervical control of the hand (Aim 1). Then, we will create neuromorphic implementations of that neural circuitry using state of the art very large scale integration chips. Special attention will be paid to implementing physiologically valid versions of alpha-gamma motoneuron interactions, and realistic plasticity rules.

We will also create a Domain Specific Language that allows the translation of general neuroanatomical circuits into neuromorphic code to make this technology accessible by the general neuroscience community (Aim 2). We will test, refine and validate the neuromorphic circuits by using the neuromorphic chips to control neuro-robotic hands using electric motors programmed to behave as muscles, and sensors to replicate the function of muscle spindles and Golgi tendon organs (Aim 3).

We will also control cadaveric human hands to validate the neuromorphic controller for the anatomy of the human hand. This will pave the way to a better understanding of hand function and disability and serve as the proof of concept for a new class of neuromechanical robotic, prosthetic and brain-controlled hands.

A companion project is being funded by the National Institute of Information and Communications Technology, Japan (NICT). This project is jointly funded by the following NSF programs: Disability and Rehabilitation Engineering, Collaborative Research in Computational Neuroscience, Robust Intelligence, and Mind, Machine and Motor Nexus program.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.