Inspired by the operation and structure of the brain, engineers at NUI Galway and the University of Ulster are developing bio-inspired integrated circuit technology which mimics the neuron structure and operation of the brain. Dr. Fearghal Morgan, Dr. Jim Harkin and Dr. Liam McDaid have used the natural architecture of the brain to create an electronic system that emulates some of the workings of a neuron.
Dr. Fearghal Morgan who is Director of the Bio-Inspired Electronics and Reconfigurable Computing (BIRC) research group, at NUI Galway says, “What we are trying to do is replicate a small brain-like structure in electronics. It is bio-inspired and is modelled on the structure of the behaviour of the brain.
“We are trying to replicate the structure of the brain in silicon. But only as a small device which will only be a fraction of the size of the brain.”
The device which was developed under the EMBRACE (EMulating Biologically-inspiRed ArChitectures in hardwarE) project operates in a similar way to the signal traffic of the neurons in the brain and how they are connected. According to Fearghal, the aim is that, “we will have a way of processing data that is different from the the typical micro-processor.”
Normally this would consist of reading instructions and taking data from several sources, some of them from the input to the device. It then manipulates that data and sends instructions to the output of the device: “It is instruction based processing and mostly sequential.” (In many systems processors can be replicated and layered into a multi-core system so there is an element of concurrence.)
“The nature of our device is that it is inherently concurrent. Hopefully, [we’ll have] thousands of neurons eventually with tens of thousands of connections which will give us a brain-like function.
“It will not be anywhere like as powerful as the brain but hopefully it will be low power compared to other computer systems.”
Electronic neurons, implemented using silicon integrated circuit technology, cannot replicate the complexity of the human brain which has 100 billion neurons and 1,000 trillion neuron connections. But the advantage of a bio-inspired processor is in the promise of much reduced power consumption in comparison to a traditional processing device. This allows for more opportunities to embed processors in a variety of locations.
“If you could open up this particular chip you would see electronic components that are connected together but physically they don’t look like neurons and the connections don’t look like synaptic connections between brain cells but the architecture is similar. I wouldn’t compare the structure of the electronics to a physical brain.”
At present the embrace chip is able to control a robot, “It can read signals from the environment. It has ultra-sonic sensors and it moves through a particular environment as quickly as possible without crashing.”
Another way that the bio-inspired processor differs from other conventional devices is that It is capable of learning: “It starts off as a bunch of neurons and we place it in a robotic environment and we allow the robot to move and it feels.” There are particular neuron configurations, “Each neuron has a connection to another neuron. Those connections may be weighted and when little pulses or spikes are passed from one neuron to another the neuron fires. They affect what is called the neuro-potential of the neuron they are connected to. Eventually when enough spikes come into a particular neuron and it reaches a threshold specific to that neuron that neuron will in turn fire.”
“So you have all these little neurons firing at different times depending on the pulses that are coming in that represent the world they are looking at.” The output of all this activity then goes to transducers which in turn control the robot.
To train a brain involves, “many thousands of attempts at possible configurations… and eventually you evolve the next generation of solutions. Slowly but surely, over hours and hours, you train a particular neuron connectivity between certain neurons and the threshold at which each of these neurons spike.”
Like a human baby this brain is constantly developing abilities, “and some of it is by trial and some of it is by error.
“There is so much more within the brain that we electronic engineers don’t understand. We need to work with neuro-physicists to understand the multiple layers of complexity.
“Self-awareness in robots is not something that we are anything close to. We are trying to put very specific functions that we would want to put into a neuron-like piece of silica. Hopefully we are moving toward more powerful neuron based systems.”