Alum Collaborates with Mason Scientists on Robotic Brain
Posted: May 17, 2007 at 1:00 am, Last Updated: November 30, -0001 at 12:00 am
By Karen Loss
“The brain is embodied, and the body is embedded in the environment.” That, explains alumnus Jeff Krichmar, now a senior fellow in theoretical neurobiology at the Neurosciences Institute in San Diego, is his research team’s catch phrase.
In 1997, Krichmar completed his PhD studies in computational sciences and informatics at Mason where he worked closely with Giorgio Ascoli and under the tutelage of Krasnow Institute director James Olds, among others. Now, Olds, Ascoli, and others at the Krasnow Institute are collaborating with Krichmar and his team at the Neurosciences Institute to study brain function by using robotic devices controlled by simulated nervous systems.
“If you understand the rules by which something is made, not just the physical act of putting it together, but actually how and why it all works together, you can understand more fully the makeup of systems. This underlies the brain-based systems research,” says Olds.
The researchers’ brain-based devices (BBDs) learn about features in their environment through experience, in much the same way as living creatures do. Higher brain functions depend on the cooperative activity of an entire nervous system, reflecting its morphology, dynamics, and interaction with the phenotype and the environment.
“If we are to make a model of how the brain works, we really need to put the model on something that has a body, that can move about the environment and actively sense from the environment what it sees, what it hears, what it smells and so on,” Krichmar says.
According to Krichmar, the BBDs start off naïve, although they do have a few reflexes to keep them from crashing into walls or falling off ledges. From then on, they have to learn from the environment what they need to know so they are able to navigate their world.
“These devices also provide the groundwork for developing intelligent machines that follow neurobiological rather than computational principles in their construction,” Krichmar explains. “Typically, computational models of brain function are run on computers that are severed from the real-world. Real brains, however, don’t work in a vacuum. I strongly believe that brains have to be attached to a physical body plan that allows interaction with the environment.”
With the BBD methodology, the researchers have learned much about how animals perceive the world, how they build and recall memories, and how they learn new motor skills.
“Maybe some of these things we’re finding out will give us a better understanding about diseases that affect motor or memory problems and just in general give us an understanding of how the brain works and how it gives rise to behavior,” notes Krichmar.
“One of the lessons I learned at Krasnow, especially working with Giorgio Ascoli, was the importance of the brain’s structure or anatomy,” says Krichmar. “So we make these models and try to mimic the different anatomical pathways within the nervous system.”
The models are built on a very large scale. The largest model to date has about 100,000 neurons and about 2 million synaptic connections between those neurons.
“Of course, that’s far smaller than a real brain, but it’s still a very complex system,” says Krichmar. The researchers are able to record the actions from every single neuron and synapse in the BBD’s brain while the device is carrying out a behavior, which, Krichmar explains, is just not possible with real brains.
One important aspect of Krichmar’s research is the network analysis tools his team has created to record and analyze the actions that occur between every neuron and the synaptic events that take place during the BBD experiments.
“When scientists do research on individual neurons in real brains, they do so with the use of electrodes attached to the neurons. They may do this with 20 different neurons and sometimes as many as 100, but that is pushing the limits as to how many they can actually observe at any one time,” says Krichmar. So for now, it is not possible to do the kind of research Krichmar and his team are doing on BBDs with living subjects.
“Our hope is that one day we will actually be able to collect this amount of data from real brains, and researchers will be able to use some of the tools that we’ve developed to understand our artificial brains,” Krichmar says.
“Without a doubt the most sophisticated behavior seen in biological or artificial agents is demonstrated by organisms whose behavior is guided by a nervous system,” Krichmar says. “By building the complete system — brain, body, and environment — our group and a few other groups in the world hope to achieve a new level of understanding of how the brain works, with the possibility of building machines that have the type of intelligence we associate with animals.”
This article originally appeared in a slightly different form in the spring 2007 issue of the Krasnow Bulletin.