Brain Based Devices - Artificial bodies with artificial nervous systems
NSI home Brain-Based Devices Publications

Darwin VII

NOMAD front view
Unlike a robot, Darwin VII is an autonomous "being" which starts with a naive brain and learns from its own experience, connected to a realistic model of the mammalian brain that is simulated on a multiprocessor computer

Darwin VII is a Brain-Based Device that is used as a tool to help study the brain. Darwin VII starts out naive and learns from its own experience. It is connected to a realistic model of a brain, which is simulated on a network of powerful computers. All of Darwin VII's behaviors are controlled by the activity of its brain cells, allowing researchers a unique window into how the human brain may work. Darwin VII can interact with its environment in several ways: by sensing light, sound and taste, and by moving around and grabbing objects.

As Darwin VII roams around the lab, the images it captures through a video camera are sent through a wireless connection to its brain running inside a computer. The brain tries to make sense of the image through the activity of its neurons. Then the brain triggers a motor command, which is relayed to Darwin VII's body.

When the researchers activate Darwin VII, it begins a "run" with a naive brain. It has an inborn preference for light and a predisposition for a certain taste, but no other experience or programming.

As it explores its environment, Darwin VII will encounter a small black block with a blob or stripe on the top. Since it is attracted to light, it will steer toward the block and pick it up. When it grabs the striped block with its metal gripper, it gets a charge of current. In the simulated brain, this conductivity is registered as good taste. Blocks with blobs, on the other hand, do not give a charge and correspond to bad taste. As Darwin VII's gripper holds the block, the brain simulation associates the taste of the block with the pattern that it sees. Through the brain simulation, Darwin VII not only learns what each block looks like, but which blocks taste good and which taste bad. After learning these things, it will no longer pick up the bad tasting blocks. It will approach the blocks, and after seeing the pattern of the blocks, Darwin VII will now remember the associated taste. If Darwin VII thinks it tastes good it will pick up the block and if it thinks it tastes bad it will avoid the block.

To date, The Institute's work has yielded revealing insights, e.g. into the importance of self-generated movement in the aquisition of perceptual capability and the almost universal role of value systems in adaptaion and learning.

The Institute's researchers believe that it is impossible to study the brain or a part of it in isolation. Feeding static data to the brain simulation running on a computer does not capture the complexity of the processing that the brain is required to do. Using Darwin VII we can perform tests with all of the noise and sophistication of the real world and still record the activity from the entire brain, something that is extremely difficult to do with live animals.

What Darwin VII "does"

  • navigates through its environment
  • senses obstacles and avoids bumping into them
  • approaches objects after "seeing" them from a distance
  • grabs objects with its gripper
  • tastes objects
  • rejects or avoids objects that it has learned will taste bad

Darwin VII Photos

Click on any photo for a larger image. All photos on this web page are used with the permission of Newsmakers online photo service.

Jim Snook and NOMAD
Chief engineer James Snook performs "brain surgery"

Jeff Krichmar, Jim Snook, and NOMAD
Neuroscientist Jeff Krichmar, Ph.D., (left) and chief engineer James Snook look over Darwin VII.



NOMAD "tasting" a block
Darwin VII picks up an object it found to see if it thinks it "tastes good" using its thinking brain as neuroscientist Jeff Krichmar, Ph.D, watches.


Darwin VII Videos

Darwin VII's Behavior and Brain Response During Learning

These movie clips show Darwin VII in action during behavioral experiments and, to the right of Darwin VII, activity of selected neural areas in the simulation (R, top left; IT, top right; A1, bottom left; Mave, bottom right, left side; Mapp, bottom right, right side). Each pixel in a selected neural area represents a neuronal unit and activity is normalized in a range from no activity (dark blue) to maximal activity (bright red). During each movie clip Darwin VII first approaches an appetitive block and then an aversive block. More details of this experiment can be found in our recent publications.
  • View Darwin VII during an early visual learning trial (mpeg, 24MB)
    Darwin VII upon early encounter with blocks having visual patterns. In this early conditioning trial, Darwin VII is shown picking up and "tasting" both the appetive and aversive block. Activity in IT is insufficient, but activity in the taste system is sufficient to drive activity in the motor behavior neural area (Mave or Mapp) above the behavioral threshold.
     
  • View Darwin VII after visual conditioning (mpeg, 19MB)
    Darwin VII upon the tenth encounter with blocks having visual patterns. After primary conditioning with visual stimuli, activity in area IT is sufficient to drive the Mave or Mapp neuronal units above the behavioral threshold triggering a motor response to taste an appetitive block and avoid "tasting" an aversive block.
     
  • View Darwin VII after auditory conditioning (mpeg, 18MB)
    Darwin VII upon the tenth encounter with blocks having only auditory cues. The stimulus blocks shown in the figure beeped, but had a pattern made up of small black and white shapes that was high contrast enough to evoke a visual tracking response, but did not have enough of a pattern to evoke a response in IT. After primary conditioning with auditory stimuli, activity in area A1 is sufficient to drive the Mapp or Mave neuronal units above the threshold to trigger a behavioral response.
     

Darwin VII In Action

This film clip shows Darwin VII in action. Darwin VII is born with certain instincts: he avoids hitting obstacles, he approaches high contrast objects, and he gets good value or taste from conductive objects. In the film clip, the blocks with striped patterns are conductive. Darwin VII learns on his own that the striped blocks are good tasting and the blob patterned blocks are bad tasting. In other words, Darwin VII has learned to associate a taste with a visual pattern. After about 10 samples of each block, Darwin VII learns to avoid tasting the blob patterned blocks.

Darwin VII In The News

This film clip shows a short news piece that gives a nice overview of the NOMAD project. It was aired by KGTV, a San Diego ABC affiliate, on May 4th, 2000. This movie clip is used with permission of KGTV.


For more information, contact Jeff Krichmar




Jump to:

Darwin VII photos

Darwin VII videos



More Information:
Brain-Based Devices Main Page
Darwin X/XI
A model of episodic, spatial, and multimodal memory formation
Segway Cerebellum
A model of predictive learning
Segway Soccer Playing BBD
Interacting with humans on the field