A Worm Brain Powering a LEGO Robot

December 19, 2014 | 02:15
A Worm Brain Powering a LEGO Robot
A Worm Brain Powering a LEGO Robot
Researchers with the OpenWorm project have uploaded the digital copy of a worm brain to a wheeled LEGObot. Without additional programming or learning the bot displays the behavior of a C elegans roundworm.

The OpenWorm project aims to create the first digital organism using the Caenorhabditis elegans roundworm as a model. The tiny 1.5 mm creature is the only life form to date who's connectome – a wiring diagram of neural connections in the brain – is entirely mapped out. A task that took scientists 12 years to complete.

In silico
Participants of the OpenWorm project construct software models based on the connectome and other scientific data gathered from decades of researching the worm. At the same time they're developing simulation platforms to run the models, all with the goal to create an in silico life form.

The research is open source with repositories available on Github and community members meeting every two weeks in Google Hangouts.

Worm in spirit
Now Timothy Busbice, one of the founders of OpenWorm, has uploaded the worm's brain map of 302 neurons and their connections as a software program to a LEGObot. The bot's sensors provide input data for the digital sensory neurons. When input data result in a certain value, the sensory neurons send a message to the motor neurons. The motor neurons, in turn, transmit a signal to control the speed and direction of the robot's motors.

The LEGObot operates without any additional programming or machine learning. Its behavior emerges after activating the connectome and resembles that of a roundworm. The front sonar sensor, for instance, emulates the worm's nose. If it senses an obstacle ahead it changes direction to avoid collision.

Limits of technology
Busbice has been working on digitizing a connectome for over 20 years. (Before the diagram of the C elegans was completed he worked with made up neural networks “that did not yield any meaningful results”, he writes).

In his emulation of the brain each neuron is represented by an individual computer program. For years Busbice ran up against the limits of technology because processors could only run a limited number of processes simultaneously. Only when 64-bit computers became available he was able to execute his plan on his Lenovo T420.

Communication between neurons
The neuron-programs exchange information using UDP, a communication method from the Internet Protocol Suite that, unlike TCP, doesn't require establishing a connection and thus allows for asynchronous messaging. Each application is identified by a unique port number.

When the frontal sonar sensor senses an obstacle it transmits an UDP packet to the sensory neurons. Just as in the biological brain, communication between neurons has a certain value or weight. When a neuron sends a message to another neuron it has 3 connections to, the weight of the signal has value 3, if it has only 1 connection the weight is 1, etc. The weight is communicated in the UDP message. The receiving neuron can receive a number of messages and when the accumulated weight exceeds a certain number, say 10, it jolts into action and sends a message of it own to the neurons it's connected to. If the threshold is not exceeded the counter goes back to zero after 200 milliseconds.

Decades of work of many scientists and engineers have gone into reverse engineering a complex biological system and recreating it in the form of a worm-brained LEGObot. So have a look at the creature in the video below and realize that's what can be accomplished by a species with about a 100 billion neurons per capita to spend.

Image by Timothy Busbice
Loading comments...
related items