worm-sim

The tiny worm C. elegans is one of the simplest multicellular organisms, and one of the only organisms in the world to have its connectome (neuron map) completed.

Despite being so simple, C. elegans exhibits several complex behaviors. They will attempt to move towards food when they sense it, and move around obstacles when they make contact. After viewing a video of a connectome being used to simulate a worm brain and control a GoPiGo robot, I wanted to see if something similar could be done in the browser.

In this demo, hunger neurons are continuously stimulated. If the worm nears food, food sensing neurons are stimulated. If the worm reaches the edge of the window, nose touch sensory neurons are stimulated.

Usage

Head on over to heyseth.github.io/worm-sim/ and watch the worm wriggle around. Click to place down food. The green dots at the top are a visual representation of the connectome, with each dot representing a neuron (increased opacity = increased activity). There are buttons to hide the connectome, reset the worm's position, and to clear placed down food.

To reiterate, the worm you see on your screen is being controlled entirely by a simulated virtual worm brain. Very cool if you ask me!

How Movement Works

The worm is modeled as a chain of linked body nodes. Each frame, the simulated connectome updates its neuron activity and converts motor neuron output into 17 body segment activation groups. Each group tracks dorsal, ventral, left, right, and turn activation values.

Dorsal and ventral muscle activity create the main wriggling motion. For each body segment, the physics code compares ventral and dorsal activation to choose a target bend, then applies forces to neighboring nodes so the body bends toward that shape. A traveling central pattern generator wave stimulates alternating dorsal and ventral muscles from head to tail, which produces the familiar worm-like undulation.

Forward movement comes from anisotropic friction. Each node has a local forward direction based on the body segment it belongs to. Velocity along that direction is damped lightly, while sideways velocity is damped much more strongly. This lets the bend wave push the worm forward instead of simply sliding sideways.

Turning is handled separately from raw left/right muscle totals. The connectome can produce asymmetric muscle activity as part of normal movement, so the simulator uses a bounded turn signal for steering. That signal is strongest near the head and fades down the first few body segments, allowing the worm to wander without a persistent one-direction spiral.

License

This project is licensed under the MIT License - see the license.md file for details.

Acknowledgments

Huge thanks to:

• Timothy Busbice, Gabriel Garrett, Geoffrey Churchill, and all other contributors to the GoPiGo Connectome.
• Zach Rispoli, for porting the connectome to JavaScript.
• ARTsinn, for their awesome canvas worm demo, which served as inspiration for this project.