Particle Life

Try the demo: https://sam-martis.github.io/Particle_Life/

Reload the page to get new interaction rules! New combinations every refresh!

This project has turned out better than I expected.

At it's core, this is an attempt to simulate 'life'. Very basic and rudimentary life, but still life.

I find it to be a better representation of how protiens and enzymes work. Often times in the simulation, you'll see recombination happening. Sometimes adding a single particle to a collection/group will completely reorient and reshape it.

Also it's cool to look at, especially how complex structures form from simple rules of interaction.

How it works

Core idea:

Unsymmetric forces. That's what drives the entire simulation. Red could attract Green but Green could repel Red. If the stength of attraction/repulsion is similar, they could form a pair that makes movement, with one constantly trying to get away while the other gets tugged along!

We can represent the coefficient of attractions as a matrix $A$ of size $n$ x $n$ The ${A_{ij}}^{th}$ entry represents the co-efficient of attraction attraction of the $i^{th}$ particle type to the $j^{th}$ particle type.

For the above simulation, the number of types of particles(ie: $n$) is 6 (can be easily increased but I didn't feel the need to do it)

The interaction matrix A is assigned random values at the beggining of each session. Hence, Whenever you refresh the page, you get new rules and new possibilities for structures to form!

More details:

Quad trees

The core computational engine is powered by k-d trees, more specifically a Quad-tree in my case. k-d trees are spatial data-structures, as they encode information about the location of elements within their structure.

This makes calculating forces more efficient as we can only query nearby particles of and ignore far-away particles.

Also, collision detction is GREATLY improved as we can just query the immediate viscinity of the particle in question.

Check out my Quad tree implementation for more details! https://github.com/Sam-MARTis/QuadTrees

Toroidal Boundry

A toroidal boundry is type of boundry condition where the entire map can be mapped onto a toroid, with the edges meeting. Basically, if you go far left, you emerge out of the right. Not as simple to setup as I expected cause I also had to change the region to query the Quad tree at.

RK2 Method

A Runge-Kutta method of order 2 is used (Also known as Heun's method) to update the velocity given the forces.

More pictures