The Cellular Automaton Method for Cave Generation

Dear reader, this post has an interactive simulation! We encourage you to play with it as you read the article below.

In our series of posts on cellular automata, we explored Conway’s classic Game of Life and discovered some interesting patterns therein. And then in our primers on computing theory, we built up a theoretical foundation for similar kinds of machines, including a discussion of Turing machines and the various computational complexity classes surrounding them. But cellular automata served us pretty exclusively as a toy. It was a basic model of computation, which we were interested in only for its theoretical universality. One wouldn’t expect too many immediately practical (and efficient) applications of something which needs a ridiculous scale to perform basic logic. In fact, it’s amazing that there are as many as there are.

In this post we’ll look at one particular application of cellular automata to procedural level generation in games.

An example of a non-randomly generated cave level from Bethesda’s The Elder Scrolls series.

The Need for More Caves

Level design in video games is a time-consuming and difficult task. It’s extremely difficult for humans to hand-craft areas that both look natural and are simultaneously fun to play in. This is particularly true of the multitude of contemporary role-playing games modeled after Dungeons and Dragons, in which players move through a series of areas defeating enemies, collecting items, and developing their character. With a high demand for such games and so many levels in each game, it would save an unfathomable amount of money to have computers generate the levels on the fly. Perhaps more importantly, a game with randomly generated levels inherently has a much higher replay value.

The idea of randomized content generation (often called procedural generation) is not particularly new. It has been around at least since the 1980’s. Back then, computers simply didn’t have enough space to store large, complex levels in memory. To circumvent this problem, video game designers simply generated the world as the player moved through it. This opened up an infinitude of possible worlds for the user to play in, and the seminal example of this is a game called Rogue, which has since inspired series such as Diablo, Dwarf Fortress, and many many others. The techniques used to design these levels have since been refined and expanded into a toolbox of techniques which have become ubiquitous in computer graphics and game development.

We’ll explore more of these techniques in the future, but for now we’ll see how a cellular automaton can be used to procedurally generate two-dimensional cave-like maps.

A Quick Review of Cellular Automata

While the interested reader can read more about cellular automata on this blog, we will give a quick refresher here.

For our purposes here, a 2-dimensional cellular automaton is a grid of cells G, where each cell c \in G is in one of a fixed number of states, and has a pre-determined and fixed set of neighbors. Then G is updated by applying a fixed rule to each cell simultaneously, and the process is repeated until something interesting happens or boredom strikes the observer. The most common kind of cellular automaton, called a ‘Life-like automaton,’ has only two states, ‘dead’ and ‘alive’ (for us, 0 and 1), and the rule applied to each cell is given as conditions to be ‘born’ or ‘survive’ based on the number of adjacent live cells. This is often denoted “Bx/Sy” where x and y are lists of single digit numbers. Furthermore, the choice of neighborhood is the eight nearest cells (i.e., including the diagonally-adjacent ones). For instance, B3/S23 is the cellular automaton rule where a cell is born if it has three living neighbors, and it survives if it has either two or three living neighbors, and dies otherwise. Technically, these are called ‘Life-like automata,’ because they are modest generalizations of Conway’s original Game of Life. We give an example of a B3/S23 cellular automaton initialized by a finite grid of randomly populated cells below. Note that each of the black (live) cells in the resulting stationary objects satisfy the S23 part of the rule, but none of the neighboring white (dead) cells satisfy the B3 condition.

A cellular automaton should really be defined for an arbitrary graph (or more generally, an arbitrary state space). There is really nothing special about a grid other than that it’s easy to visualize. Indeed, some cellular automata are designed for hexagonal grids, others are embedded on a torus, and still others are one- or three-dimensional. Of course, nothing stops automata from existing in arbitrary dimension, or from operating with arbitrary (albeit deterministic) rules, but to avoid pedantry we won’t delve into a general definition here. It would take us into a discussion of discrete dynamical systems (of which there are many, often with interesting pictures).

It All Boils Down to a Simple Rule

Now the particular cellular automaton we will use for cave generation is simply B678/S345678, applied to a random initial grid with a fixed live border. We interpret the live cells as walls, and the dead cells as open space. This rule should intuitively work: walls will stay walls even if more cells are born nearby, but isolated or near-isolated cells will often be removed. In other words, this cellular automaton should ‘smooth out’ a grid arrangement to some extent. Here is an example animation quickly sketched up in Mathematica to witness the automaton in action:

An example cave generated via the automaton rule B678/S345678. The black cells are alive, and the white cells are dead.

As usual, the code to generate this animation (which is only a slight alteration to the code used in our post on cellular automata) is available on this blog’s Github page.

This map is already pretty great! It has a number of large open caverns, and they are connected by relatively small passageways. With a bit of imagination, it looks absolutely cavelike!

We should immediately note that there is no guarantee that the resulting regions of whitespace will be connected. We got lucky with this animation, in that there are only two disconnected components, and one is quite small. But in fact one can be left with multiple large caves which have no connecting paths.

Furthermore, we should note the automaton’s rapid convergence to a stable state. Unlike Conway’s Game of Life, in practice this automaton almost always converges within 15 steps, and this author has yet to see any oscillatory patterns. Indeed, they are unlikely to exist because the survival rate is so high, and our initial grid has an even proportion of live and dead cells. There is no overpopulation that causes cells to die off, so once a cell is born it will always survive. The only cells that do not survive are those that begin isolated. In a sense, B678/S345678 is designed to prune the sparse areas of the grid, and fill in the dense areas by patching up holes.

We should also note that the initial proportion of cells which are alive has a strong effect on the density of the resulting picture.  For the animation we displayed above, we initially chose that 45% of the cells would be live. If we increase that a mere 5%, we get a picture like the following.

A cave generated with the initial proportion of live cells equal to 0.5

As expected, there are many more disconnected caverns. Some game designers prefer a denser grid combined with heuristic methods to connect the caverns. Since our goal is just to explore the mathematical ideas, we will leave this as a parameter in our final program.

Javascript Implementation, and Greater Resolution

One important thing to note is that B678/S345678 doesn’t scale well to fine grid sizes. For instance, if we increase the grid size to 200×200, we get something resembling an awkward camouflage pattern.

A 200×200 grid cave generation. Click the image to enlarge it.

What we really want is a way to achieve the major features of the low-resolution image on a larger grid. Since cellular automata are inherently local manipulations, we should not expect any modification of B678/S345678 to do this for us. Instead, we will use B678/345678 to create a low-resolution image, increase its resolution manually, and smooth it out with — you guessed it — another cellular automaton! We’ll design this automaton specifically for the purpose of smoothing out corners.

To increase the resolution, we may simply divide the cells into four pieces. The picture doesn’t change, but the total number of cells increases fourfold. There are a few ways to do this programmatically, but the way we chose simply uses the smallest resolution possible, and simulates higher resolution by doing block computations. The interested programmer can view our Javascript program available on this blog’s Github page to see this directly (or view the page source of this post’s interactive simulator).

To design a “smoothing automaton,” we should investigate more closely what we need to improve on in the above examples. In particular, once we increase the resolution, we will have a lot of undesirable convex and concave corners. Since a “corner” is simply a block satisfying certain local properties, we can single those out to be removed by an automaton. It’s easy to see that convex corners have exactly 3 live neighbors, so we should not allow those cells to survive. Similarly, the white cell just outside a concave corner has 5 live neighbors, so we should allow that cell to be born. On the other hand, we still want the major properties of our old B678/S345678 to still apply, so we can simply add 5 to the B part and remove 3 from the S part. Lastly, for empirical reasons, we also decide to kill off cells with 4 live neighbors.

And so our final “smoothing automaton” is simply B5678/S5678.

We present this application as an interactive javascript program. Some basic instructions:

  • The “Apply B678/S345678” button does what you’d expect: it applies B678/S345678 to the currently displayed grid. It iterates the automaton 20 times in an animation.
  • The “Apply B5678/S5678” button applies the smoothing automaton, but it does so only once, allowing the user to control the degree of smoothing at the specific resolution level.
  • The “Increase Resolution” button splits each cell into four, and may be applied until the cell size is down to a single pixel.
  • The “Reset” button resets the entire application, creating a new random grid.

We used this program to generate a few interesting looking pictures by varying the order in which we pressed the various buttons (it sounds silly, but it’s an exploration!). First, a nice cave:

An example of a higher resolution cave created with our program. In order to achieve similar results, First apply B678/S345678, and then alternate increasing the resolution and applying B5678/S5678 1-3 times.

We note that this is not perfect. There are some obvious and awkward geometric artifacts lingering in this map, mostly in the form of awkwardly straight diagonal lines and awkwardly flawless circles. Perhaps one might imagine the circles are the bases of stalactites or stalagmites. But on the whole, in terms of keeping the major features of the original automaton present while smoothing out corners, this author thinks B5678/S5678 has done a phenomenal job. Further to the cellular automaton’s defense, when the local properties are applied uniformly across the entire grid, such regularities are bound to occur. That’s just another statement of the non-chaotic nature of B5678/S5678 (in stark contrast to Conway’s Game of Life).

There are various modifications one could perform (or choose not to, depending on the type of game) to make the result more accessible for the player. For instance, one could remove all regions which fit inside a sufficiently small circle, or add connections between the disconnected components at some level of resolution. This would require some sort of connected-component labeling, which is a nontrivial task; current research goes into optimizing connected-component algorithms for large-scale grids. We plan to cover such topics on this blog in the future.

Another example of a cool picture we created with this application might be considered a more “retro” style of cave.

Apply S678/B345678 once, and increase the resolution as much as possible before applying B5678/S5678 as many times as desired.

We encourage the reader to play around with the program to see what other sorts of creations one can make. As of the time of this writing, changing the initial proportion of live cells (50%) or changing the automaton rules cannot be done in the browser; it requires one to modify the source code. We may implement the ability to control these in the browser given popular demand, but (of course) it would be a wonderful exercise for the intermediate Javascript programmer.

Caves in Three Dimensions

It’s clear that this same method can be extended to a three-dimensional model for generating caverns in a game like Minecraft. While we haven’t personally experimented with three-dimensional cellular automata here on this blog, it’s far from a new idea. Once we reach graphics programming on this blog (think: distant future) we plan to revisit the topic and see what we can do.

Until then!

Advertisements

Turing Machines and Conway’s Dreams

Additional Patterns

Last time we left the reader with the assertion that Conway’s game of life does not always stabilize. Specifically, there exist patterns which result in unbounded cell population growth. Although John Conway’s original conjecture was that all patterns eventually stabilize (and offered $50 to anyone who could provide a proof or counterexample), he was proven wrong. Here we have the appropriately named glider gun, whose main body oscillates, expelling a glider once per period. Here is an initial configuration:

An initial position for the glider gun

And its animation:

This glider gun was the first one of its kind ever discovered. To distinguish it from the now large class of “gun” patterns, it is called Gosper’s glider gun. It has the smallest initial population of any known gun (hint, hint: find a smaller one and get famous!).

Second, we have examples of moving patterns which leave stationary patterns as they travel. These are commonly called puffers. For the sake of amazement, we give the coolest puffer we could find, which actually lays Gosper guns! (credit to Wikipedia for the image)

At the end of the animation, the red colored cells are the puffer, the green are Gosper guns, and the blue are the emitted gliders.

So (after the work of many in searching for these patterns), we see that under special circumstances Life can grow without bound. This has an interesting connection to computability. Specifically, any model of computation in which every computation is guaranteed to stop (in the case of cellular automata, this is reaching a stable state) cannot be Turing-complete.

[Note: the details on Turing machines are covered in this blog’s primer on the theory of computation, but the reader may recall that a system for computation which is Turing-complete can do any computation that can be done on any other Turing machine. This includes performing arithmetic, simulating Conway’s Game of Life, and performing the functions of a web browser.]

So colloquially, being able to simulate an infinite loop (or infinite recursion) is required to do interesting computations. More rigorously, if an automaton is to simulate a Turing machine, then it must be able to loop infinitely, because a Turing machine can.

But we have just found that Life can simulate infinite loops. Specifically, a Gopser gun or the puffer above both simulate an infinite counter, counting the number of emitted/laid patterns. Admittedly, we can’t do much with just an infinite counter, but it gives us the hint that we may be able to construct the elementary pieces of a real computation engine. We conjecture now that Life is Turing-complete, and will prove it by construction. While it would be amazing to fit such a proof in a blog post, in reality we will explain a sketch the proof, elaborate on certain parts, and defer to the large body of work already done on this problem to assert our claim.

Bits and Gates

Recall there is a sufficient pair of conditions for a computational model to be Turing-complete: the ability to implement arbitrary logic functions and the existence of a model for random access memory (the read-write tape and write head).

In standard computers, these logic functions are built up via elementary logic gates. High and low current, representing 1 and 0, respectively, are sent through the logic gates, which output the appropriate level of current corresponding to the logic function. The easiest set of complete logic gates are And, Or, and Not, from which any arbitrarily complex truth table can be built.

On the other hand, it is not hard to prove that the Nand (X Nand Y := Not (X And Y)) function alone is sufficient to implement all logic functions. And so in contemporary circuit design, Nand gates (which are cheap to manufacture) are used in the billions to implement all of the necessary logic in a computer chip.

Hence, the ability to simulate a complete set of logic gates in Life is necessary for its Turing-completeness. From our investigation of the patterns above, there is one obvious candidate for current: a Gosper gun. A stream of emitted gliders corresponds to high current, and an absence to low. We include a special “eater” stationary pattern which controls the current.

The Gosper gun current, with an eater to control flow. The glider passes through iff the red cell is alive.

Further, another copy of this eater can be used to manage the output, and multiple eaters can be combined to handle two input streams, thus implementing logic. Indeed, the construction is very detailed, and requires a lot of tinkering to understand. Here we present the And gate, and send the reader to LogiCell for the designs of Or and Not.

Logical And gate

A and B represent the input currents to the gate, while C is a continuous stream. The two eaters at bottom center allow a current to pass through if and only if the current hitting it comes from B and only B. Current coming from A collides with C, cancelling both streams. If A is off (as it is in the diagram above), then B cancels with C and the eaters simultaneously, and no gliders get through. If A is off but B is on, then A cancels with C, and B still does not hit get through. However, if A and B are both on, then everything works great.

So building up from the And, Or, and Not pieces, we may implement every possible logic function. Thus, as in the original proof of Life’s Turing-completeness, we can model a finite state machine attached to two counters, which itself is Turing-complete.

[The proof of the two-counter Turing-completeness is sketched as follows. Every Turing machine can be simulated by two stacks. i.e., If H is the position of the read-write head of a Turing machine, then the head of one stack corresponds to the values to the right of H including H, while the second stack corresponds to the values strictly to the left of H. Then, any stack can be simulated by two counters, where the bits of one counter are the bits in successive cells in the stack, and the second number is required extra space for stack operations. Hence, a two stack machine can be simulated by four counters. Finally, four counters may be simulated by two counters, where one counter contains a number x=2^a3^b5^c7^d, where a,b,c,d correspond to the integer values of our four simulated counters, and the second counter is used in the arithmetic to modify x. Therefore, one counter contains the information of all four stacks. Working backward, a finite state machine which can control two counters has the same computing power as a Turing machine. It is thus Turing-complete.]

A Gargantuan Pattern

Now, proving Turing-completeness and implementing a computing machine within Life are two very different things. For one thing, we require some system of registers and memory access. Amazingly enough, researchers have created fully universal Turing machines (which can simulate other Turing machines). We point the reader to a picture of the initial configuration for a small Turing machine and a very detailed description of its parts and instruction set.

The glorious promise of Turing-completeness must be taken with a grain of salt. If we were to run within Life some wonderful computation that we might actually find useful as human beings, it would probably take longer than the life of the sun to complete. Indeed, we don’t actually care about computational speed or the prettiness of its output. Our true goal was the theoretical proof that this model is equivalent in power to a Turing machine. This has a number of profound implications (most of which have been voiced before by the Big Names of mathematics).

Specifically, we initially started with a very simple set of rules. From this, we observed much chaotic behavior, but found some order in still patterns, gliders, and oscillators. After thoroughly mastering these pieces, we suddenly found the ability to compute anything that can be computed! All of this order was hiding in chaos.

Furthermore, given an infinite grid with random initial configuration, a Turing machine sub-pattern is guaranteed to exist in it with probability 1 (see the Infinite Monkey Theorem). Not only that, but there are guaranteed to exist sub-patterns corresponding the Turing machines for every possible computation. This includes the Facebook social graph, Halo 3, the proof of the Four Color Theorem, and every program that will ever be written in the future. All in the same grid.

So with the minuscule initial design of a few simple rules, and given enough randomness, there is guaranteed to be order and elegance of the most magnificent and mind-boggling nature. Not even in Conway’s wildest dreams would we find such beauty! This is a gem of mathematics. We leave it to the reader to extrapolate philosophy and debate theories of intelligent design; we are content to admire.

At some point in the future, we wish to investigate using genetic programming to search for “interesting” Life patterns. Furthermore, the idea came upon us to run a Life-like game with k-regular cells where k is arbitrary. For large k, tessellation of these cells is only possible in the hyperbolic plane, but with the appropriate geometric software, this may give an interesting visualization of a variant of Life where, say, each cell has eight, nine, or ten neighbors (hexagonal cells has been done, as tessellation is easy in the Euclidean plane). Of course, even though a hyperbolic tessellation is indeed infinite, the cells grow exponentially smaller as they near the edge of the plane, effectively restricting our working space. Implementing this variant would require a bit of research, so we will likely write on other topics in the mean time.

Until next time!

The Wild World of Cellular Automata

So far on this blog we’ve been using mathematics to help us write interesting and useful programs. For this post (and for more in the future, I hope) we use an interesting program to drive its study as a mathematical object. For the uninformed reader, I plan to provide an additional primer on the theory of computation, but for the obvious reason it interests me more to write on their applications first. So while this post will not require too much rigorous mathematical knowledge, the next one we plan to write will.

Cellular Automata

There is a long history of mathematical models for computation. One very important one is the Turing Machine, which is the foundation of our implementations of actual computers today. On the other end of the spectrum, one of the simpler models of computation (often simply called a system) is a cellular automaton. Surprisingly enough, there are deep connections between the two. But before we get ahead of ourselves, let’s see what these automata can do.

A cellular automaton is a space of cells, where each cell has a fixed number of possible states, and a set of rules for when one state transitions to another. At each state, all cells are updated simultaneously according to the transition rules. After a pedantic, yet interesting, example, we will stick to a special two-dimensional automata (n \times n grids of cells), where the available states are 1 or 0. We will alternate freely between saying “1 and 0,” “on and off,” and “live and dead.”

Consider a 1-dimensional grid of cells which has infinite length in either direction (recalling Turing Machines, an infinite tape), where each cell can contain either a 0 or 1. For the sets of rules, we say that if a cell has any immediately adjacent neighbor which is on, then in the next generation the cell is on. Otherwise, the cell is off. We may sum up this set of rules with the following picture (credit to Wolfram MathWorld):

The state transition rule for our simple cellular automaton.

The first row represents the possible pre-transition states, and the second row is the resulting state for the center cell in the next generation. Intuitively, we may think of these as bacteria reproducing in a petri dish, where there are rigorous rules on when a bacteria dies or is born. If we start with a single cell turned on, and display each successive generation as a row in a 2-dimensional grid, we result in the following orderly pattern (again, credit to Wolfram MathWorld for the graphic):

The resulting pattern in our simple cellular automaton.

While this pattern is relatively boring, there are many interesting patterns resulting from other transition rules (which are just as succinct). To see a list of all such elementary cellular automaton, see Wolfram MathWorld’s page on the topic. Indeed, Stephen Wolfram was the first to classify these patterns, so the link is appropriate.

Because a personification of this simulation appears to resemble competition, these cellular automata are sometimes called zero-player games. Though it borrows terminology from the field of game theory, we do not analyze any sort of strategy, but rather observe the patterns emerging from various initial configurations. There are often nice local or global equilibria; these are the treasures to discover.

As we increase the complexity of the rules, the complexity of the resulting patterns increases as well. (Although, rule 30 of the elementary automata is sufficiently complex, even exhibiting true mathematical chaos, I hardly believe that anyone studies elementary automata anymore)

So let’s increase the dimension of our grid to 2, and explore John Conway’s aptly named Game of Life.

What Life From Yonder Automaton Breaks!

For Life, our automaton has the following parameters: an infinite two-dimensional grid of cells, states that are either on or off, and some initial configuration of the cells called a seed. There are three transition rules:

  1. Any live cell with fewer than two or more than three living neighbors dies.
  2. Any dead cell with exactly three living neighbors becomes alive.
  3. In any other case, the cell remains as it was.

Originally formulated by John Conway around 1970, this game was originally just a mathematical curiosity. Before we go into too much detail in the mathematical discoveries which made this particular game famous, let’s write it and explore some of the patterns it creates.

Note: this is precisely the kind of mathematical object that delights mathematicians. One creates an ideal mathematical object in one’s own mind, gives it life (no pun intended), and soon the creation begins to speak back to its creator, exhibiting properties far surpassing its original conception. We will see this very process in the Game of Life.

The rules of Life are not particularly hard to implement. We did so in Mathematica, so that we may use its capability to easily produce animations. Here is the main workhorse of our implementation. We provide all of the code used here in a Mathematica notebook on this blog’s Github page.

(* We abbreviate 'nbhd' for neighborhood *)
getNbhd[A_, i_, j_] := A[[i - 1 ;; i + 1, j - 1 ;; j + 1]];

evaluateCell[A_, i_, j_] :=
  Module[{nbhd, cell = A[[i, j]], numNeighbors},

   (* no man's land edge strategy *)
   If[i == 1 || j == 1 || i == Length[A] || j == Length[A[[1]]],
    Return[0]];

   nbhd = getNbhd[A, i, j];
   numNeighbors = Apply[Plus, Flatten[nbhd]];

   If[cell == 1 && (numNeighbors - 1 < 2 || numNeighbors - 1 > 3),
    Return[0]];
   If[cell == 0 && numNeighbors == 3, Return[1]];
   Return[cell];
   ];

evaluateAll[A_] := Table[evaluateCell[A, i, j],
   {i, 1, Length[A]}, {j, 1, Length[A[[1]]]}];

This implementations creates a few significant limitations to our study of this system. First, we have a fixed array size instead of an infinite grid. This means we need some case to handle live cells reaching the edge of the system. Fortunately, at this introductory stage in our investigation we can ignore patterns which arise too close to the border of our array, recognizing that the edge strategy tampers with the evolution of the system. Hence, we adopt the no man’s land edge strategy, which simply allows no cell to be born on the border of our array. One interesting alternative is to have the edges wrap around, thus treating the square grid as the surface of a torus. For small grids, this strategy can actually tamper with our central patterns, but for a large fixed grid, it is a viable strategy.

Second, we do not optimize our array operations to take advantage of sparse matrices. Since most cells will usually be dead, we really only need to check the neighborhoods of live cells and dead cells which have at least one live neighbor. We could keep track of the positions of live cells in a hash set, checking only those and their immediate neighbors at each step. It would not take much to modify the above code to do this, but for brevity and pedantry we exclude it, leaving the optimization as an exercise to the reader.

Finally, to actually display this code we combine Mathematica’s ArrayPlot and NestList functions to achieve a list of frames, which we then animate:

makeFrames[A_, n_] := Map[
  ArrayPlot[#, Mesh -> True]&, NestList[evaluateAll, A, n]];

animate[frames_] := ListAnimate[frames, 8, ControlPlacement -> Top];

randomLife = makeFrames[RandomInteger[1, {20, 20}], 200];
animate[randomLife]

Throwing any mathematical thoughts we might have to the wind, we just run it! Here’s the results for our first try:

What a beauty. The initial chaos almost completely stabilizes after just a few iterations. We see that there exist stationary patterns, the 2×2 square in the bottom left and the space-invader in the top right. Finally, after the identity crisis in the bottom right flounders for a while, we get an oscillating pattern!

Now hold on, because we recognize that this oscillator (which we henceforth dub, the flame) is resting against the no man’s land. So it might not be genuine, and only oscillate because the edge allows it to. However, we notice that one of the patterns which precedes the flame is a 3×3 live square with a dead center. Let’s try putting this square by itself to see what happens. In order to do this, we have an extra few lines of code to transform a list of local coordinates to a pattern centered in a larger grid.

patternToGrid[pts_List, n_] :=
  With[{xOff = Floor[n/2] - Floor[Max[Map[#[[2]] &, pts]]/2],
        yOff = Floor[n/2] - Floor[Max[Map[#[[1]] &, pts]]/2]},
   SparseArray[Map[# + {yOff, xOff} -> 1 &, pts], {n, n}, 0]];
square = {{1, 1}, {1, 2}, {1, 3}, {2, 1}, {2, 3},
  {3, 1}, {3, 2}, {3, 3}};

Combining the resulting two lines with the earlier code for animation, we produce the following pattern:

While we didn’t recover our coveted flame from before, we have at least verified that natural oscillators exist. It’s not hard to see that one of the four pieces above constitutes the smallest oscillator, for any oscillator requires at least three live cells in every generation, and this has exactly three in each generation. No less populated (static or moving) pattern could possibly exist indefinitely.

Before we return to our attempt to recreate the flame, let’s personify this animation. If we think of the original square as a densely packed community, we might tend to interpret this pattern as a migration. The packed population breaks up and migrates to form four separate communities, each of which is just the right size to sustain itself indefinitely. The astute reader may ask whether this is always the case: does every pattern dissipate into a stable pattern? Indeed, this was John Conway’s original question, and we will return to it in a moment.

For now, we notice that the original square preceding the flame grew until its side hit a wall. Now we realize that the wall was essential in its oscillation. So, let us use the symmetry in the pattern to artificially create a “wall” in the form of another origin square. After a bit of tweaking to get the spacing right (three cells separating the squares), we arrive at the following unexpected animation:

We admit, with four symmetrically oscillating flames, it looks more like a jellyfish than a fire. But while we meant to produce two flames, we ended up with four! Quite marvelous. Here is another beautiful reject, which we got by placing the two squares only one cell apart. Unfortunately, it evaporates rather quickly. We call it, the fleeting butterfly.

We refrain from experimenting with other perturbations of the two-square initial configuration for the sake of completing this post by the end of the year. If the reader happens to find an interesting pattern, he shouldn’t hesitate to post a comment!

Now, before returning to the stabilization question, we consider one more phenomenon: moving patterns. Consider the following initial configuration:

A few mundane calculations show that in four generations this pattern repeats itself, but a few cells to the south-east. This glider pattern will fly indefinitely to its demise in no man’s land, as we see below.

Awesome. And clearly, we can exploit the symmetry of this object to shoot the glider in all four directions. Let’s see what happens when they collide!

Well that was dumb. It’s probably too symmetric. We leave it as an exercise to the reader to slightly modify the initial position (given in the Mathematica notebook on this blog’s Github page) and witness the hopefully ensuing chaos.

Now you may have noticed that these designs are very pretty. Indeed, before the post intermission (there’s still loads more to explore), we will quickly investigate this idea.

Automata in Design

Using automata in design might seem rather far-fetched, and certainly would be difficult to implement (if not impossible) in an environment such as Photoshop or with CSS. But, recalling our post on Randomness in Design, it is only appropriate to show a real-world example of a design based on a cellular automaton (specifically, it seems to use something similar to rule 30 of the elementary automata). The prominent example at hand is the Conus seashell.

A Conus shell.

The Conus has cells which secrete pigment according to some unknown set of rules. That the process is a cellular automaton is stated but unsupported on Wikipedia. As unfortunate as that is, we may still appreciate that the final result looks like it was generated from a cellular automaton, and we can reproduce such designs with one. If I had more immediate access to a graphics library and had a bit more experience dealing with textures, I would gladly produce something. If at some point in the future I do get such experience, I would like to return to this topic and see what I can do. For the moment, however, we just admire the apparent connection.

A Tantalizing Peek

We have yet to breach the question of stabilization. In fact, though we started talking about models for computation, we haven’t actually computed anything besides pretty pictures yet! We implore the reader to have patience, and assert presciently that the question of stabilization comes first.

On one hand, we can prove that from any initial configuration Life always stabilizes, arriving at a state where cell population growth cannot continue. Alternatively, we could discover an initial configuration which causes unbounded population growth. The immature reader will notice that this mathematical object would not be very interesting if the former were the case, and so it is likely the latter. Indeed, without unbounded growth we wouldn’t be able to compute much! Before we actually find such a pattern, we realize that unbounded growth is possible in two different ways. First, a moving pattern (like the glider) may leave cells in its wake which do not disappear. Similarly, a stationary pattern may regularly emit moving patterns. Next time, we will give the canonical examples of such patterns, and show their use in turning Life into a model for computation. Finally, we have some additional ideas to spice Life up, but we will leave those as a surprise, defaulting to exclude them if they don’t pan out.

Until next time!