Stickers

Stickers was primarily inspired by Karl Sims famous Evolving Virtual Creatures. His project consisted of taking a Connection Machine (a massively parallel super computer), and simulating a physically accurate 3D world populated with creatures that were constructed of rectangular blocks that were hinged against one another to form a creature of segmented block-limb type ornaments. In addition to physical structure these creatures had a nervous system for controlling the hinges. In other words they would flail around. Sims applied a standard genetic algorithm to drive the creatures' evolution such that they evolved both physical form and a nervous system for moving that form. The goal he was selecting for was energy efficient movement. His results were spectacular. The 3D block creatures were elegant in structure and moved and graceful patterns.

I was curious whether I could reproduce Sims' project and then expand on it. I didn't have a supercomputer and the physics equations for Sims project seemed pretty immense, so I designed a simpler universe. The result was Stickers. Stickers is a 2D universe consisting of what would be analagous to a liquid medium. In other words, it is a viscous environment where a push will produce a measurable thrust, unlike the gas environment that Earth's land-dwelling organisms inhabit. The stickers themselves are constructed of stick segments that can hinge around their end-points. The basic idea is identical to Karl Sims. I even used his directed-graph implementation for storing the genotype and then growing the phenotype of each sticker.

Why did I try to reproduce someone else's highly successful project? A particularly good question in light of the fact that I dumbed down the universe so much. Karl Sims did a good job of evolving individual creatures against a fitness function. He did some minimal expansions on this idea such as having two creatures evolve against each other in persuit of a designated goal, but that is as far as he went. I wanted a living world on my computer screen in front of me. The idea of starting with physics and evolving organisms that are adapted to that physics and then interact and compete for the basic resources of life in the context of that physics struct me as extremely life-like. My long-term goal with Stickers was to allow large populations of organisms to evolve diverging cohabitating species based on different survival strategies, all living and competing together in a single physically vivid world.

I imagined simple mindless tiny creatures that would act more or less like microbes or plankton. I imagined a higher level of slighter larger creatures that might scoop these little guys up like popcorn. I imagined larger creatures still, a whole food chain. I imagined strategies such as sitting still and grabbing what comes by, like a carnivorous plant of some sort. These ideas were all perfectly "allowed" by the basic physics of the simulation.

Some basic points about Stickers: It is a closed energy system. There are only two kinds of matter, food-dots and stickers. Stickers can eat food-dots or can eat other stickers with a success rate proportional to the difference in size of the two stickers. Since Stickers is a closed energy system, all the energy of a sticker that has just died is converted into an equivalent amount of energy in the form of food-dots. As stickers burn up kinetic energy in muscle exertion their expended "heat" is accumulated by the environment and redistributed in the form of food-dots as well. The same goes for all other kinds of energy expenditure such as reproduction and a basic metabolism tax. Some ideas were vague and were attempted at certain times throughout the project. These included sexual reproduction, a variable litter size when reproducing, certain segments designated "mouths" which were the only segments capable of consuming food, etc.

Stickers never really took off. It certainly got off the ground as can be seen from the animation on this page. I successfully created an amazing universe that would fill up with beautiful creatures, but there were two problems, one of which I believe was the result of the other. Physical laws were the basis for the universe. Everything else depended on a good interesting physics and even in two dimensions I found the forward dynamics equations to be too dauting to implement successfully. I tried for weeks and got incremental improvements but never utter success. The flawed physics resulted in the second major problem, poor evolution. If the organisms couldn't move appropriately then they couldn't get food and if the physics was broken, they couldn't move appopriately which means they certainly couldn't evolve intelligent locomotion ability on their own. As a result, everything else I hoped for cascaded down to these problems. No complicated hierarchies of food chains or sexual competition or survival strategies and niches ever developed because the basic locomotion remained beyond the stickers' evolutionary reach. They are capable of flailing about and moving through the environment, but the bad physics makes for a weak correlation between the morphology/nervous system and controlled methodical movement. As a result, the most successful stickers are at the microbe level. These stickers, usually consisting of three segments either in a line or more commonly as three spokes, just spin like crazy and slowly ooze around the environment accidentally falling across food-dots randomly. Because they are small their metabolic tax is minimal. Because they have very few hinges (muscles), their kinetic cost is similarly low. They outsurvive larger stickers every time even though they would provide a tasty meal for larger carnivore. No sticker has ever demonstrated the ability to sense the direction of a food-dot or prey sticker and direct its movement trajectory purposefully in that direction.

It should be noted that other people have had better success. There are programs with a strikingly similar approach to Stickers. They have a 2D liquid environment with segmented stick-creatures much like Stickers. I doubt if the genotype is specified in an identical fashion but the basic idea is there. These other people have had better success with the physics engine and have subsequently produced relatively interesting worlds, although they still lack the complex ecosystem that I envision.

I don't have a downloadable version of Stickers because the program is only in an experimental stage of completion. The user interface is obfuscated to say the least, not the mention the the fact that I consider it to be a relatively unusable program due to the flawed physics engine. Sorry.