Current Headway

an investigation of differences in [the perception of] time between architecture and [first-person / brief] games, and how this impacts social encounters

Thursday, 16 June 2011

cities through algorithms

BELOW: I stumbled upon this article online that focuses on a wonder in gaming that I found curious. The generating of cities made using mathematical algorithms. By definition an abstraction of real world architecture to gaming architecture is taking place, which brings up several questions, as to how on earth somebody could pull something like this of? and how convincing it could be? Apparently the benefits of such a thing, saves valuable time and resources, and often yields very interesting results compared to creations made by artists. These algorithms use hundreds of parameters, and sometimes by simply altering these parameters, very unique fiction can be created. What I am personally interested in is the implications of such a phenomena on the physical game world itself. Using AI and allowing the player to explore and interact with NPC's, how does the change of parameters affect the actual behavior of people inside the world, or what it could imply… I'm not just talking about road networks, and traffic changes, (although that in itself is interesting too) bur rather the way it would influence social encounters. Take Paris for instance, it is no shocker that the center of the city is the busiest and perhaps safest area, and that the slums exists radially in the outskirts. As the city grows so does its epicenter and once slums are now middle-class residencies. How can an AI be configured to simulate such a phenomena through changes in parameters. After all what would make a city more real than for the player to feel tendencies to stick to busier streets rather than slums, and allow for the NPC's to behave like real people in large cities do. A simulation of AI in relation to the changes of parameters within a modeled city. In a way, this phenomena could carry on what "The Sims" tried to do, attempt to model the behavior of people in spatial environments more realistically then previous games (GTA, MAFIA, lA NOIRE etc.) have done. On the other hand such a simulation would never truly mimic the spontaneous behavior of people in reality, but it can always be used as a tool, especially in architecture.

  
Procedural Destruction and the Algorithmic Fiction of the City

 

[Image: From Procedural Modeling of Cities by Yoav Parish and Pascal Müller].

Note: This is a guest post by Jim Rossignol.

In 2001 Yoav Parish and Pascal Müller spoke at the SIGGRAPH conference in Los Angeles, California, to present a mathematical city. Their presentation contained an algorithmic approach for modeling city-like topologies. The results were remarkably realistic, and were one among a host of city-like generative systems to appear at the start of the decade.

Another, Jared Tarbell's Substrate (pictured) remains a fantastic example of how a mathematical approach to generating apparently urban patterns can also be artful.

[Image: From Jared Tarbell's Substrate].

But it was looking at the work of Parish and Müller that inspired game designer Chris Delay to develop his most recent project: the cryptic (and as-yet-unexplained) Subversion, of which little is known, other than it relies on large, procedurally generated cities for the backbone of its game world.

Having already been burned by the problems of creating content "by hand," Delay set out to let algorithms do the work of building buildings in his new game. Not only that, but he was determined to create an artistically interesting experience without artists.

[Images: From Chris Delay's Subversion].

Of course, videogames have long been the home of procedurally generated landscapes where numbers and mathematical equations played the role of the visual designer. Early paranoid classic The Sentinel made use of these techniques to create an astonishingly atmospheric 10,000 levels in simple vector graphics, from just a few kilobytes of data. Other games have used similar techniques as a shortcut to creating solar systems and vast fractal landscapes.

But when it came to cities, well, it took a long time for anyone to take up the challenge.

[Image: From The Sentinel by Geoff Crammond].

Rather than opt for procedural techniques, game designers usually elect to build their cities by hand, often with startling results. The re-imagined contemporary New York that features in last year's Grand Theft Auto 4 required a small army of well-paid artists and designers to hand-craft the entire world. Their accomplishment is unmatched, but the cost to the company behind the project is in the tens—and perhaps hundreds—of millions of dollars. To build up a living city from blank polygons is one of the most expensive possible projects in game design.

Delay, whose project is being undertaken with a tiny budget and by just a handful of staff based in Cambridge, UK, does not have the luxury of vast content teams. His vector-drawn city is far less realistic than Rockstar's textured, heaving metropolis, but there's nevertheless a beauty to it. It's a kind of mathematical map of the essential urban environment: there are roads, sidewalks, and a no-man's land of corporate moats around great skyscrapers...

Identify the key equation that define urban patterns, and you, too, can summon a city into existence.

Delay has begun to show off how his cities emerge from the ground up in a series of videos, and he spoke to me about the process.

"I started out with road layouts, and then began to modify the parameters," he explained. "Sometimes you'll get lovely radial, spiral patterns, or you can tell it to create a really rigid Manhattan-style grid." One set of numbers delivers the block logic of American cities, another is rather more like the spirals of Medieval European sprawls. The two merge to create something even more believable. "Every subsequent layer builds on the previous layer," Delay points out, "so the very next layer looks for the spaces between layers, and makes judgments about 'is this likely to be a skyscraper, or to be a house?' Then you zoom in, and carry on. You do another procedural generation process for each layer of detail, filling in that world."

[Image: From Chris Delay's Subversion].

A few weeks after speaking with Delay I attended Thrilling Wonder Stories—a seminar at the Architectural Association in London, curated by Liam Young and BLDGBLOG—where I watched conceptual designer Viktor Antonov explain how he had created a science-fictionalized Paris (for a now-cancelled videogame called The Crossing).

Antonov approached the problem by altering just a few parameters in the standard architectural model. For instance, Antonov had noticed a few fundamental details about how the mid-nineteenth century neo-classical core of Paris had been constructed: big street-level floors, smaller attic spaces, complex chimney stacks. By increasing the emphasis on the lower floors, and stretching them out—and by emphasizing the height and complexity of the chimneys—Antonov was able to create a thematically consistent science fiction Paris.

Simply by altering a few basic architectural parameters, he said, you were able to fictionalize the city, whilst at the same time retaining its fundamental identity. His designs were still recognizably—even mathematically—Parisian, in other words, but they were also otherworldly.

[Image: By Viktor Antonov].

This idea instantly connected back to Delay's project: what parameters would we need in order both to understand and create a science fiction Edinburgh, or Sao Paulo, or Vancouver? Identify the necessary fantasy logic within a procedural city-building system and you could recreate cities with their alternate identity in an instant. An accelerated future Moscow, or a retropunk Venice, instantly sprawling out of the monitor.

And perhaps this is not such an outlandish thing to aim for—especially when you consider the speed at which procedural city projects have been appearing across the tech landscape. Could one of these cities potentially be refitted to allow for this type of radical tweak?

Projects like Shamus Young's impressive PixelCity, or Marco Corbetta's Structure seem ripe for such strange fictions. Corbetta's system is particularly impressive in its verisimilitude: he aims to create a basic engine for rapidly generating the kinds of cities that games like Grand Theft Auto 4 require, and consequently doing so for much cheaper.

Could Corbetta's engine come with a Paris or a Barcelona preload, which could then be put through Photoshop-style filters for alternate reality logic in its architecture? A stronger skyline, weirder street furniture.

[Image: From Marco Corbetta's Structure].

More exciting, at least for the thrill-seeking gamer in me, is the fact that Corbetta is aiming one notch higher than any of his peers: he's aiming to make these cities procedurally destructible. His site contains a demonstration video of neatly arrange office interiors and a domestic library being blown to pieces with a machine-gun. What good is an imaginary city if you can't go inside the buildings? What good is a virtual downtown if you can't go crazy with a bazooka? Corbetta's work preempts these questions.

Further, it conjures visions of massive demolition exercises in parallel worlds—entering an Antonov-algorithm for neo-Rome, where gladiatorial escapades see us going through the walls of the coliseum and into the randomly generated plazas beyond.

That, perhaps, is the greatest promise of procedural cities: that soon they'll be real enough that their destruction will seem like tragedy.

1 comment:

  1. Very interesting article! This use of AI and parametric strategies in the auto-generation of urban form could be quite relevant to your topic, particularly where the parameters have to do with events and social encounters.

    ReplyDelete