Small groups form to play games, to talk to one another, or just as a result of passing through constricted areas. These groups might consist of adults, big kids, small children, or a mix of all three, each mix with a different dynamic.

The large group, or entire crowd on the playground, has its own rhythm and movement. Children choose which games to play or which equipment to use, form lines to wait, move around and through other groups; people come and go. All this is often done at a fast pace and in a chaotic way; children might shove through lines or get easily bored and frequently switch from one activity to another.

This quarter

I propose that the large group activity on a playground is both tractable and interesting, and that we explore such activity this quarter. To do so, we would build a small playground environment, with some different kinds of terrain, such as sand, grass, hills, and concrete, and a few kinds of equipment; we would provide for children to come and go and give each behavioral rules and motivations as well as rules for interacting with each part of the environment.

Research objectives

The primary goal of the project is to explore group behavior in an environment different from the standard city scenes or disaster response scenarios. The playground scenario involves the interaction of agents with differing motivations and abilities, navigating through a rich environment. This is interesting both because of the underlying complexity, but also because of the desired character of the result; a playground should look "fun". Beyond these kinds of inherent interest, such a simulation might be useful for people developing parks or other leisure environments, or in the design of video games.

A secondary goal is to develop an integrated system for doing medium-scale simulations like this one. I am thinking here of a game-style system for animation research, in the general vein of things like SimCity, Populous, The Sims, Second Life, Croquet, or Spore, but on the other hand with full programmability, more like a programming language or a Wiki page where the primitives are characters and environmental objects, and the abilities are geared for animation. We have talked elsewhere about a "Massive"-style system for use in research; this would be perhaps along similar lines, but a little less massive; call it "bantamweight". The idea would be to provide a macro-environment backdrop for doing (future) research on individual character movement, and a micro-environment backdrop for doing research on larger groups, in one integrated space.

Major tasks

I am thinking of this as a small group project, for (say) three people, but it is likely possible to scale that number up to four or five or down to two.

We will need to define some scenarios and in doing so narrow down the vague goals above to some more concrete ones. Of course these will affect and to some extent define the rest of the tasks.

We will need an environment, with varied terrain and equipment. We could initially use simple geometric equipment, with ladders, slides, bridges, and the like. Interaction with the equipment could be physically simulated or rules-based.

We will need agents of some kind. These will need a system for making choices and interacting with the environment.

The agents will of course move in some way. I think it is best if the agents look individually playful, but I think considering realistic synthetic motion for children is likely to take longer than one quarter to do with any kind of quality. (Some related research is here.) We may be able to use geometric characters with evocative behaviors; for an example, see Ken Perlin's Polly. Motion capture would also be an option, but it is possible (in an uncanny valley sense) that evocative geometry will in fact look more appropriate than realistic characters.

There should be plenty of research from other fields on how children think and interact with one another. (In our field, this has been studied extensively for topics in programming and HCI, but perhaps not as much for large-muscle or group activity). This research would likely be helpful to our efforts.

Some related work

There has been much recent work in group animation, mostly involving large groups or navigation through restricted environments (for example, for building evacuation). Some notes on crowd animation from this winter's 888 are here. Matt Boggus's page on the "Massive" idea also has a number of useful links.

There are also several recent papers in basic motion planning, some in complex environments. For example, Lau and Kuffner present a method based on motion capture segments that allows characters to navigate around, over, and under changing obstacles.