People Power

For the last 4 or 5 years we've been researching some fundamental new techniques for bringing scenes to life with people. Our focus hasn't been on "hero characters" as much on secondary visualization of realistic crowds. We call this initiative, "People Power” and it comprises two thrusts: one around ergonomics (how a single person interacts with designs) and how groups of people interact with design (crowds).

What you’re going to see now is part of our crowd research. This is very much a research project, so don’t expect it as a feature in any form that you see it now. We’re trying to solve some very fundamental problems around the animation and control of crowds. We have three primary goals:

- Figure out how to display thousands of characters as interactively as possible
- Figure out how to generate believable motion with a minimum of effort
- Figure out how to deliver believable interactions amongst crowd members (avoidance being a key one)

We’ve already filed for patents on most of this stuff, so we’re really talking some leading-edge efforts. You might have heard about our Project Geppetto, well, that is one component of what we’re trying to do here. That component is what allows us to take an “ocean of motion” and create believable transitions between motion clips that avoid all the foot sliding and sudden impossible movements of all crowd systems we’re aware of. You rarely see a camera in a fixed location watching a crowd system for any length of time because at some point, there is inevitably a large “glitch” where a character does something that defies our understanding of what humans do or can do. If we do our job right, you won’t see anything like that with People Power.

Another core technology is our Kynapse path-finding AI technology. We’ve incorporated that to help define how the space is navigated and to put a lot of spatial awareness at the hands of the system (like how close to a wall can someone walk).

There’s a lot more to People Power, but that’s all I’m going to talk about at this point. You can ask questions, but let me forewarn you that I’m not likely to talk too much about what you see here at this time. There will be a time to talk more about it. I hope you’re going to Autodesk University this year…

Some notes:
- We're only showing a limited subset of motion data that we've captured
- We're only showing two segmented characters (male/female), these are just placeholders for something else...
- We're showing some concepts for controlling flow, these are just concepts
- We haven't finished our crowd avoidance engine, this is just showing some of our techniques
- We haven't finished our representation system, this is just where we're at right now
- The segmented characters are pretty much interactive in the viewport - up to around 100 or so

Enjoy!

Note: The building model in the video is courtesy of TRO Jung|Brannen, Inc