Simulating a crowd with the help of Flocking Artificial Intelligence
Don’t you just love watching a flock of birds in the sky; twisting and turning, creating the most stunning patterns? This natural occurrence is not just for our own selfish enjoyment; it is in fact a natural emergent behaviour, also common in species such as insects and fish.
Who would have thought that something so simplistic and beautiful could influence the way that films and games are made these days!
How does flocking work?
For simplicity sake, I will use a flock of birds in the following example.
When a bird flys in a flock, it pays close attention to the actions of its neighbours, and then adjusts its own position accordingly. There are three main factors involved:
1. Separation – To avoid the flock getting squashed, the bird will keep an even distance apart
2. Alignment – Each bird will travel in the general direction of its neighbours. The bird at the front will often fly to the back and allow another to take it's place and vice versa. Please note however that flocking is not determined by any one solitary bird
3. Cohesion – The position of
the bird will also change depending what the other birds are doing and the way they are facing
These three factors work in synergy with each other to
create the effects of a flock - just the way a swarm of insects or a school of fish would. There is neither one factor more dominant than
the other since it is through such complex interactions that order is born. Flocking is all about cause and effect.
Another example is a crowd full of people, making their way from one part of town to another. Sometimes it is easy to follow a crowd of strangers without consciously thinking of our actions. We have all done it – ended up somewhere totally unexpected (I guess that is why flocking AI has played an important influence in zombie crowd simulation in games)!
Here is a little animation I made in Adobe Flash to demonstrate the impact of separation, alignment and cohesion in a school of fish.
Why do creatures travel together in flocks, schools and swarms etc?
In nature, creatures will often stay close together for increased body warmth, protection against predators, reproduction and ultimately to assist in navigation.
Flocking AI has acted as a valuable tool for scientists and nature lovers alike because it has helped one gain a greater understanding of the way creatures interact with each other.
Simulating Flocking behaviour
Although it is not impossible to simulate flocking behaviour using manual animation techniques, there are by far easier and more realistic ways of doing it using the help of flocking algorithms.
Craig Reynolds was the first person to simulate flocking behaviour using his program called “Boids” in 1986; pretty dated to look at now, although Reynolds has totally revolutionised the modern entertainment industry. Tim Burton’s “Batman Returns” and Disney’s “The Lion King” were amongst the first films to feature Reynolds ‘Boid model’ to create realistic swarms and stampedes.
If you want to have a go at making your own flocking simulations, you can do it relatively easily these days with the help of 3d animation software. A majority of these programs will have plug-ins available that do all the hard calculations for you. The simplest method would be to animate a single object, then to make several instances of it played back at random. The more variety the animation, the more natural the finished flock or crowd will be.
Autodesk Maya for instance has a crowd maker utility that can handle crowds of up to 1000 (on the condition that your computer can handle it of course)! It is also worth noting that Adobe After Effects has a nifty little plug-in called Trapcode Particular which can transform any 2d picture into a full 3d flocking simulation. Not only can you make realistic flocking animations using Trapcode, you have the aid of around 100 variables to play around with – each creating unique effects.
The algorithm in depth
Although this article offers you a brief outline of Flocking artificial intelligence, you can find out how the flocking algorithm works in more detail using the link below (courtesy of Conrad Parker).
- Boids Pseudocode
Homepage of Conrad Parker
- Boids (Flocks, Herds, and Schools: a Distributed Behavioral Model)
Background and update on BOIDS, the model of group motion in flocks, herds, schools and related phenomena. Includes a Java-based demonstration and many links to related research and applications.