Game Career Guide is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

Get the latest Education e-news
  • Book Excerpt: AI for Game Developers

    - Glenn Seeman and David Bourg

  • Feature!Classic Flocking

    Craig Reynolds coined the term boids when referring to his simulated flocks. The behavior he generated very closely resembles shoals of fish or flocks of birds. All the boids can be moving in one direction at one moment, and then the next moment the tip of the flock formation can turn and the rest of the flock will follow as a wave of turning boids propagates through the flock. Reynolds' implementation is leaderless in that no one boid actually leads the flock; in a sense they all sort of follow the group, which seems to have a mind of its own. The motion Reynolds' flocking algorithm generated is quite impressive. Even more impressive is the fact that this behavior is the result of three elegantly simple rules. These rules are summarized as follows:


    Have each unit steer toward the average position of its neighbors.


    Have each unit steer so as to align itself to the average heading of its neighbors.


    Have each unit steer to avoid hitting its neighbors.

    It's clear from these three rule statements that each unit must be able to steer, for example, by application of steering forces. Further, each unit must be aware of its local surroundings—it has to know where its neighbors are located, where they're headed, and how close they are to it.

    In physically simulated, continuous environments, you can steer by applying steering forces on the units being simulated. Here you can apply the same technique we used in the chasing, evading, and pattern movement examples earlier in the book. (Refer to Chapter 2 and, specifically, to Figure 2-7 Chapters 2 and 3. Although particles are easier to handle in that you don't have to worry about rotation, it's very likely that in your games the units won't be particles. Instead, they'll be units with a definite volume and a well defined front and back, which makes it important to track their orientations so that while moving around, they rotate to face the direction in which they are heading. Treating the units as rigid bodies enables you to take care of orientation.
    and surrounding discussion to see how you can handle steering.) We should point out that many flocking algorithms you'll find in other printed material or on the Web use particles to represent the units, whereas here we're going to use rigid bodies such as those we covered in

    For tiled environments, you can employ the line-of-sight methods we used in the tile-based chasing and evading examples to have the units steer, or rather, head toward a specific point. For example, in the case of the cohesion rule, you'd have the unit head toward the average location, expressed as a tile coordinate, of its neighbors. (Refer to the section “Line-of-Sight Chasing” in Chapter 2.)

    To what extent is each unit aware of its neighbors? Basically, each unit is aware of its local surroundings—that is, it knows the average location, heading, and separation between it and the other units in the group in its immediate vicinity. The unit does not necessarily know what the entire group is doing at any given time. Figure 4-1 illustrates a unit's local visibility.

    Figure 4-1. Unit visibility


    Figure 4-1 illustrates a unit (the bold one in the middle of the figure) with a visibility arc of radius rr, and the angle, θ. Both parameters affect the resulting flocking motion, and you can tune them to your needs.
    around it. The unit can see all other units that fall within that arc. The visible units are used when applying the flocking rules; all the other units are ignored. The visibility arc is defined by two parameters—the arc radius, r, and the angle, θ. Both parameters affect the resulting flocking motion, and you can tune them to your needs.

    In general, a large radius will allow the unit to see more of the group, which results in a more cohesive flock. That is, the flock tends to splinter into smaller flocks less often because each unit can see where most or all of the units are and steer accordingly. On the other hand, a smaller radius tends to increase the likelihood of the flock to splinter, forming smaller flocks. A flock might splinter if some units temporarily lose sight of their neighbors, which can be their link to the larger flock. When this occurs, the detached units will splinter off into a smaller flock and could perhaps rejoin the others if they happen to come within sight again. Navigating around obstacles also can cause a flock to break up. In this case, a larger radius will help the flock to rejoin the group.

    The other parameter, θ, measures the field of view, so to speak, of each unit. The widest field of view is, of course, 360 degrees. Some flocking algorithms use a 360-degree field of view because it is easier to implement; however, the resulting flocking behavior might be somewhat unrealistic. A more common field of view is similar to that illustrated in Figure 4-1, where there is a distinct blind spot behind each unit. Here, again, you can tune this parameter to your liking. In general, a wide field of view, such as the one illustrated in Figure 4-2 in which the view angle is approximately 270 degrees, results in well formed flocks. A narrow field of view, such as the one illustrated in Figure 4-2 in which the view angle is a narrow 45 degrees, results in flocks that tend to look more like a line of ants walking along a path.

    Figure 4-2. Wide versus narrow field-of-view flock formations


    Both results have their uses. For example, if you were simulating a squadron of fighter jets, you might use a wide field of view. If you were simulating a squad of army units sneaking up on someone, you might use a narrow field of view so that they follow each other in a line and, therefore, do not present a wide target as they make their approach. If you combine this latter case with obstacle avoidance, your units would appear to follow the point man as they sneak around obstacles.

    Later, we'll build on the three flocking rules of cohesion, alignment, and separation to facilitate obstacle avoidance and leaders. But first, let's go through some example code that implements these three rules.


comments powered by Disqus