Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


Swarm Demo

Copyright © 2012 Bart Massey

This program is licensed under the "MIT License". Please see the file COPYING in the source distribution of this software for license terms.

This is a simple Java 2D graphical physics simulation built around "bugs", little motile thingies with a tail that points away from their facing. Build it with just "javac *.java". You can invoke it with "java Swarm 10" to get a swarm of 10 bugs. If you are going to use more than about 100 bugs, you should probably reduce their size substantially or the simulator will never place them all and they won't be able to move very well.

The idea is to subclass the AI class of the bugs to provide an AI for a particular Agent in the simulation. The AI gets sensor data and can exert control over the Agent's motion. The default AI just tries to drive the bug to the center of the screen. More interesting AI's are possible. It is also relatively straightforward to add new Agents or even passive Things to the simulation, and to give the new stuff useful behavior. In short, this is just a framework, not a final answer.

The sensor and actuator data is mediated by the View classes. Take a look at how Agent sets up views to get some idea of what's happening here.

The physics of the simulation is controlled by a number of per-entity constants; minimum and maximum reverse and forward acceleration AMIN and AMAX, minimum and maximum right and left turn velocity VTMIN and VTMAX, and a coefficient of drag CDRAG for a velocity-dependent, size-squared-dependent drag term.

The whole simulation takes place in a Double coordinate space ranging from 0..1. The physics simulation timestep is governed by Playfield.DT, which is echoed as "dt" in various classes. Some effort has been expended to make sure that the whole thing runs in roughly "real-time"; one simulated second per wall-clock second. The frame rate of the display is normally less than the timestep rate, for efficiency. The display frame rate is normally 200ms or faster, which gives an adequate animation.

The Agent physics are currently built to model x and y position, facing angle t, velocity v in the facing direction, angular velocity vt, and acceleration a in the facing direction. The only currently controllable parameters are a and vt, but of course this can easily be changed.

Remember that an object with acceleration a and velocity v will move forward at a rate

    v + a * dt

Thus, the object will move a total distance of

    v * t + 0.5 * a * t * t

in time t. The damping term mentioned above will squash this a bit, though.

For dealing with angles, it is useful to recall that

    dx = dr * Math.cos(t)
    dy = -dr * Math.sin(t)
    t = Math.atan2(dy, dx)

It is also good to remember that angles are typically measured in radians, 2 * Math.PI of them around the unit circle. The class method AI.angleDiff() will compute the difference between two angles "the short way around" so that you don't make extra-long left turns.

Currently, when bugs collide with each other or a wall, all bugs involved in the collision instantly lose all their acceleration and velocity. They still turn at the normal rate, but they typically get "stuck" and can't move until they either back up or turn around.

There is a Makefile for building this. It's kind of kludgy, but it works. Say "make" to build the demo. Say "make javadoc" to build javadocs in the "doc" subdirectory.

Mostly, this code is supposed to be fun to play with. Enjoy!