The Anatomy of a Squirrel

• Space - Managing the world, and the navigation
• Squirrel - Managing the squirrel, and it's behavior
• Tree - Managing the tree, and it's behavior
• Nut - Managing the nut, and it's behavior

Module ownership

General things

Modules

A typical Meme module loads some geometry from disk, and sets up at one or more tasks to manage that objects behavior. A module for a spinning ball would load the ball geometry in it's creation routine, and set up a task for the spinning.

Compound objects

A compound object is written such that all it's geometry isn't liked to world as you would normally have it, but to ModuleWorld. The ModuleWorld object represents the object under which this compound object is placed when it is instantiated.

If you have a standalone "spinning ball" module, turning it into a compound object requires you to:

• Remove any camera declaration and rendering loop
• Link all geometry created to ModuleWorld instead of world

You now have a spinning ball object that can by used anywhere.

The ability to rescale and move the object at will, or to link it into the object heirarchy at any point, gives you lots of power. If you have written a wall-clock module, you can scale it down and use it as a wristwatch for your avatar, if you want to.

The Compound Objects of the Squirrel demo

Click the modules name to see the source of each module. Note that it isn't an entry for a code beauty contest and that there is old junk, testing code, and yet-to-be-used stuff in it!!

The Space module

The space module loads the floor (the mesh on which the saga takes place), sets up a rendering task and a navigation task.

Then it loads two Tree compound objects, and a Squirrel compound object at predefined locations.

A keyboard task is also set up, which monitors the keyboard for the keystrokes 'S' and 'T', which create a new Squirrel and Tree object respectively, giving them a random location and scalefactor.

The Tree module

The tree module is the simplest of them all. It simply loads the tree-trunk and crown as geometries, and sets up a simple "brain" task.

The "brain" of the tree doesn't doo very much. It simply waits a random time, and then creates a Nut compound object at a random location in it's crown. Then it waits again, and creates another nut, and so on....

The Nut module

The nut isn't very intelligent either. It simply has a "brain" that calculates a descendance curve towards the ground, and upon impact, a bounce. After a few bounces the nut comes to rest, and the task is put to sleep.

The Squirrel module

The squirrel is the most advanced module. It contains five tasks, each for:

• blink...the blinking of the eyes, at pseudo-random intervals (initiall active).
• brain...the "brain" of the squirrel (initially active), which "decitions" the squirrel makes.
• walk...walking along the ground, doing the up-down motion, wiggling the tail, turning the legs. (initially dormant)
• turnhead...turning the head to a predefined yaw angle (initially dormant).
• munch...eating a nut (initially dormant).

The Brain task is the interesting one - it is the "scheduler" that decides which of the three sleeping tasks to execute.

The brain selects one of four basic behaviors randomly (with equal probability), and a random timespan under which to execute this behavior. The four behaviors are:

• Sit - the squirrel does nothing. The walk, turnhead and munch tasks are dormant.
• Walk - the squirrel decides a random direction, calculates the position of the nearest tree (to avoid it), and wakes the walk task.
• Look - the squirrel calculates a random head yaw angle, and launches the turnhead task.
• FindFood/Eat - this fourth behavior has a few conditions:
  • If the squirrel is already holding a nut, then eat it by launching the munch task.
  • If the squirrel is not holding a nut, but is close to a nut, then pick it up.
    This is done by sending a pickup-message to the nut, (which stops the nut's bouncing behavior) and attach the nut to the squirrels arm.
  • If the squirrel isn't close to a nut, then calculate the angle and distance to the closest nut, and launch the walk task. Note that the brain might preempt the squirrels "walk to a nut" by getting a new idea before the squirrel actually gets to the nut!
  The closest nut is found by scanning thru the object tree, and using messages inquiring each object if it is of the type "Nut" or not.

A Multi-User Squirrel

Essentially, the squirrel follows the thoughts in my behavior proposal in that it uses a "brain" that makes the indeterministic decitions, and an "engine" (in this case the other tasks) that deterministically carry out the orders of the brain.

To make the squirrel fully multi-user, would simply mean to put a message layer between these two. This is fairly easy to do, but lack of time forced me not to. I started to implement it, but have only gotten as far as a bunch of declared message structures that arn't used anywhere....

Essentially, each "decision" the brain makes would need to be communicated to the "engine" task with a message. When that is accomplished, the squirrel would be able to drive one or more remote replicas identically across the net.

All possible, but not done.

Conclusion

Until Later...

/Z