PereGaea
Laslo Godel
REACTIONS
We can now look at the Dynism's ability to Act which, apart from the Reflexes and Production Rules we saw earlier, I've really said little about.
Unlike their weightless swimming predecessors, land-going Dynisms will usually need to keep most of their effectors in a constant state of operation just to maintain static configurations, like this basic four-footed stance, against the full force of PereGaean gravity. Their Joints must therefore each acquire Reflex Loops that will allow them to maintain their angular positions against all but the strongest forces. We'll call these Reflexes `Static Reflexes'. If a joint can rotate in two axes it will contain two such Static Reflexes, if it can rotate in three it will contain three.
A Dynism must obviously however be able to change its joint angles when necessary so that it can move its components in order to perform its necessary tasks like pursue Prey and avoid Predators.
To do so the Dynism continues to use Output Patterns basically like those it used back in Oceanic times. However, instead of controlling the Effectors directly, let's suppose they now `alter the settings' for the angular positions at which the Static Reflexes maintain the skeletal joints. These Reflexes themselves will then move the components to their new positions and hold them there so that the Dynism adopts a new configuration. Then, as we saw when we looked at Actions, when these Configs are executed with sufficient speed, the inertia and momentum of the components involved turn these `rigid' Configs into the smooth, even rhythmic motions required for such Actions as Walking, Swimming, Climbing, and Running.
Certain Configs cannot however be performed directly after others. For instance when the Dynism extends its front leg, resuming the `stand' Config when stimulated to do so poses no problems. However, should it instead be stimulated, by an internal sensor perhaps, to lie down from that Config, it will need to resume its Stand Config, then perhaps sit down using its hind legs, then lower itself the rest of the way using its front legs. This means that to move from the first Config to the last, it must pass through two `intermediate' Configs on the way.
Along with the Config Output Patterns therefore, the Dynism must also acquire in its ROM what amounts to a `Sequence Map'. This contains the sequences of Intermediate Configs it must perform in order to move from one `end' Config to the next. Here I've used square symbols to represent Start or End Configs, round ones to represent Intermediate Configs, and diamonds to represent Intermediate ones from which more than one Config can be reached. Most Configs, the Stand or Sit ones especially, will allow the Dynism to perform various minor Configs simultaneously such as reaching out from a Stance as we've already seen, or closing its Jaws to seize a Prey.
As you can see from the Map and from this illustrated extracted from it, Configs can form loops and branches just as they can in observed Actions. Most are unlikely to be executable in reverse as the arrows in the Route Map suggest. Inertia and momentum also have their price; such Actions will usually need to be entered and left via Preliminary Actions like these Go and Whoa ones for instance.
Here too `Speed Numbers' may be attached to an Action to control the speed at which it is executed. The faster an Action is performed however, the fewer the modifying `Overlay Actions' that can be executed simultaneously with it; running quickly may well only allow gentle turns for instance. In some cases, especially with very brief Actions, the force required to execute them may be more important than their actual completion, the Denarian equivalent of a kick in self-defense perhaps.
Eventually the Dynism will enclose itself in a single large cube made up of subcubes just as it did with its Predators and Prey. This cube too is fixed relative to the Dynism's thorax and moves and rotates with it.
While the Dynism can use this cube to store its own Output Config Patterns, it also comes to use it in another way which is just as important. It allows the Dynism to acquire its first crude `Targeting Reflexes', as we'll call them. When the Dynism perceives an object that it is stimulated to contact or grasp in some way, a Stone to step on perhaps, it can use the grid cubes of this new 3D `Target Map' to place the relevant body component, be it pincer or foot, directly into a cube intersecting the object's surface. To perform this reach-and-touch Action, it uses the basic Reach Out Config we saw earlier, but the Static Reflexes can now also carry out the actual `targeting' of the extremity. Indeed, any of the Dynism's Action Configs can now become `Transitive' in this way.
As before, a Dynism does not always need to be able to see a surface to touch it. It can in effect apply a `Hidden Half Assumption' of its own to objects it perceives to be cylindrical or spherical. This allows it to place a pincer behind such objects to probe for Prey perhaps or to grasp the object itself. To aid in the location of surfaces relative to its Target Map, the Dynism now acquires Stereoscopic Vision, if it hasn't already done so. Two eyes set some optimal distance apart insert their Stimulus Images into two Image Buffers. Their positions within these Buffers will differ from each other by a factor that depends on distance; The closer the object is, the greater this factor will be, the more distant the lesser. A Distance Tester can then use this factor to derive the object's distance and size.
Stereoscopic Vision now also helps slightly more advanced Dynisms to determine some aspects of a RAM Object's three-dimensional shape as well. For instance, it may now be able to determine whether a surface is `flat' enough to step on without slipping.
The Target Map is also very useful to the Dynism in another way. Let's imagine that the Dynism is walking from an area of flat ground towards ground that becomes increasingly broken until it becomes strewn with sizable rocks.
Once these Rocks exceed a certain minimum size, the Dynism substitutes a Stepping Action for the Walk Action. This is like a Walk Action except that each foot now has to be Targeted into a Subcube that is a compromise between where it would `normally' go, and one containing the nearest surface of such size and `flatness' that it can be safely stepped on. Such Stepping Actions can as a consequence usually only be performed slowly.The Dynism also uses the Target Map to ensure the Step Action places the Dynism's rear feet into exactly the same Safe Surface subcubes as its front feet when it moves forward one body length. To enable this to happen, it must use a sub-Mapper that ensure that the Target Map cube location for each Safe Surface moves back one subcube as the Dynism moves forward one. New Safe Surfaces thereby enter the cubes at the front of the Map while those at the rear drop out of it. If the Dynism turns or climbs up or down, new such Surfaces move into or out of the Map's corners according to the direction of the turn.
But the point will inevitably be reached where such Safe Surfaces are too far apart for Step Actions; the Dynism can progress no further.If however the broken surface is followed by a relatively flat one, the Dynism may be able to select one or another `Jump Action' which involves moving its entire body in a single Action. Since this `Safe Surface' will lie partly or completely outside the Target Map, the Dynism must also acquire a subtester which determines whether it is large enough to take all four of the Dynism's feet and contains enough empty space to accommodate its body. The Surface must also not be too far away horizontally or vertically, and there must be no intervening obstacles.
We will look at how the Dynism might come to make such `Decisions' later on, in the meantime we will assume it makes them entirely at random.
