There is no direct mapping from into the framework because Wilson does not offer a complete structural definition of the . Instead he presents a diagram Fig.1 on page 2 of his paper [347] and in the accompanying text he explains the different components of the diagram in an informal way. To explain our embedding we have to show here a copy of his diagram with additional markings added from the author of this text.
Figure 4.5:
Structural comparison between zcs and EF framework
|
Figure 4.5 shows that all the main components of the framework can be identified in the drawing of Wilson.
- Environment: Wilson assumes an environment which is the source of binary strings encoding the sensory inputs for the zcs. These binary encodings are the result of a translation process from certain kinds of 'objects' in his wood1-world (for a more detailed description of this translation process see the section 4.4). The environment is also the destination for the actions of the zcs.
- Interface: Wilson does not explicitly talk about an interface between the environment and the agent . But he distinguishes detectors -labeled
in the framework-, which receive input from the environment as well as effectors -labeled in the environment-, which send action signals to the environment. These assumptions imply some mappings from the environment to the detectors as well as from the effectors to the environment. Furthermore does Wilson assume explicit rewards from the environment to the agent. In the framework these mappings are explicitly named by ainp(), aout(), fit(). They together constitute the interface .
- Agent: The agent according to Wilson is a structure containing some sets including the detector and effector values and some mappings. In the framework a general system function is assumed which maps the input to the output. The details of this system function are left open for different kinds of differentiations.
Thus while the general structure from zcs maps well to the framework are the details of the zcs-system function special. We will reconstruct this function below. For this we work out the theoretical definitions and simultaneously a software version. In some cases we will use the details of the software functions as defining operations for the theoretical mappings as well.
Gerd Doeben-Henisch
2012-03-31