Delivered for the IEEE Africon2011 13.-15.Sept.2011 Conference, Track03: 'Computational Intelligence and Computational Semiotics“ by G.Doeben-Henisch, G.Abrami, M.Pfaff, M.Struwe
This text describes the simulator software, which is used for the above cited paper.
The simulation-software, called APS-Simulator is construct to evaluate and test the theoretical concept (theory).
The simulator based on the theoretical-concept, which is explained before. First of all the simulator is split in two parts. First part is the pure implementation of the simulator-concept and comprised the algorithms to simulate the content of our research. As we see in the paper for the conference, we illustrate our theory with results of many simulations done by the APS-Simulator.
The second part of the APS-Simulator is the visualization. The visualization is based on OpenGL and shows the current simulation-process.
The structure of the APS-Simulator is based on the theory of conscious based learning agents and implements these completely. Of course there are some classes to build a framework for the simulation to operate. For example the GUI (Graphical User Interface) or the I/O - Methods for the file system, et cetera.
The Simulator is written in C++ and contains Q-Classes because we develop the APS-Simulator with the development-environment QTCreator from Nokia.
The environment is the “world” where our agents live. The environment is a field of “Objects” or “Non-Objects” with a size of N x M fields. The agents can interact with the environment and vice versa. The environment contains every object (food, block) and the agents (Agent0, Agent1, Agent2, AgentN, .. AgentN+1). Every field in the N x M - Matrix wich contains no object or no agent are “empty” and therefore “free fields” that can be enter by a agent.
An agent is the implemented theoretically concept of a agent-structure (Agent0, Agent1, … AgentN, AgentN+1) and contains many attributes and methods to operate. An agent can interact with the environment and vice versa. An agent can eat on a “Food-Field” if its drive to “eat something” is active. The structure of the agents have been described in the Paper.
The mapper is the interface between the agent and the environment. With the mapper we can translate an input from the environment to the agent and an action from the agent to the environment. Via the mapper we can implement the structure and the concept of every agent on every stage. The mapper translate the Information from the environment and the agents to each other. Furthermore the Mapper-Class is our interface to demonstrate the simulation-process with our visualization.