Abstract
According to the importance of artificial life in simulating natural processes, we have modeled an ecosystem as a virtual world of predating events. In this regard, on the basis of emergence of symbolic communication, we have modeled assigning a sign to an object by an autonomous agent in the environment as a symbol, which is the relation of symbol to object. In the next step, we have modeled symbols composition, which is the relation of one symbol to the other ones or among symbols themselves, i.e. language grammar or syntax. So, we have designed an artificial prey-predator ecosystem to have reinforcement learning of a symbolic language in preys, so that it can generate an appropriate escaping behavior after hearing the symbol of each predator. Finally, we will have a self-organizing system in which symbolic communication will emerge a language of symbols without any external knowledge or central control. We have used a neural network for reinforcement learning in the first stage, and the other neural network for the second stage.
