Example of simple FCM

A very simple map can be defined to model an agent perceiving and reacting to its distance from a foe. The closer the foe, the more frightened the agent. Depending on this distance and also on the fear level, the agent will decide whether or not to evade. The more frightened the agent, the faster the evasion. An FCM corresponding to this example is given in the next Figure. In this example there are two sensitive concepts (foeClose and foeFar), one internal (fear), and one motor (evasion). There are also three influence edges: closeness to a foe excites fear, distance to a foe inhibits fear, and fear causes evasion. Activations of the concepts foeClose and foeFar are computed by fuzzification of the real value of the distance to the foe, and the defuzzification of the activation of evasion tells us about the speed of the evasion. With the FCM model it is possible to distinguish the perception from the sensation: The sensation is the real value coming from the environment, and the perception is the sensation modified by the internal states. For example, it is possible to add three edges to the previous map: one autoexcitatory edge from the concept fear to itself, one excitatory edge from fear to foeClose, and one inhibitory edge from fear to foeFar. A given real distance to the foe seems higher or lower to the agent depending on the activation level of fear. Also, the fact that the agent is frightened at time t influences the level of fear of the agent at time t + 1. This kind of mechanism gives the possibility of modeling a degree of paranoia and a degree of stress for the agent. It also allows the agent to memorize information from previous time steps: fear maintains fear. It is therefore possible to build very complex dynamic systems involving feedback and memory using an FCM, which is what is needed to model complex behaviors and abilities to learn from evolution. (The example is coming from : Tisseau, J. (2001). Réalité virtuelle—Autonomie in virtuo. Unpublished habilitation a diriger les recherches dissertation, University of Rennes, France.)