Effects of Simulation Parameters on Na?ve Creatures Learning to Safely Cross a Highway on Bimodal Threshold Nature of Success

摘要

A model of simulated cognitive agents (na?ve creatures) learning to safely cross a cellular automaton based highway is described. These creatures have ability to learn from each other. We investigate how the creatures’ learning outcomes are affected by the model parameters (e.g., the traffic density, creatures’ ability to change a crossing point, creatures’ fear and desire). We observe and study a bimodal nature in the number of successful creatures in various creature populations at simulation end. This value is either low or high depending on the values of the model parameters.