Four Decision Algorithms for Cognitive Agens Learning How to Cross an Intersection of Minor and Major Roads and Analysis of Their Performance

Cognitive agents can be used to simulate autonomous robots, which have the ability of perceiving, reasoning, judging and learning information and then reacting to this information and to the environment. In this thesis a model of simple cognitive agents learning to cross an intersection of minor and major roads is introduced. In the model the major and minor roads are modeled by cellular automata. The cars on the major road move in accordance with the modified rules of the Nagel-Schreckenberg model and the cognitive agents on the minor road learn to cross the major road using an observational social learning process. The thesis introduces four new decision algorithms for the cognitive agents to learn to cross the intersection and investigates cognitive agents’ learning performance for these algorithms and for the realistic values of the car traffic parameters. The thesis investigates the effects of the model parameters on the cognitive agents’ learning performance using various statistical performance indicators. Also, the thesis compares the performance of the four new decision algorithms with the performance of the previous decision formula used by the cognitive agents. The results obtained from the statistical analysis indicate that the four new decision algorithms have distinctively different performances for the different selection of the values of the model parameters.