Design and evaluation of a multiagent interaction protocol generating behaviours with different levels of complexity

The design of a multiagent system based on simple interaction rules is presented, which can generate different types of overall behaviours, from asymptotically stable to chaotic, verified by two tests for chaos. Exogenous perturbations are analysed, showing that very small changes can have a great impact on the evolution of the system. Some methods of controlling such perturbations in order to have a desirable final state are investigated. Also, endogenous perturbations and the effect of alternative decisions on the evolution of agent utilities are examined. Different methods are suggested for describing the behaviour of the multiagent system.