Topic: Learning-based dynamical decision-making for multi-agent systems

Dynamical decision-making (DDM) for multi-agent systems (MASs) in the presence of uncertainty is central to the field of artificial intelligence as well as many other fields, i.e., localization, programming, control, and optimization. The primary aim of DDM for MASs is to allow interactive agents to choose local action with private observation and adjust coupled-strategy accordingly. This is accomplished by suitable problem formulation and rational mechanism. The challenges are to design a framework able to resolve conflicts and obtain strategy under the model uncertainty and interaction uncertainty among agents. The conflicts resolution can happen with the help of Game Theory (GT). GT is an optimization-based framework that uses utility influenced by multiple actions/strategies. Learning is an important branch of Artificial Intelligence that provides agents the ability to automatically learn and improve strategy from the interaction. Learning solution to action/strategy plays the role of decision in various uncertainties. These learning algorithms may be applied to dynamical decision-making for multi-agent systems.