Abstract
A theory of active agents is presented which aims to simulate dynamical complex networks. As complex systems involve many-to-many interactions, the ability to predict the evolution of the system is greatly affected by the ability to take into account the simultaneity of nodes interactions. An active agent in a discrete system is an entity capable of activity. The state of an active agent at a certain time point is defined by its activity and its energy, which is the propensity of the agent to perform its activity. The state vectors is a data structure whose instances are context-specific, i.e. they can be implemented by adding any feature deemed useful to describe the dynamical system. The update of states of the active agents constitutes an evolution of the system, that is, a discrete dynamical system. The relation among entities are represented in a multilayer graph that efficiently describes systems interconnected through different categories of connections. The update is determined by the relations of the entities, and by the activity and energy state of the active agents. The states of the entities are updated in a synchronous manner. Active-agents modelling framework is a model-free approach, which can be used when knowledge of physical laws is scarce or partial, when the system to be simulated is complex in terms of size and density of interactions and can be parallelised without the need to implement a race condition.