Agent-based modeling and simulation - Modeling and Simulation of Complex Systems - page 37

Two drawbacks are identified: First, although the simulation model
itself is formalized (via XML Schema), the execution of a simulation is
defined in a textual way only [137, 107]. Second, the AORS approach
assumes a simulation execution with fixed time increments.
2.3.6 Further related work
De Vries et al. (2001)
De Vries et al. present a model of agents with a flexible cycle of
sensing, reasoning and acting. They explicitly deal with actions with
duration and true concurrency , which means that concurrent actions
are treated as happening simultaneously [30].
Right at the beginning, De Vries et al. state that though 'interaction
is a key concept for agents, these issues are not properly treated in
many agent programming languages' [30]. In the following they
state furthermore that 'typically, actions are atomic and take zero
time, events from the environment aren't explicitly modelled and
interference or synergy of actions is impossible' [30]. Their model
aims to address these issues.
They notice that most agents perform some kind of sense-reason-act
cycle . Therefore they propose a flexible execution cycle consisting of
reasoning and interaction, with observation being just a special kind
of interaction. The only requirement imposed on this cycle is that in
every iteration the reasoning precedes the interaction. Furthermore,
they point out that the agents have to observe their environment
explicitly and that their perception can differ from the true state of
the environment.
In order to allow actions with duration, actions are composed of
atomic sub-actions consuming one time unit each. Based on this
approach, parallel execution is defined for various cases (parallel
internal actions, parallel external actions, parallel internal and external
actions). In this way, De Vries et al. completely abstract from the
inner functioning of the agents and focus on concurrent execution.