Information Technology Reference
In-Depth Information
1. Constraints as limiting (environmental) conditions
Constraints are used to limit the possibility of an action to be
executed successfully. Constraints of this type will most probably
involve more than one action and define under which conditions
two (or more) actions of different agents are influencing each other.
Used with this intention, constraints are clearly an essential part
of a model.
2. Constraints as assertions
Constraints can also be used to assert specific properties. In
this case, constraints might define which preconditions have to be
fulfilled for execution of specific actions or, more generally, which
conditions have to be fulfilled during the entire execution time of
any action.
Depending on the specific constraint, the effect or consequence of an
unfulfilled (or violated) constraint may differ. Considering constraints
as limiting conditions, a violated constraint results in the creation of
a specific event. This event in turn may directly alter the involved
agents (e.g., limited mobility after a collision) and eventually the
corresponding location within the environment. In case of constraints
as assertions, which will most likely be used during development
or verification of a model, the possible response to an unfulfilled
constraint ranges from simply creating an entry in an error log to
aborting the current simulation (
fail-fast
behavior [119]).
The basic assumption is that no constraints are defined at all.
Therefore, all sensor and effector actions of an agent are always
executed successfully and no two actions interfere with each other.
Nevertheless, constraints are an integral part of an agent-based model
and may help a lot to increase overall quality of the model. This is
mainly due to two reasons:
First, constraints are an appropriate and elegant way to make
implicit assumptions explicit. Besides documenting assertions as
well as pre- and post-conditions for specific actions, constraints
allow to explicitly express these assumptions. Furthermore, this can
