Multi-Agent System Research Map

Environment

An agent can interact with its environment in several different ways. It observes its environment with its sensors and modifies it with the help of its actuators. Additionally an agent has the possibility to move inside its environment. These kinds of interaction vary depending on considering a software agent or a robot that resides in a "real" environment.

Following Russel and Norvig [RN03] the environemnt can be classified by these properties:
accessible vs. inaccessible Accessible environemts allow the agent to gain complete information of the world state. Most real-world environments are not accessible in that sense.
deterministic vs. non-deterministic If the environment is deterministic an agents action leads to a guaranteed effect and predestined world state.
static vs. dynamic A static environement does not change by itself, but only by the agents actions.
discrete vs. continuous An environement is discrete when a fixed number of actions and percepts exist in it.

Depending on the kind of agent - eg. a software agent vs. a robot - and the problem domain the environment exhibits different properties and may additionally require quick or real time behaviour from the agent. Considering a software agent two different kinds of movements can be observed. On the on hand the agent can change its location through code migration and on the other hand the agent can be located on a mobile device that is moved through its environment.

[RN03] S. Russel, P. Norvig. Artificial Intelligence: A Modern Approach, Second Edition, Prentice Hall, 2003.