Agent-oriented programming

Agent-oriented programming is a style of programming that is based on the theory of agent-oriented paradigms of Yoav Shoham and was mainly used in the declarative programming language PROLOG.

1993 Shoham has published agent-oriented paradigms that are based on so-called " mental states " (mental states ), which is internal, that is invisible from the outside condition of software agents meant. A formal language describes the " mental states ", which are determined by assumptions ( beliefs ) and obligations ( commitments ).

  • Actions ( actions ) are described, for example, REQUEST, INFORM.
  • Beliefs are assumptions that the agent has about its environment, so all the data available to him.
  • Commitments are promises which are the agent of its environment ( other agents ). They state that if the agent receives a particular message while it is in a certain mental state, it performs a certain action.
  • Conventional programs are wrapped in a wrapper, so that they can be regarded as agents. Shoham speaks of " Agentifizierung " ( agentification ).
  • It is used a time point-based temporal logic.

Languages ​​for agent-oriented programming

PLACA is a development of more theoretical programming language AGENT -0. PLACA was developed at Stanford University in California by Sarah Rebecca Thomas. Unlike AGENT- 0 PLACA has the advantage that not only requests from other agents answered and inquiries can be addressed to that, but in PLACA also direct mechanisms are in place to track a goal and several actions to perform to achieve a goal to reach. Each agent in PLACA has an ordered list of intentions ( goals ) and an ordered list of plans. To manage these lists PLACA has the following syntactic structure:

  • ( INTEND x) - The PLACA Agent tries to make the statement true y. Therefore, x is added to the list of intentions.
  • ( ADOPT x) - Adds the intention or plan to list the intentions and plans added.
  • (Drop x) - Deletes the intentions of x and x the plan from the list of intentions or plans.
  • (CAN -DO x) - The agent reports that it can perform the action x.

CAROL5 was implemented in C to implement group learning systems ( social learning systems).

Temporal logic of actions
34123
de