|
2APL (A Practical Agent Programming Language) is a modular BDI-based programming language that supports the development of multi-agent systems. 2APL provides a rich set of programming constructs allowing direct implementation of concepts such as beliefs, declarative goals, actions, plans, events, and reasoning rules. The reasoning rules allow run-time selection and generation of plans based on declarative goals, received events and messages, and failed plans. 2APL can be used to implement muti-agent systems consisting of software agents with reactive as well as pro-active behaviours. == Overview == 2APL provides programming constructs to specify both multi-agent systems and individual agents. Multi-agent systems are specified in terms of individual agents and the environments with which they interact. Individual agents are specified in terms of the following ingredients. * Beliefs: It implements an agent's initial information about its environments and other agents with which it interacts. An agent's beliefs may change during its execution. * Goals: It implements an agent's initial objectives. Each objective denote a state the agent desires to achieve. A goal will be removed as soon as it is achieved. Different goals may not be achievable at the same time as they may denote conflicting states. * Basic Actions: 2APL provides different types of actions, among which, * * ''Belief Update Action'' is to update an agent's beliefs. * * ''Communication Action'' is to pass a message to another agent. * * ''External Action'' is to interact with an environment. * * ''Abstract Action'' is to encapsulation a plan by a single action. * * ''Belief Test Action'' is to query an agent’s beliefs. * * ''Goal Test Action'' is to query an agent's goals. * * ''Adopt Goal Action'' is to add a goal to an agent’s goals. * * ''Drop Goal Action'' is to remove a goal from an agent’s goals. * Plans: A plan consists of basic actions composed by operators such as sequence, conditional choice, conditional iteration, and a unary operator to identify (region of) plans that should be executed atomically, i.e., the actions should not be interleaved with the actions of other plans of the agent. * Reasoning Rules: Three types of (practical) reasoning rules are provided to implement the generation of plans. The rules have a belief condition indicating when the rule can be applied. * * ''Planning Goal Rule'' is to generate a plan to achieve a goal. * * ''Procedural Rule'' is to generate a plan to react to either an event (received from environment) or a message (received from an agent). This rule can also be used to relate an abstract action to the plan it encapsulate. * * ''Plan Repair Rule'' is to generate a plan to replace a failed plan. * Modules: A 2APL agent’s program can be developed in separate modules. Each module encapsulates cognitive components such as beliefs, goals, plans, and reasoning rules. In practice, a 2APL module can be used to program a specific functionality, such as a ''role'' or an ''agent profile''. A programmer can perform a wide range of operations on modules, e.g., ''creating'' a module instance, ''updating'' it, ''executing'' it, and ''testing'' its state. * Environments: A 2APL environment can be implemented as a Java object. The methods of such an object correspond to agents' external actions. The body of a method implements the effect of the corresponding action. 抄文引用元・出典: フリー百科事典『 ウィキペディア（Wikipedia）』 ■ウィキペディアで「2APL」の詳細全文を読む スポンサード リンク
|