Abstract
Agent-Oriented Programming (AOP) is a relatively new programming paradigm, proposed by Yoav Shoham, which views software systems as consisting of a set of agents that interact with one another to solve problems beyond their individual capabilities. Since the inception of the paradigm, a number of AOP languages have been proposed. This chapter focuses on one such language, the Agent Factory Agent Programming Language (AFAPL), a practical rule-based language that has been applied to a wide range of problem domains including robotics, virtual and mixed reality environments, and mobile computing. AFAPL is placed in context through a general introduction to the state-of-the-art in AOP. The chapter finishes with a discussion of some future trends for AOP and some concluding remarks.
TopIntroduction
Agent-Oriented Programming (AOP) is a relatively new programming paradigm introduced by Yoav Shoham (1993) in which software systems are viewed as consisting of a set of agents that interact with one another to solve problems that are beyond their individual capabilities. More specifically, agents are viewed as high-level autonomous software entities that encapsulate a set of capabilities and whose internal state is comprised of a set of mental components such as beliefs, capabilities, choices and commitments. This view of agents as mentalistic entities is a common perspective within multi-agent systems research and underpins many of the most prominent agent theories (Cohen & Levesque, 1990; Rao & Georgeff, 1991; Wooldridge, 2000). These theories model the internal decision-making process of an agent in terms of the interplay between the constituent components of the underlying mental state. Their objective is to define how an agent is able to act in a rational goal-directed manner and to tease out various desirable properties that emerge from that action. Thus, the objective of AOP is to present a framework for developing a new class of programming languages that are derived from these theories.
This view of programs that consist of components whose state is defined by a set of mental qualities and in which computation is realized through speech act based message passing is not unique to Shoham. John McCarthy, who is widely acknowledged as being one of the first to associate mental qualities with machines (McCarthy, 1979) had, as early as 1990, written an draft proposal for a programming language he entitled Elephant 2000 (McCarthy, 1992). That said, McCarthy recognized at the time that his language could be implemented, and to this day, no implementation of Elephant 2000 exists (McCarthy, 2007). In contrast, a number of AOP languages have been proposed, implemented and successfully used to build a range of agent-oriented applications.
This chapter focuses on one such AOP language, entitled the Agent Factory Agent Programming Language (AFAPL) (Collier, 2001; Ross & Collier & O’Hare, 2004; Collier & Ross & O’Hare, 2005). AFAPL has its origins as part of a larger software engineering framework entitled Agent Factory (O’Hare, 1996; Collier, 1996; O’Hare & Collier & Conlon & Abbas, 1998; Collier, 2001; Collier & O’Hare & Lowen & Rooney, 2003). Both AFAPL and the associated framework have been designed specifically to support the fabrication of agent-oriented applications, and have been used extensively in the development of a number of prototype systems in areas such as: robotics (O’Hare & Duffy & Collier & Rooney & O’Donoghue, 1999; Dragone & Holz & O’Hare, 2007), mobile computing (O’Hare & O’Grady, 2003; Muldoon & O’Hare & Phelan & Strahan & Collier, 2003), virtual and mixed reality (Duffy & O’Hare & Campbell & Stafford & O’Grady, 2005), wireless sensor networks (Marsh & Tynan & O’Kane & O’Hare, 2004; O’Hare & O’Grady & Marsh & Ruzzelli & Tynan, 2006) and information retrieval (Peng & Collier & Mur & Lillis & Toolan & Dunnion, 2004; Lillis & Collier & Toolan & Dunnion, 2007).
The remainder of this chapter starts with an overview of Agent-Oriented Programming. This is followed by a general overview of the Agent Factory framework. The next section presents details of AFAPL, and after that some of the tools that have been developed to support its use are discussed. The penultimate section, discusses future trends in this area, and finally, some concluding remarks are presented.
Key Terms in this Chapter
Platform Service: A shared resource that can be used by all agents residing on the given platform (assuming access to that resource has not been restricted).
Intentional Agent: An agent whose reasoning process is modeled as the interplay between a set of mental attitudes - such as beliefs, desires, intentions, obligations, or commitments - that together comprise the agents’ mental state.
Agent Communication Language: A language that formalizes how agents may interact with one another and which is typically based on Speech Act Theory.
Agent-Oriented Programming: A programming paradigm in which agents are viewed as fundamental computational units, whose state consists of a set of mental attitudes, and which interact by passing messages using some Agent Communication Language.
Agent Platform: The underlying apparatus that is provided to support the deployment and management of agents.
Message Transport Service: A type of platform service that provides support for inter-agent communication via some form of underlying network protocol (e.g. HTTP, IIOP, XMPP).
Commitment: An agreement that is made by an agent to do something at a given point in time, for a given agent (possibly itself), and which may in some cases only be maintained only so long as an associated set of conditions remains valid.
AFAPL2: An Agent-Oriented Programming language that is part of the Agent Factory framework.
Agent: An autonomous software and/or hardware entity that is situated in some environment, and which is capable of both reacting to changes in that environment and taking the initiative.
Belief: A fact about the state of the agents’ environment (this may include the agent itself).
Belief-Desire-Intention (BDI): A well-established mental state model in which beliefs represent the current state of environment, desires represent the ideal state of the environment, and intentions represent the subset of the agents’ desires that it is committed to bringing about.
Commitment Rule: A core programming construct that defines situations (conjunctions of positive and negative beliefs) in which the agent should adopt a commitment.
Actuator: An effecting unit that agents can use to manipulate their environment.
Module: A private resource that can only be utilized by the agent that it is associated with.
Perceptor: A sensing unit that agents can use to generate beliefs about the current state of their environment.