A Compiler Design for the Agent-Based Programming Language

In this paper, we introduce a prototype compiler for the Agentbased Programming Language (APL) using the Belief-DesireIntention (BDI) model. The compiler we developed can translate an agent-based program written in APL into the Java byte code, which is executable on the Java Virtual Machine (JVM). This paper proposes a simple syntax for Agent-based Programming Language (APL). APL is designed based on the BDI model, which is a natural way to describe the behaviors of an intelligent agent. By using the associated database and reflection concepts, the APL can use different intentions intelligently and dynamically. This work shows how nicely and naturally the APL programming language can represent intelligent software agents and how we can translate the APL program into the Java byte code. In this work, we use the Belief-Plan mapping table and Java reflection feature to implement dynamic linking which makes it possible for APL to link to different code fragments at runtime without recompiling the APL source code. This feature of APL makes the language unique and it may change the way of programming.