Software Programmed by Artificial Agents toward an Autonomous Development Process for Code Generation

Current automated software developments mostly deal with automatic code generation. This approach usually involves the software synthesis from design models, and pre-defined policies and fixed rules. This paper proposes an approach that goes beyond the above software development automation by presenting a self-directed process that is able to make decisions to develop software. An ontology-enabled agent plays the role of human developer by performing software development activities autonomously. The knowledge captured in an ontological database enables high-level reasoning to interpret, design and synthesize the system logic based on the physical system configuration. The methodology is implemented in a graphic computer tool. A case study of control software development for a distributed fuel management system for aircraft is presented. Diverse fuel systems can be easily described and the code can be generated from a visual layout. This paper describes details of the autonomous ontology-based methodology to generate software code. Some experimental results are also shown.