Self-replicating and self-modifying programs in fraglets

The inherently decentralized nature of artificial chemical computing models makes them particularly attractive for building bio-inspired software with self-organizing and emergent properties. Yet it is not straightforward to construct such chemical programs, either manually or automatically. We are exploring the potential of chemical programming models for automatic programming, in the context of autonomic environments where software must operate unsupervised for unlimited periods of time. We are enhancing the Fraglets chemical language to support intrinsic genetic programming, such that programs can replicate and modify themselves during execution. The Fraglets language was originally designed to express communication protocols. We first show a few extensions towards more generic computations, then show how self- replicating and self-modifying programs can be created. This is a first step towards programs that can repair and optimize themselves in an autonomic way. We reveal a number of features and shortcomings of the language, suggesting fixes and future directions.