Design and compilation of an object-oriented macroprogramming language for wireless sensor networks

• Published in: 39th Annual IEEE Conference on Local Computer Networks Workshops pp. 574 - 582
• Main Authors: Oppermann, Felix Jonathan, Romer, Kay, Mottola, Luca, Picco, Gian Pietro, Gaglione, Andrea
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.09.2014
• Subjects: Compilers; Instruction sets; Java; Object oriented modeling; object-oriented languages; Peer-to-peer computing; Programming; Resource management; Wireless sensor networks
• DOI: 10.1109/LCNW.2014.6927705

Wireless sensor network (WSN) programming is still largely performed by experts in a node-centric way using low-level languages such as C. Although numerous higher-level abstractions exist, each simplifying a specific aspect of distributed programming, real applications often require to combine multiple abstractions into a single program. Using current programming frameworks, this represents a difficult task. In previous work, we therefore defined a conceptual framework that facilitates abstraction composition by defining sound compositional rules among few fundamental abstraction categories. The framework is extensible: programmers can add new abstractions within the boundaries determined by the compositional rules. In this paper we describe the design of a language-called MPL-that instantiates this conceptual framework. To support the extensible nature of the framework, the language is object-oriented, which allows programmers to add new abstractions by inheriting from existing classes that implement predefined interfaces. We modeled the syntax after Java, to make it more palatable to inexperienced embedded programmers. Compared to Java, we modified the language to enable efficient execution on WSN devices. We designed and implemented a compiler that translates MPL language into executable C code, which spares the overhead of a virtual machine. By comparing MPL implementations against functionally-equivalent Contiki/C implementations of several benchmark applications, we determined that the performance overhead of MPL is limited, and yet the programming task is simplified.