Ramses

AbstractWireless Sensor Networks (WSNs) consist of networks composed of tiny devices equipped with sensing, processing, storage, and wireless communication capabilities. WSN nodes have limited computing resources and are usually powered by batteries. First generations of WSNs were designed to attend requirements of a unique target application usually with a single user, who was also the infrastructure owner. However, the rapid evolution in this area and the increasing of the complexity of the sensors and applications pose new challenges to WSN solutions, which can be addressed by specific middleware platforms for these networks. Existing middleware systems provide suitable mechanisms to define the high-level application logic and to deal with heterogeneity and distribution issues of WSN, but most of them do not provide explicit mechanisms to define the underlying autonomic behavior of the system, an essential feature of this kind of network. In this perspective, Autonomic Computing (AC) appears as a promising option to meet autonomic requirements in WSN middleware design. This paper presents the consolidated specification of RAMSES, a reference architecture of a self-adaptive middleware for WSNs. RAMSES was conceived in light of a well-stablished Reference Architecture Model, the RAModel. It follows the autonomic computing model MAPE-K, and presents a mapping of AC conceptual model to a set of software components. We claim that, with the aid of a middleware that supports the autonomic computing principles, a WSN becomes an autonomous WSN by design. RAMSES realizes our vision by providing: (i) an architectural template with core aspects of the self-adaptive systems from which is possible to build concrete middleware instances for self-adaptive WSNs, and (ii) a specification of the reference architecture using a formal architecture description language (Pi-ADL), which enables the representation of dynamic software architectures as required by WSNs. A scenario-based qualitative analysis and a checklist survey conducted with experts demonstrate the effectiveness of RAMSES. Moreover, a concrete WSN middleware instance derived from RAMSES was implemented as a proof of concept.