Sensomax: An Agent-Based Middleware For Decentralized Dynamic Data-Gathering In Wireless Sensor Networks

In this paper we describe the design and implementation of Sensomax, a novel agent-based middleware for wireless sensor networks (WSNs), which is written in Java and runs on networks of various Java-enabled embedded systems ranging from resource-constrained Sun Spot nodes to resource-rich Raspberry Pi boards. Programming WSNs tends to be a complex task for developers, as it requires detailed knowledge of underlying hardware resources as well as their firmware or operating systems. Although many solutions have been proposed up to this date, only a few of those are capable of satisfying challenging demands such as serving multiple user applications and reprogramming the network at run-time in popular high-level languages such as Java. Sensomax presents a novel combination of several best practices from existing solutions, facilitating fully distributed and decentralized bulk programming and/or updating of sensor nodes; serving multiple simultaneous applications deployed by single or multiple users; allowing dynamic run-time changes in the application requirements; and offering on-the-fly switching between time-driven, data-driven, and event-driven operational paradigms. Sensomax provides a sophisticated set of APIs, a feature-rich desktop application, a web application for cloud-based distributed networks, and a simulator. We demonstrate Sensomax in operation on a real network of 12 Sun Spots deployed as an environment-monitoring system, and 600 virtual Sun Spot nodes running continuously over periods of several weeks, using a novel statistically rigorous adaptive change-point detection algorithm to identify significant "anomalous" changes in the monitored data-streams.