In this work the authors propose a distributed algorithm for building approximate Minimum Spanning Trees (MSTs) for energy-efficient gathering and aggregation of sensed data in randomly-distributed, multi-hops, wireless sensor networks (WSNs). The MST created, obtained through global organization based on local interactions among each sensor and its neighbors, exploits an innovative joining mechanism that guarantees, at the same time, the creation of a near-optimal energy-efficient MST and the avoidance of cycles. The led simulations have demonstrated that the proposed solution gives, in terms of expected quality of built trees, better performance then the UDG-Nearest Neighbor Algorithm (UDG-NNT) that is, currently, the best known solution for building approximate, energy-efficient, MSTs in multi-hop WSNs.
I. INTRODUCTIONWireless Sensor Networks (WSN) are used in a lot of different applicative scenarios, including environmental, industry and military monitoring, object movement tracking, both for safety (floods or fire prevention) or security (intrusion detection) reasons. Independently from the considered applicative scenario, WSNs have the same scope: they sense the covered area in order to acquire data and delivery them to base stations, where data are elaborated and made available for external users. The performance of each sensors network, in terms of quality (precision, tolerance) of sensed data, delivery time, fault tolerance, robustness and energy consuming are influenced by very different aspects. As well as the hardware features of sensors, in terms of electronic circuits and radio technology used, the overall behavior of WSN is affected by the requirements of specific applicative scenario, that could imply synchronous and asynchronous events-driven monitoring, continuous data stream acquisition, and the quantity of sensors deployment in the sensed area, their distance and the interaction among them. Surely, one of the challenges of the WSN research is to find an effective strategy for the power management of sensors. Indeed, even if the low cost and small size of the sensors have allowed their ubiquitous distribution and a great flexibility in their use, on the other hand, they have implied stringent limits in terms of computational capabilities and, above all, of energy available. The power management is