Issues in designing a compilation framework for macroprogrammed networked sensor systems
InterSense '06 Proceedings of the first international conference on Integrated internet ad hoc and sensor networks
| Hi-index | 0.00 |
The proliferation of networked embedded devices such as wireless sensor nodes ushers in an entirely new class of computing platforms, which unlike traditional systems, interact with the real world. The focus of sensor research till now has been on design of hardware and networking protocols to enable remote sensing and data collection. The next challenge in sensor system design is to exploit the in-network computing capabilities to make them intelligent and capable of autonomous operation. The focus of this thesis is on investigating models and algorithms for facilitating distributed computation in wireless sensor systems. A model of computation is defined to abstract the class of dense, uniformly deployed, wirelessly-connected, localized, two-dimensional, static sensor systems. The model is used for design and analysis of algorithms for solving several non-trivial computation problems in sensor systems. The benefits of our approach are demonstrated by end-to-end design of a sensor system, which exploits the computation and storage capabilities of the sensor nodes for efficient routing and resolution of topographic queries. We define the cluster-based, heterogeneous model for wireless sensor systems (COSMOS), which abstracts the network as an asynchronous, distributed system consisting of a set of uniquely-identifiable nodes, organized in a clustered mesh topology. It models several features specific to sensor systems, not abstracted by the traditional computation models, such as the variable radio range and power management of the sensor nodes. Time and energy optimal, and energy-balanced algorithms are discussed for sorting and selection in single-hop sensor systems---for single-channel and multi-channel networks. We also present time efficient and energy optimal algorithms for sorting and summing in multi-hop sensor systems. We discuss energy-efficient and fault-tolerant resolution of topographic queries in sensor systems. Our approach is based on prior construction of the topographic map of features of interest in the network. Once constructed, the map is used for resolving a large number of queries efficiently. We show that our approach is time and energy efficient, and recovers more reliably from node failures in the network in comparison to the state-of-the-art techniques.