| Recent advances in embedded technology have made possible the deployment of wireless networks composed of hundreds or thousands of energy efficient sensors and actuators equipped with their own CPU, memory, and wireless radio. These wireless networks of sensors and actuators (WNSA) have found applications in many domains among which building automation, personal health care, strategic military assistance, and environmental surveillance. Despite the large number of WNSA applications, the number of actual WNSA deployments is limited by several factors including programming difficulties and the low reliability of resulting applications. Many approaches have been proposed to ease the programming of WNSAs: macro-programming languages, toolkits, virtual machines, and programming abstractions. While these approaches introduce valuable programming features and architectural organization models, their scope is typically limited to a single WNSA, composed of identical nodes, all managed by a single individual that runs a single application at a time.;In this research, we provide a framework, named Physicalnet, which assumes nothing of the sort. Physicalnet applications can simultaneously use the same sensors and actuators; they can seamlessly access the services implemented on a heterogeneous set of nodes; and they can involve nodes from multiple remote WNSAs. Physicalnet nodes are free to move within a network and across networks; and each node can have a different owner that can specify fine-grained and complex access rights involving environmental measurements, node locations, or any other application variables. Finally, Physicalnet also addresses canonical WNSA requisites: ease of programming, scalability, robustness, low application launch time, low application response time, energy conservation, and low memory usage.;Physicalnet is implemented on resource constrained MICAz wireless sensor nodes, and personal computers. Using a combination of simulation, testbed, and numerical evaluation, we estimate the following quantitative metrics for multiple applications and multiple deployment scenarios involving mobile nodes, multiple networks, and concurrent applications: number of lines of code (used an indicator of conciseness), memory usage, energy consumption, application launch time, application response time, and percentage of lost sensor data. We compare Physicalnet with the related work using several of these metrics and conclude on the practicality of Physicalnet and its potential impact. |
