Contributed editorial appearing in
Scientific Computing 23:5, April 2006, pg. 12.
Butter, sugar, eggs, vanilla, flour, baking soda and salt may at first glance appear to be standard items on a typical shopping list. Include the final item of semisweet chocolate morsels and you may recognize this collection as the constituents required to make chocolate chip cookies. But having the correct ingredients is not enough to obtain the final product – we also need the correct recipe. Each component must be mixed with the others in a specific order and sufficiently heated to enable the endothermic chemical reactions that convert this collection of individual parts into a coherent whole. The taste of warm cookies chased down by a cold glass of milk is an "emergent system property" achieved only when the process is executed correctly. Plugging the most exquisite ingredients into a poor recipe often morphs into a recipe for failure. "Playing well together as a team" is the goal of every successful coach, director, choreographer and system designer.Our ever-advancing increase in technological capability has enabled a dramatic decrease in the size and power requirements of detectors, transducers and sensors. Traditionally, field samples are collected in accordance with rigorous procedures and are transported to a central laboratory for analysis by cumbersome instruments. This permits the monitoring of multiple locations by a single device at the expense of long delay time between repeat analyses. Mobile, field-deployable system can shrink this delay by moving among sampling locations. Multiple stationary field sensors can be installed at specific sites and monitored continuously, but are often limited by the power and communication needs of the units. A natural next step in this technology is the hybridization of mobile and stationed sensors. The target hybrid would consist of multiple sensors providing continuous monitoring of various locations, while retaining the ability to deploy and reconfigure the positions without having to provide power and communication lines. These features describe those of a "wireless sensor network" (WSN).
The ingredients for a WSN include the proper sensors, miniaturized control and logic circuits, adequate power supplies, rugged enclosures and wireless commmunication transceivers. Not unlike the cookie problem, simply developing the proper wireless sensors does not address the issue of networking. A large amount of effort and ingenuity must be expended for the distributed sensor network to function as a coherent whole. Professor Anish Arora and this research group in the Department of Computer and Information Science at The Ohio State University have focused on the development of dependable distributed and networked systems.
Among the group's many projects enjoying support from the U.S. Defense Advanced Research Projects Agency (DARPA), Microsoft Research and the U.S. National Science Foundation (NSF), is one concerned with how a community of wireless sensors governs itself and elects its local representatives. Having each individual sensor communicate back to a central administrator creates several problems. First, the power required to communicate over long distances can quickly exhaust the reserves of each sensor. Second, the message of each sensor can be relayed or "hopped" by other sensors that are closer to the administrator, akin to the cutoff man between the outfield and home plate in baseball; however, sensors close to the controller would expend a majority of their energy relaying messages not their own. Third, the administrator would quickly become overwhelmed with messages from individual sensors, analogous to the White House having to respond to hourly email from each American citizen, regardless of the level or scope of their concerns.
Representative government addresses these problems through the clustering of citizens into a hierarchy of increasingly local groups, as in regions, states, counties, cities and townships. Instructions and goals can be relayed from the central adminstrator to the individual through their elected officials. Individuals can resolve issues locally with their representatives and, if they are of higher importance, the issues can be passed along up the chain of command.
Following this pattern, Professor Arora and his group have developed an algorithm called the fast local clustering service (FLOC), wherein each wireless sensor or node contains instructions for electing or serving as the representative of its local cluster, known as a "clusterhead." This method provides for the dynamic reconfiguration of local clusters such as is needed when the sensors are on mobile platforms, or when an existing clusterhead fails or exhausts its power supply. Because clusterheads report to upper nodes in the hierarchy, individual values can be summarized into concise reports that reduce the traffic load to the top administrator. Additionally, local elections, node additions and sensor failures are handled at the cluster level without instructions from the administrator, providing for a level of failure isolation and local healing for the entire network.
Recent developments in Professor Arora's group and others offer improvements and competing systems for governing distributed wireless networks. As is the case of making cookies, politics can be a messy business.
