Overview

The Systems Infrastructure Area works closely with domain scientists from across CENS (e.g., geophysics, plant biology) who would like to conduct a sensor network deployment experiment to: select appropriate results from systems and sensor research; identify missing elements and develop them as necessary; integrate existing components; and ultimately, deploy complete sensing systems and provide the domain experts with life-cycle tools for their management.  We ensure that domain-specific systems infrastructure needs are met, and yet maintain a multidisciplinary perspective so that broader cross-domain re-use is enabled.

Our work is necessarily interdisciplinary, but for reporting purposes we categorize our work as generalized deployment software tools, testbed-specific software tools, and low-power hardware platforms.  Generalized deployment software tools include ESS, Emstar, new routing algorithms (Hyper, Centroute), and SensorBase.  Testbed-specific tools have been developed for seismic and acoustic application areas.  Low-power hardware platform development has occurred at sensor interface board, mote, and microserver levels.

Generalized deployment software tools

• ESS is a software application designed to support the environmental monitoring needs of diverse domains, encompassing botanists working in forested settings such as the James Reserve in southern California mountains or lush botanical gardens, as well as civil engineers tracking groundwater chemical components in the Mojave Desert and Bangladesh. During the past year, we have deployed nine transects, totaling over a hundred ESS nodes and over 400 sensors in a variety of settings.  Through these extensive deployments, we have identified weaknesses and made the system robust.

• The Emstar sensor network development framework is a family of tools, libraries, and services that provides a unique environment in which to design, develop and deploy heterogeneous sensor network applications.  Emstar encourages and supports the use of simulation in the early stages of design and development by providing a range of simulated sensor network components—including radios—that provide the same interfaces as actual components, thus allowing literally the same code to execute in simulation mode (with no actual sensor network physical components), hybrid mode with some actual components (eg, radio) and some simulated components, and full native mode with no simulated components.  This allows basic code functionality to be extensively tested in a friendly simulated arena, and then steadily moving into more realistic (and more stressful) settings, eventually emerging as a real deployment.  Emstar is the foundation for virtually all sensor network development in the CENS Systems Laboratory.  Emstar currently has over 100 active users spanning at least 14 universities, 10 companies, 6 countries, and 4 continents.

• Routing: Extensive work with the original routing algorithm in ESS indicated that inherent deficiencies in that approach (“multihop”, a blend of two previous methods) were likely sources of occasional instabilities and would potentially limit scalability.  Two new research avenues, mobile microservers and multiple microservers, both needed to overcome these issues and so we have developed two new routing algorithms (Hyper, Centroute) to support these efforts.  Hyper uses a distributed algorithm to determine high-quality routing trees; Centroute uses a centralized algorithm at the root to construct ‘optimal’ trees.  We are currently engaged in an all-pairs evaluation program to assess the claimed relative merits of each of the three methods.

• SensorBase.org: Sensor networks often use different data storage and management mechanisms. SensorBase.org is platform for common data storage and management system for sensor networks. It provides users a uniform and consistent method for publishing sensor network data. It allows users to define data types, groups, and permission levels. It is a sensor network specific search engine, which allows users to query for specific data sets based on geographic location, sensor type, date/time range, and other relevant fields.  Users that want to publish data sets create a project description and define what types of sensor readings are allowed for the project. Users can also create new measurement types, sensor types, raw to actual calibration equations, permission levels, add users to the group, and add trust reference count to users.

Testbed-specific tools

• Seismic testbed: ESS is an ideal tool for applications that periodically sense, and then forward their sampling data to a centralized point for archiving.  Other modalities, such as multi-channel 100Hz sampling of seismic data, or high-frequency acoustic sampling as a basis for real-time distributed multilateration and localization of sources, fundamentally require different approaches to data sampling and processing.  The 50-node seismic deployment in Mexico further suffers from difficult RF connectivity issues and other failure modes, which demands new approaches to distributed data collection and control.  DTS (Disruption Tolerant Shell) is a new tool that explores control in this difficult mode.

• The acoustic testbed has been constructed using prototypes of the low-power LEAP technology described below, and focuses on distributed localization.  This task is inherently distributed at the signaling level, and our implementation further uses distributed processing to first localize the nodes themselves (creating an artificial 3-D coordinate system), and then to localize acoustic sources of interest to domain scientists (e.g., localizing woodpeckers and triggering a camera).

Low-power hardware platform development

Limitation on energy availability is a nearly ubiquitous concern in embedded sensor networks.  Domain scientists have a steady stream of new applications that demand higher-performance and lower energy consumption components.  We have made three significant strides this year:  a new low-power, high sampling frequency sensor board; a 3^rd-generation solar-powered mote assembly; and a new low-power, energy-aware processing platform suitable for use as a microserver.