Sensornets promise to extend automated monitoring and control into industrial processes. In spite of great progress made in sensornet design, installation and operational costs can impede their widespread adoption---current practices of infrequent, manual observation are often seen as sufficient and more cost effective than automation, even for key business processes. In this paper we present two new approaches to reduce these costs, and we apply those approaches to rapidly detect blockages in steam pipelines of a production oilfield. First, we eliminate the high cost of bringing power to the field by generating electricity from heat, exploiting the high temperature of the very pipelines we monitor. We demonstrate that for temperature differences of 80 °C or more, we are able to sustain sensornet operation without grid electricity or batteries. Second, we show that non-invasive sensing can reduce the cost of sensing by avoiding sensors that pierce the pipeline and have high installation cost with interruption to production. Our system instead uses surface temperature to infer full or partial blockages in steam pipelines and full blockages in hot water pipelines. Finally, we evaluate our "steam-powered sensing" system to monitor potential blockages in steam pipeline chokes at a production oilfield. We also show the generality of our algorithm by applying it to detect water pipeline blockages in our lab. To our knowledge, this paper describes the first field-tested deployment of an industrial sensornet that employs non-solar energy harvesting.

Detailed diagnostic data is a prerequisite for debugging problems and understanding runtime performance in distributed wireless embedded systems. Severe bandwidth limitations, tight timing constraints, and limited program text space hinder the application of standard diagnostic tools within this domain. This work introduces the Log Instrumentation Specification (LIS), which provides a high level logging interface to developers and is able to create extremely compact diagnostic logs. LIS uses a token scoping technique to aggressively compact identifiers that are packed into bit aligned log buffers. LIS is evaluated in the context of recording call traces within a network of wireless sensor nodes. Our evaluation shows that logs generated using LIS require less than 50% of the bandwidth utilized by alternate logging mechanisms. Through microbench-marking of a complete LIS implementation for the TinyOS operating system, we demonstrate that LIS can comfortably fit onto low-end embedded systems. By significantly reducing log bandwidth, LIS enables extraction of a more complete picture of runtime behavior from distributed wireless embedded systems.

In many cases, sensornets require continuous monitoring, 24x7, at remote, inaccessible locations making energy management a critical part of most sensornets. The sensornet research community has explored energy conservation and energy harvesting to address this problem of long-lived sensornets. Energy conservation is a primary concern in almost all sensornet work, and techniques from low-power hardware and OSes to coordinated network protocols and applications. Complementing energy conservation, energy harvesting gathers new energy from the environment.

An algorithm to estimate velocity fields from low-resolution video recordings is presented. The algorithm does not attempt to identify and track individual vehicles, nor does it attempt to estimate derivatives of the field of pixel intensities. Rather, a frame is compressed by obtaining an intensity profile in each lane along the direction of traffic flow. The speed estimate is then computed by searching for a best matching profile in a frame at earlier and later times. It is shown that the estimate is equivalent to a weighted median of velocities of vehicles. Because the algorithm does not need high-quality images, it is directly applicable to a compressed format digital video stream, such as .mpeg, from conventional traffic video cameras. The procedure is illustrated using a 15-min-long Video Home System recording to obtain speed estimates on a 1-mi stretch of highway I-80 in Berkeley, CA

Image processing, specifically automatic target recognition (ATR) in synthetic aperture radar (SAR) imagery, is an application area that can require tremendous processing throughput. In this application, data comes from high bandwidth sensors, where the processing is time-critical. There is limited space and power for processing the data in the sensor platforms or in battlefield groundstations. DoD's strong push for using commercial-off-the-shelf (COTS) technology, the very high non-recurring engineering (NRE) costs for low volume ASICs, and evolving algorithms limit the feasibility of using custom special purpose hardware. In addition, a scalable system is required as the different sensor platforms have different image pixel rates and different mission requirements have different target recognition throughput needs per pixel. In this paper, we describe an ATR algorithm implementation using FPGA accelerators. We first describe the ATR algorithm that was implemented, the implementation on a single FPGA, how the FPGA nodes are connected to make a scalable system, and compare the performance to current scalable microprocessor-based implementations.