Back to Main Page

Sensing Everywhere!

DEBORAH ESTRIN, UC - Los Angeles Slides - 2.5 MB Download - 287 MB Watch the Talk (18:23)

Computing that leverages sensing allows us to create programmable observatories of the physical world that can begin to address environmental, community, and personal concerns. These systems have the ability to reveal the previously unobservable and in doing so help us understand and manage our interactions with the physical world, with scarce resources, and with one-another.

The Internet, machine learning, and computational models are among the many foundational technologies that have brought computing everywhere. The task today is to investigate how computing extends beyond the processing and sharing of knowledge embedded in text and numbers, to direct measurement, management and manipulation of physical phenomena.

One example is the role of sensing in the management of ecosystems, which are changing rapidly in response to human activities. Timely information about their structure and function and the services they provide is vital to conservation of habitat and biodiversity and mitigation of the effects of urbanization. However, ecosystems are complex and dynamic. They exhibit tremendous spatial variation and understanding them requires simultaneous measurement of physical, chemical, and biological parameters both above and below the Earth's surface.

Current efforts rely heavily on sensors that must operate over large distances. Remote sensing in satellites above the Earth provides sense data that has transformed our understanding of environmental and urban processes. However, we are limited by the resolution, or the amount of information that a pixel represents (often the median value spread out over meters or kilometers). Remote sensing alone cannot answer questions regarding biological phenomena that occur within the resolution of the pixel.

While we need advances in sensor development in the long term, fusion of available sensor data streams with one another, along with computational models and clever interpretation and mining of those data, can lead to drawing important inferences about the raw measurements we can take now. In addition, advances in miniaturization enable combinations of sensing, computation and wireless communication that can be placed in devices that are located in situ and up close to physical phenomena. The ability to observe physical occurrences with high spatial and temporal fidelity allows us to create models and make predictions, and thereby manage our increasingly stressed physical world. We need distributed sensing systems that collect large-scale, ground-based, and high spatial resolution measurements. Both the National Ecological Observatory Network (NEON) and the Ocean Observing Initiative (OOI) are excellent examples of this new approach enabled by information technology and sensing.

Imaging plays a key role in the interpretation of sensor streams. Wireless optical devices (smart web-cams) can act as biological sensors for species distribution, timing of leafing and flowering, vegetative health and disease, evidence of synthesis or calibration of carbon cycle models. Analyses, made possible by advances in digital imaging hardware devices and computer algorithms, can be used to automatically interpret those images. For example, one project collects daily images of over 1600 public web-cams that are pointed to areas of vegetation and uses a computer algorithm to quantify the greenness levels of those images to explore blossoming patterns during the onset of spring change in relation to weather patterns in particular micro-climates. Simple greenness images provide spatial resolution orders of magnitude greater than remote sensing systems.

Market forces have led to the adoption of mobile phones across all regions of the world. Such devices offer us a tool to observe the world on an unprecedented scale. By associating those observations with location and time, mobile phones extend the vision of leveraging imaging for ecosystems. For every remote sensing satellite in the sky, it is possible to have thousands of human eyes to refute or refine what's actually happening on the ground.

Another area of application of distributed sensing is in human systems. Always on, always worn devices represent the first broadly available and affordable technology to provide individualized location and activity-based observation in a manner that scales to very significant populations. The resulting data-streams provide significantly higher resolution when compared to the old system of form-based, retrospective self-reporting. The combination of personal data and models can now reveal how daily patterns interact with the environment and that feedback can inform personally and globally sustainable choices.

As we delve further into sensing, privacy concerns must be addressed. Over time, these personal behavior devices, combined with geo-spatial tracking systems, provides a body of minable data about an individual's life that is truly unprecedented. Do we need legislation protecting us from systematic discrimination based upon activity and location based information from these devices? How can systems that capture and share data at unprecedented granularity and scale practically address privacy concerns and retain their intended value?