Pollution sensing and diagnosis have long been separated from the people most impacted by them. It was conducted by specialists with expensive and scarce equipment. As a result, testing was infrequent and decisions on mitigation were made by central planners with limited access to data. Worse yet, individuals with the ability to dramatically limit dangerous exposure via minor changes in behavior have been left blind to the relationship between their daily actions and exposure to pollution. Advances in computing, sensing, and wireless communication technologies have the potential to allow those most affected by pollution to participate in pollution sensing, rational cost assessment, and mitigation.
This project focuses on developing a distributed mobile system for socially-collaborative environmental monitoring, which greatly reduces the problem of environmental sensing data scarcity, supports richer environmental sensing data analysis, and enables better environment awareness and protection via social collaboration. This project will develop a system composed of inexpensive sensing and computation devices purchased by individuals for their own edification and protection. These embedded systems will communicate with each other and aggregate data, enabling multi-sensor localization of pollution sources and quantification of the potential damage by each polluter. By measuring pollution and modeling its impact, it will be possible to associate pollution sources with the costs they impose. Furthermore distributed networking will allow individuals to actively participate in and socially collaborate on environmental monitoring and protection.
The people most impacted by air quality have long been separated frominformation to help them understand how their own behavior influences exposureto pollution. In part, this was because air quality sensing systems weregenerally large, expensive, and stationary, and required experts to operatethem. Even though many of these systems could accurately measure pollutantconcentrations, only a few locations could be monitored, and most people spentmost of their time at other locations, locations that were "in the dark" fromthe perspective of air quality sensing science. A major goal of this project has been to develop technologies to supportinexpensive, personal, air quality monitoring. To this end, we have worked on embedded sensing hardware and software design, indoor localization, sensor calibration and characterization, methods for increasing the accuracy of pollution estimates, in-lab experiments and field trials, as well as education and outreach activities. In the following paragraphs, we summarize our efforts and contributions in these areas. Embedded sensing hardware and software: We have designed, fabricated, andopen-sourced the hardware/software embedded system necessary to supportpersonal air quality monitoring. This hardware is compact and containsmultiple air quality sensors (e.g., for carbon monoxide, carbon dioxide, andvolatile organic compounds) as well as other sensors (e.g., temperature andhumidity) to help correct for susceptibility of the air quality sensors tochanges in environmental conditions. This embedded system communicates withcommodity smartphones via Bluetooth, allowing data to be stored, aggregated,and analyzed on our servers. In addition to serving as gateways between oursensing hardware and remote servers, smartphones also allow users to see therelationship between their behavior (e.g., motion patterns) and pollutionexposure, and to see position-dependent pollution data from other sources,including other mobile sensors. Accurate indoor localization: One pressing challenge for our studies was developing a method of accurately estimating the locations of sensors when indoors, where people spend most of their time and GPS is generally notavailable. We developed and published a number of new ideas on this topic, including accurate and fully-automated Wi-Fi fingerprinting for indoor room localization, culminating with a method of automatically estimating building floorplans by observing the motion patterns of participants. Sensor calibration and characterization: To make personal air quality sensingaccessible, sensing hardware must be easy-to-carry, inexpensive, and requirelittle professional maintenance. This, unfortunately, rules out large andexpensive sensors. For many pollutants, the types of sensors appropriate forcompact personal air quality sensing systems are subject to errors as a resultof the presence of other pollutants (cross-sensitivity) or due to changes inenvironmental conditions such as temperature and humidity. We have studiedthese dependencies in order to develop techniques to compensate forcross-sensitivity and changes in environmental conditions, and our findingshave been published. Methods for increasing the accuracy of pollution estimates: Inexpensive,portable sensors are susceptible to cross-sensitivity to other pollutants, aswell as dependencies on environmental parameters such as temperature andhumidity. In addition, many such sensors have readings that gradually driftover time. This project has yielded three main techniques to improve theaccuracy of pollution estimates. (1) We developed a drift compensationtechnique allowing mobile sensors to opportunistically recalibrate wheneverthey are near other mobile or stationary sensors. We developed a method ofusing knowledge of the rate of drift to optimally weight the conflictingreports of different sensors during the recalibration process. This ideaimproves the accuracy of personal mobile sensors without requiring annoyingexplicit recalibration by study participants. (2) We developed a technique toestimate pollution concentrations based on knowledge of room-to-room airflow patterns in indoor environments. (3) Finally, we have developed agraphical model based method of estimating pollutant concentrations in thepresence of cross-sensitivities. This method is also capable of reportingwhether particular estimates are high- or low-confidence, allowing outliers to be eliminated. Lab experiments and field trials: Extensive lab experiments and multiple field trials have been carried out throughout our project to learn more about real-world pollution exposure and to evaluate the research ideas described in the preceding paragraphs. These include studies in Boulder, Denver, and Fort Collins in Colorado, and within the Navajo Nation, which were closely integrated with our outreach efforts. A number of the project's publications were motivated and/or validated by our lab experiments and field trials. We have conducted extensive education and outreach activities with the focus on getting the tools and knowledge learned in this project into the hands of tribal college students, high schools in rural Colorado, and even broadly to the public in the Front Range of Colorado. These include workshops and scientific affairs supporting high school and middle school students, international collaboration with China and Ghana, and numerous in-field deployment studies. These activities have been essential in establishing and strengthening the collaboration between the research team and the U.S. Environmental Protection Agency (EPA) and other research/industrial/nonprofit groups, as well as fostering interdisciplinary research and training of both cyber and environmental science and engineering.
