PhD Research

Coordinating measurements for air pollution monitoring in uncertain participatory sensing settings

My goal is to influence urban planning decisions and policy making by understanding potentially harmful environmental phenomena such as noise and air pollution. To do that I developed novel coordination algorithms to involve citizens in the intelligent collection of data by taking informative measurements when and where it is needed.

It is estimated that there are more than 7 billion mobile phone devices active worldwide. This radical growth of mobile technology is starting to be exploited by experts for cheap large-scale data collection. In this research, we are interested in environmental data, such as radiation, noise and air pollution, which is crucial for public health. The traditional approach of collecting environmental data typically requires expensive to obtain and maintain equipment, as well as a number of environmental sciences experts to administer them. On the other hand, by exploiting the wide availability of mobile devices, fine grained sensor data can be collected for a large geographical area like a big city. This data can be used to create detailed heat maps providing insight to experts about the environmental phenomenon, which in turn will affect the authorities in decision making and urban planning. In more detail, we are interested in the concept of participatory sensing, where people contribute information from low-cost mobile devices they carry with them. This information can include measurements taken from mobile phones such as sound amplitude to measure noise pollution or air quality measurements taken with the assistance of specialized equipment. However, even though collecting data through people's mobile devices is a very effective and cheap, people are self-interested actors that have only local information about the environment and pursue their own agenda. This means measurements may be taken in a suboptimal way. In particular, participants often do duplicate work, i.e., different people take a number of measurements at the same location and time or they do not explore the whole map of interest, which leads to a partial or false picture of the environment.

To address these challenges, a coordination system is needed to guide or suggest where, when and who should take measurements. Specifically, the use of intelligent algorithms can solve this problem by coordinating and assisting humans to take more informative measurements as well as fill the gaps for areas that are not covered yet and avoid duplicate work. Moreover, since humans are often predictable in their daily routines the system can exploit this fact in order to make more informative suggestions to people. In particular, a key aim in this work is to ensure that people can get suggestions about taking measurements at times and locations that are least intrusive to their daily life.

Against this background, we provide a complete participatory sensing framework with algorithms for coordinating measurements for environmental monitoring. Our algorithms use local search, heuristics, clustering techniques and stochastic simulations to map participants to observations that needs to be taken. In particular, our algorithms intelligently search through the space of possible solutions to find mappings that will maximize the total information learned about the environment in a given time period.

We empirically evaluate our algorithms on real-world human mobility and air quality data and show that they outperform the state-of-the-art greedy and myopic algorithms.