Mobile Sensing Architecture for Air Pollution Monitoring

Abstract

Industrial growth has brought unforeseen technological advances to our society. Unfortunately, the price to pay for these advances has been an increase of the air pollution levels worldwide, affecting both urban and countryside areas. Typically, air pollution monitoring relies on fixed monitoring stations to carry out the pollution control. However, this method is too expensive, not scalable, and hard to implement in any city. The Mobile Crowdsensing (MCS) approach, a novel paradigm whereby users are in charge of performing monitoring tasks, allows environment monitoring to be made using small sensors embedded in mobile vehicles. The possible scenarios can be divided into two: urban scenarios, where a wide set of vehicles are available, and rural and industrial areas, where vehicular traffic is scarce and limited to the main transportation arteries. Considering these two scenarios, in this thesis we propose an architecture, called EcoSensor, to monitor the air pollution using small sensors installed in vehicles, such as bicycles, private cars, or the public transportation system, applicable to urban scenarios, and the use of an Unmanned Aerial System (UAS) in rural scenarios. Three main components compose our architecture: a low-cost sensor to capture pollution data, a smartphone to preprocess the pollution information and transmit the data towards a central server, and the central server, to store and process pollution information. For urban scenarios, we analyze different alternatives regarding the design of a low-cost sensing unit based on commercial prototyping platforms such as Raspberry Pi or Arduino, and Commercial Off-the-shelf (COTS) air quality sensors. Moreover, we analyze and propose a process to perform pollution monitoring using our architecture. This process encompasses four basic operations: data reading, unit conversion, time variability reduction, and spatial interpolation. For rural scenarios, we propose the use of an Unmanned Aerial Vehicle (UAV) as a mobile sensor. Specifically, we equip the UAV with sensing capabilities through a Raspberry Pi microcomputer and low-cost air quality sensors. Finally, we propose an algorithm, called Pollution-driven UAV Control (PdUC), to control the UAV flight for monitoring tasks by focusing on the most polluted areas, and thereby attempting to improve the overall accuracy while minimizing flight time. We then propose an improvement to this algorithm, called Discretized Pollution-driven UAV Control (PdUC-D), where we discretize the target area by splitting it into small tiles, where each tile is monitored only once, thereby avoiding redundant sampling. Overall, we found that mobile sensing is a good approach for monitoring air pollution in any environment, either by using vehicles or bicycles in urban scenarios, or an UAVs in rural scenarios. We validate our proposal by comparing obtained values by our mobile sensors against typical values reported by monitoring stations at the same time and location, showing that the results are right, matching the expected values with a low error. Moreover, we proved that PdUC-D, our protocol for the autonomous guidance of UAVs performing air monitoring tasks, has better performance than typical mobility models in terms of reducing the prediction errors and reducing the time to cover the whole area.Moreover, we analyze and propose a process to perform pollution monitoring using our architecture. This process encompasses four basic operations: data reading, unit conversion, time variability reduction, and spatial interpolation.