Welcome to OpenSense at ETH Zurich
We are part of the OpenSense research project which is funded by Nano-Tera. OpenSense aims at investigating community-based sensing using wireless sensor network technology to monitor air pollution.
OpenSense in the Media
The project receives regular attention in both national and international media. Please refer to our media section for a list of articles.
Smartphone Application (iOS and Android)
We use the ultrafine particle maps created with the measurements from our mobile air pollution monitoring network to build the iOS and Android applications hRouting–The Health-Optimal Route Planner. Get it for free in the iTunes App Store and Google Play Store. The app provides urban dwellers in Zurich a health-optimal routing service. Citizens can reduce their exposure to ultrafine particles by not taking the shortest path between origin and destination but a healthier and slightly longer alternative route computed by the app.
The code of the iOS application is open-source and can be found on GitHub.
The code of the Android application is open-source and can be found on GitHub.
OpenSense Project Overview
Wireless sensor networks and publishing of sensor data on the Internet bear the potential to substantially increase public awareness and involvement in environmental sustainability. These technologies enable capturing sensor data by involving public authorities and the general public, and furthermore making real-time information on environmental conditions available to a wide public. Air pollution monitoring in urban areas is a prime example of such an application as common air pollutants have direct effect on the human health. However, bringing the vision of public involvement in environmental monitoring to a reality poses substantial technical challenges for the communication and information systems infrastructure, to scale up from isolated well controlled systems to an open and scalable infrastructure where many nano-scale sensors generate terabytes of data.
Challenges that are not well addressed today:
- Dealing with the heterogeneity and widely varying characteristics of the sensor equipment, measurements and data analysis.
- Supporting and exploiting mobility of sensors.
- Involving the community in a trusted, fair and transparent manner into the monitoring activity.
Air pollution monitoring is exceptionally suited to study these challenges as they are particularly pronounced in this scenario. A wide variety of sensors (meteorological data, air pollutants such as O3, NO2, NO, SO2, VOC, and fine particles) are used and measurements have to account for atmospheric transport phenomena. Sensors are frequently mobile (public and private vehicles, personal devices, airborne vehicles) and air pollution is today mostly a matter of public interest.
Datasheets of the sensors used in the OpenSense deployment:
- O3 sensor from SGX Sensortech
- CO sensor from Alphasense
- NO2 sensor from Alphasense
- Ultrafine particle sensor from Matter Aerosol
OpenSense addresses key research challenges in the domain of information and communication systems related to community-based sensing using wireless sensor network technology in the context of air pollution monitoring.
For more information on the project please visit:
Measurement Stations in Zurich
The OpenSense measurement platform is based on the prototype platform developed within the Nano-Tera project X-Sense (as part of the cooperation between the projects) and further extended for monitoring air pollution. The core of the prototype measurement station is a Gumstix embedded computer running the Linux operating system. The station supports GPRS/UMTS and WLAN for communication and data transfer. A GPS receiver supplies the station with precise geospatial information. Localization in cities is a challenging task due to street canyons, multi-path effects, and often low number of directly visible satellites. Thus, the measurement station is equipped with an accelerometer and receives the door release signal once installed on a tram to assist recognition of halts and tram stops to minimize the positioning uncertainty. The weight of the developed OpenSense station is approximately 4.5kg and energy consumption is 40W. The station is supplied with power from the tram.
Every station is equipped with an O3, CO, (later NO2), and a ultrafine particle sensor. The ozone sensor-a metal oxide semiconductor gas sensor-performs measurements by heating up the surface of a small microchip with a thin layer of a semiconducting metal oxide to several 100°C. When ozone gas is present, the electric conductivity of the semiconductor is altered. The CO and NO2 sensors are electrochemical gas sensors that measure the concentration of a target gas by oxidizing and reducing the target gas at the electrode. The mounting points of all gas sensors are covered with a thin Teflon layer to minimize interference of the target gases with the dust cover of the measurement station. The lifetime of all gas sensors is up to 3 years. As part of our cooperation, the University of Applied Sciences Nordwestschweiz (Fachhochschule Nordwestschweiz) provides ultrafine particle sensors and the necessary expertise. Furthermore, we monitor temperature and humidity in the enclosure.
At the current project state, we maintain two installations of measurement stations in Zurich: 10 stations on top of 10 trams in the city of Zurich and one station at the national air pollution monitoring network NABEL station in Dübendorf. Both deployments are briefly described below.
Thanks to the great support of VBZ (Verkehrsbetriebe Zürich), the first measurement station was installed on top of a tram in the city of Zurich at the end of September 2011. Currently we have 10 measurement stations travelling through the city on top of trams. The measurement schedule turns the stations off during nights when the trams are in their respective depots, hence, no energy is used when the trams are on battery power supply. Since the impact of mobility on the measured concentration was the subject to investigation, the early version of the schedule performed measurements at the stops rather than during the tram drive. Recognition of movement was performed based on accelerometer data. In the current scheduler implementation, the O3, and CO sensors are sampled every minute. The ultrafine particle sensor is sampled with 20Hz and 5 second averages are transmitted to the server. Additionally, we regularly receive high resolution sensor measurements from fixed NABEL and OstLuft stations in Zurich to perform reference measurements and sensor calibration.
The second station is statically positioned next to the NABEL station in Dübendrof and used as a long-term deployment in cooperation with EMPA/BAFU who kindly helped us with the installation and provide us with reference data. This deployment is running successfully since April 2011. We use the reference data obtained by this station to calibrate our sensors and to evaluate their performance under a wide range of weather conditions, which is difficult to achieve in a laboratory environment.
For calibrating the sensors, we implemented three calibration schemes for mobile sensor nodes. We investigate single-hop and multi-hop calibration given a reference station which can be reached by the mobile stations from time to time. The first scheme implements a standard way of calibrating gas sensors while the other two approaches show different trade-offs between measurement accuracy and calibration delay. We showed though experiments that when using these calibration schemes for ozone sensors we are able to measure ozone concentrations with an average error of 2ppb compared to the measurements done by the NABEL station. This is remarkable as the accuracy given in the datasheet of the sensor is 20ppb. Furthermore, we found a linear dependency of the calibration accuracy on the number of calibration hops. The accuracy loss is tolerable as long as the number of calibration hops is rather limited which is the case in public transport networks.
Besides the deployments described above, we also have a prototypical implementation of a smartphone-based measurement device. We connected a small-sized, low-cost ozone sensor to an off-the-shelf smartphone running the Android OS. The Android application assists the user with sensor calibration, displays sensor readings, stores them on the memory card, and uploads the stored data to a server for further data processing and visaulization.
Test Datasets (all data):
Data format: unix timestamp, concentration, temperature(°C), relative humidity(%), latitude, longitude(WGS-84).
Route scheduling is a problem of selecting a subnetwork of a timetable network to install measurement stations with the goal to optimize coverage of the city given a limited number of measurement stations. Since the measurement stations are equipped with gas sensors which need reference measurements from time to time, we demand that the subset of selected vehicles allows comparing measured sensor values across different sensors. This is possible only if sensing takes place at the same time and same location, meaning that the vehicles selected for the deployment should meet each other from time to time. The set of aggregated meeting points builds a network. Efficient design of this network contributes to the system’s fault tolerance by recognizing sensor malfunctioning, sensor precision loss due to sensor aging, and provides the necessary support for sensor calibration with a reference station. On top of the network of meeting points, we investigate different schemes to periodically calibrate the sensors.
On-the-fly calibration of mobile sensors given a few stationary reference stations (one NABEL and four OstLuft stations in Zurich). Manual calibration of gas sensors is an elaborate and time-consuming task. However, almost all gas measurement instruments require periodic calibration and reference measurements. Therefore, we primarily focus on automatic on-the-fly sensor calibration solutions. We exploit the fact that certain transport vehicles periodically meet each other or pass by static reference stations. Thus, spatially and temporally related measurements are used to adjust calibration parameters, which is necessary to filter out possible sensor aging effects.
Measurement scheduling. Given a route plan, a timetable, and a measurement density function. For a given mobile vehicle, regular sampling might lead to a suboptimal coverage of a city. We investigate measurement-scheduling schemes to provide optimal coverage given a route plan and a timetable. Since every measurement degrades the sensor due to involved chemical reactions, the provided problem solution assumes a limited number of measurements per sensor.
Personal sensing. Given the broad availability of personal mobile phones, it is obvious to use these devices to involve the general public into community sensing to sense common air pollutants. We investigate the possibilities and the benefits of such personal sensing.
- OpenSense: 2010 - 2013
- OpenSense II: 2014 - 2017
Partner Projects and Contributors
OpenSense closely collaborates with X-Sense, a Nano-Tera project managed by the Permasense consortium. The basic hardware platform was designed, developed, and used to monitor various environmental phenomena in the Swiss Alps within the Permasense project.
Valuable support is given by partners and friends: