Group ProfileOur research addresses several aspects of network systems and services, including participatory sensing, software-defined networking, technologies for disaster preparedness and responses, indoor location based applications and services, and collaborative data projects. These areas are described below.
1. Participatory Sensing
We study the Internet of Things (IoT) and participatory sensing systems, and have combined the two concepts to build a large-scale system, called AirBox, for fine particulate matter (PM2.5) monitoring. The project engages citizens to participate in environmental sensing and enables them to make low-cost PM2.5 sensing devices on their own. It also facilitates PM2.5 monitoring at a finer spatio-temporal granularity and enriches environmental data analysis by making all measurement data freely available. As of mid-2020, we have deployed more than 15,000 devices in over 58 countries, and we have established collaborations with international and domestic researchers on several key issues regarding IoT systems and the interdisciplinary topics of environmental monitoring systems.
Specifically, we have investigated the data quality issue of the AirBox system, and we have conducted an anomaly detection framework to detect anomalous events in the real-time data stream. We have developed a cluster-based approach for short-term PM2.5 concentration forecast, and we have designed a clean air routing algorithm to provide path navigation with minimal PM2.5 exposure. We have carried out an adaptive sensor calibration mechanism and conducted a data fusion algorithm to integrate multiple PM2.5 data sources of different accuracy levels. Our ongoing work are to improve spatio-tempoal data analysis and identify additional intrinsic properties of AirBox data, and to extend our approaches to support other sensing systems of different environmental factors, such noise, odor, and radiation.
2. Software Defined Networking
SDN enables flexible routing for the traffic engineering of unicast and multicast flows to increase network scalability. However, the problem is much more challenging for multicast since the number of possible multicast groups is exponential to the number of nodes in the network. To tackle this issue, we formulated a series of new optimization problems with practically important SDN constraints, proved the NP-Hardness and inapproximability, and designed new algorithms with the tightest approximation ratios. Specifically, we first proposed an approximation algorithm with the tightest bound to improve the scalability for multicast group communications in SDN, and a new reliable multicast tree was introduced for SDN to reduce the recovery cost under the link and node capacity constraints. Next, the proposed multicast tree was extended to support multiple multicast groups for online SDN traffic engineering with network function virtualization (NFV) by embedding the service chain consisting of a sequence of virtualized network functions (VNFs). Finally, low overhead rerouting schemes were proposed to support online SDN traffic engineering with and without segment tree routing. The proposed δ-approximation and δ-competition algorithms can achieve the tightest approximation ratios. In addition to sophisticated theoretic analysis, the above algorithms were also implemented in real SDN networks with HP OpenFlow switches and Floodlight for YouTube traffic to show their practical applicability, and the results have been transferred to Inventec and EstiNet.
3. Technologies for Disaster Preparedness and Responses
For many years, the multi-disciplinary Sustainability Science Research project DRBoaST (Disaster Resilience through Big and open data and Start Things) provided us with opportunities and support to collaborate with research fellows and faculty members in earth science, urban planning and computer science on information and communication technologies for disaster management. At the start, our focus was on technologies for building open and sustainable disaster management information systems. Our results in this direction include prototypes, models and algorithms for responsive emergency trustworthy information brokerage service; disaster resilient heterogeneous and plug-n-play networks and dynamic logical information exchange; and models and tools for fusing data from intelligent things and people.
Our more recent emphasis was on the generation and use of data. One thrust was directed towards developing methods and tools for generation and collection of data needed for disaster risk reduction. Examples include CROSS (CROwdsouring Support system for disaster Surveillance). A typical disaster surveillance system needs to make critically important decisions within minutes or hours before disasters strike. When it is necessary to crowdsource human sensor data, an emergency manager needs help to recruit and manage trained volunteers, select participants from volunteers, plan tours for them to cover locations where observational data are needed, and fuse data collected by them in real-time with physical sensor data to improve the quality of sensor coverage. CROSS was built to meet these needs. We solved realistic variants of underlying participant selection, tour planning and symbiotic data fusion problems and produced solutions not only of practical utility to CROSS, but also of theoretical significance. Selected components were integrated into the well-known platform Ushahidi. CROSS has been used by our earth science colleagues who use trained volunteers after each significant earthquake in Taiwan to collect data on new geo-hazards for sake of assessing risks of earthquake-triggered compound disasters.
Another major thrust of our work was directed toward developing active smart embedded devices, mobile applications and services/systems in smart homes and buildings that can automatically process alerts from authorized senders and building safety systems and take location specific actions to minimize chance of injuries and reduce property damages when disasters strike. In particular, each device (or application) selects its action (or response instructions) in response to an alert based on not only the type and parameters of the alert but also on attributes of the building, interior layout and nearby objects around the device (or application). Our work has demonstrated that easy to customize and maintain active smart devices/applications for diverse purposes can be built on a common architectural framework from reusable components and alerts messages pushed asynchronously over the Internet can meet the end-to-end delay requirements of time-critical alerts. Our experimentation with a prototype active emergency response system in an office building during a simulated strong earthquake has demonstrated the effectiveness of such systems. As a step towards enabling the pervasive use of active devices and applications, a proof-of-concept prototype building/environment data and information (BeDI) system (or BeDIS) was built as a part of the ICT infrastructure needed to support location-specific, active emergency preparedness and response within large public buildings.
4. Indoor Location Based Applications and Services
Our effort after DRBoaST project ended in December 2018 has focused on transforming BeDIS prototype into a ready-for-deployment platform. BeDIS is a system of Lbeacons (location beacons). They are Bluetooth low energy devices with directional antennas pervasively installed throughout the building. During normal times, each Lbeacon broadcasts its own locally stored 3D coordinates and a brief location description. This capability enables hundreds and thousands of people in the building to locate themselves and navigate amidst dense crowd and moving objects via their mobile phones. Each Lbeacon also collects and timestamps MAC addresses of Bluetooth tags carried by objects of interest such as valuable equipment, critically needed tools, people requiring care, and so on. This capability enables BeDIS to locate the objects and track their movements in real-time. Location accuracy of 3-5 m or 6-10 m is achieved by adjusting the ranges and beam widths of Lbeacon antennas. The response time of the current BeDIS is 2-3 seconds nominally.
BeDIS is unique among competing platforms by being both an infrastructure for indoor positioning and indoor navigation (IPIN) applications and a platform for indoor location based services (ILBS). Furthermore, BeDIS has several distinct advantages compared with IPS (indoor positioning systems) built on competing technologies. They include that it degrades acceptably small in location accuracy and response time in presence of large crowd, and it is easy to configure and low cost to deploy and maintain. Moreover, Lbeacon can be easily configured to be a micro data server and used in an extended BeDIS that supports active disaster preparedness and response. When triggered by a disaster/emergency alert from responsible government agencies or the building safety system, an extended BeDIS functions as a system of micro data servers for delivering location- and situation-specific emergency response instructions to people and attributes of the building, interior layout and objects in their immediate vicinities to support the choices of response actions of active devices and applications within fractions of a second to seconds.
To date, we have ready for market the following applications and services:
- BeDIS: an infrastructure/platform for IPIN and RT-ILBS applications;
- WPIN (Waypoint-based indoor navigation) applications with head-up-display style GUI (graphical user interface) and voice based navigation directions on diverse smart phones for general public and special-purpose navigators for managing outpatients flows during their visits in hospitals and health clinics to minimize the total time required by each patient to complete his/her visits; and
- BOT (BeDIS object tracking) and geo-fencing services for locating objects and tracking people indoors in general, within hospitals and other care-providing institutions specifically.
Several field trials are being conducted to assess their functionality and usability. The test sites include National Taiwan University Hospital Taipei and Yunlin Branch (BOT in ER and patient wards and WPIN in areas visited by outpatients.) Other sites include Changhua Christian Hospital, Taipei City Hall and Yunlin Veteran Home.
5. Collaborative Data Projects
We work on tools, systems, and services for collaborative data production, access, and preservation. Resource-constrained research projects often rely on third-party services for on-going data sharing and archiving. Such practices may not work well when the project involves a large number of collaborators participating in a fluid modality. In citizen science projects, for example, thousands of people may participate together. We list below a few collaborative data projects undertaken jointly with other research units (within and external of Academia Sinica), with the Institute taking a leading role.
depositar — Research Data Repositories for All.
depositar (https://data.depositar.io/) is a research data repository built upon CKAN, a free software package for publishing open data. We added several extensions: 1) rich metadata support, 2) spatiotemporal annotation and query, 3) preview and overlay of spatial datasets, 4) Wikidata-powered keyword, and 5) open registration for all. The services have been used by many for active research data management. The source code and user manual of depositar are freely available. We are also working with the Ministry of Science and Technology to develop research data management policies and guidelines for research projects supported by grants from the Ministry.
Taiwan Roadkill Observation Network (TaiRON).
TaiRON (https://roadkill.tw/) is a collaborative project on collecting (smartphone) photo observations on roadkill and other animal mortality in Taiwan. It started at the Taiwan Endemic Species Research Institute in 2011, initially using Facebook Group as the means for data collection. We redesign the data workflow so that participants first upload observation records to a dedicated project website. The website then generates digests and relay them to the Facebook group for species identification and community interaction. The new workflow allows for better data management and analysis, as well as individual control of observation records. TaiRON datasets have been instrumental in monitoring unusual animal deaths in Taiwan, and in designing new measures to reduce roadkill. The project received a 2019 National Agricultural Science Award in the Sustainable Environment category. TaiRON currently has more than 5,000 active participants.
The Sunflower Movement Archive.
This public archive ( http://public.318.io/) is a collection of artifacts, images, and videos from the 2014 Sunflower Movement. Most of the materials were acquired by Academia Sinica when the students and activists were vacating from the occupied chamber of Taiwan’s Legislature. The archive permits anyone to search and identify objects in the collection, and it makes possible for owners to release high-resolution images of their creations to the public for reuse. The Sunflower Movement signifies a turning point in Taiwan’s recent history, and it continues to influence the political landscape and societal reflection of the island. The entire collection had been transferred to the National Museum of Taiwan History, and was featured in a special exhibition on Social Movements in Post-War Taiwan.