Abstract
In this demo we present INSIGHT, a system that provides traffic event detection in Dublin by exploiting Big Data and Crowdsourcing techniques. Our system is able to process and analyze input from multiple heterogeneous urban data sources.
You have full access to this open access chapter, Download conference paper PDF
Similar content being viewed by others
Keywords
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.
1 Introduction
In this demo we present a traffic monitoring system that is currently deployed in Dublin and utilized for city event detection. The purpose of this demo is to show, that, using novel data mining techniques we are able to monitor diverse data coming from city-wide infrastructures and extract useful information to present to the city operators. We collaborated with Dublin City Council (DCC) and designed a system that is able to process real-time data from diverse input sources such as sensors mounted on top of buses, traffic sensors embedded in street intersections or even citizen’s tweets. INSIGHT identifies events of interest such as traffic congestion, construction works and accidents [1].
Although sensor data are available to smart city authorities, it is very difficult for human operators to monitor the vast amount of information. Figure 1 shows the DCC control center, where one of the screens displays INSIGHTFootnote 1. Our system identifies events and aids operators to react in a timely manner by providing tools that: (1) Automatically identify the locations where abnormal sensor behaviour occurs, (2) Analyze historical data and build models that capture the sensors’ normal behaviour, and (3) Exploit ubiquitous human sensors to complement the existing data sources.
2 System Description
Our system, as described in Fig. 3 consists of multiple components; each is responsible for monitoring an individual data source. Data are generated from a set of heterogeneous traffic sensor networks in Dublin. Each stream is processed by an Intelligent Sensor Agent (ISA) that identifies anomalies. The detected anomalies are fused to infer events (see Round Table) and are presented to the traffic operators. In case of uncertainty, Crowdsourcing tasks are initiated.
Data Sources. Our system analyzes heterogeneous streaming data from the following input sources: (i) Buses that transmit their GPS location and information regarding their route. (ii) SCATS sensors that are deployed at intersections and transmit traffic flow measurements and the degree of saturation. (iii) Tweets that are posted by users in the Dublin area. (iv) Crowdsourcing input from users that provide feedback by reporting an event or answering to dynamic system queries using the CrowdAlert app from their mobile devices (see Fig. 2).
Architecture. Our system receives raw data from the heterogeneous data sources and channels processed information to the corresponding ISA Footnote 2. The SCATS ISA checks whether the SCATS sensor’s behaviour deviates significantly from its neighbours behaviour, using a multivariate ARIMA model.
The Bus ISA is responsible to detect anomalies, monitoring the bus data. This component exploits the Lambda architecture, using StormFootnote 3, to transmit data to concurrently running Complex Event Processing (CEP) engines. These engines monitor whether the streaming data trigger the set up rules. The Bus ISA enables the automatic adjustment of needed resources (elasticity), estimating the system load in upcoming time windows, using Gaussian Processes [4].
The Twitter ISA analyzes tweets that are geo-located in the Dublin area and identifies those that describe an event (traffic, flood, or fire), using a SVM text classifier. The Twitter API sets a number of constraints in terms of how many users, keywords or locations can be tracked at the same time. INSIGHT utilizes a dynamic filtering approach that continuously evaluates candidate keywords, users, and locations and selects an optimal subset.
The output of the ISAs is integrated in the Round Table component, which aggregates information [3], infers about what is happening and forwards the detected events either to the DCC operators through a web interface or to the crowdsourcing users requesting clarification. The Crowdsourcing Server acts as a middleware among the system and the users. Its main responsibilities are: (i) to keep track of the active users that are able to participate in the crowdsourcing tasks, (ii) to assign tasks to the Crowdsourcing users, in order to resolve possible ambiguities [2], (iii) to receive reports asynchronously from the users when an event occurs, and (iv) to provide feedback to the users regarding ongoing events near their location. The Crowdsourcing App is an Android application, called CrowdAlertFootnote 4 (illustrated in Fig. 2) that enables the human crowd to monitor the ongoing traffic events and provide feedback to the authorities.
User Evaluation. INSIGHT has been evaluated by the Traffic Management team at DCC. A set of employees with diverse responsibilities were selected, evaluated the system and filled a questionnaire. The questions regarded the usefulness and accuracy of the system and how it affected their workflow. The team members were quite happy with the system that provided them accurate events in real time and a set of tools that visualize real-time and historical dataFootnote 5.
3 Demo Description
We will demonstrate the web interface and the CrowdAlert app that visualize the detected events, from the different ISAs. During the demonstration, both, real-time, and historical data will be available, to highlight multiple aspects of the system. An example of an identified fire event is shown in Fig. 4.
Equipment. We will be using a laptop in order to demonstrate the front-end of INSIGHT, which is currently used by DCC. We will also be using Android devices to demonstrate the CrowdAlert app.
Demo Plan. The users will be able to view and interact with the system that is currently deployed and running at DCC. They will be able to overview information about the anomalies and events identified by the analysis components in real time. Moreover, the users will be able to select past time periods, redefine the analysis thresholds and parameters and re-run the system. Users will be able to generate plots that visualize data originating from the Buses and the SCATS sensors. Finally, we will also demonstrate the CrowdAlert app using smartphones and tablets. The users will be able to use CrowdAlert and interact with our system. Thus, they will be able to observe events that take place in Dublin in real-time as well as to report events and provide feedback through crowdsourcing tasks.
Notes
- 1.
http://www.insight-ict.eu/ - INSIGHT was funded by an FP7 EU grant.
- 2.
- 3.
- 4.
- 5.
References
Artikis, A., Weidlich, M., Schnitzler, F., Boutsis, I., Liebig, T., Piatkowski, N., Bockermann, C., Morik, K., Kalogeraki, V., Marecek, J., Gal, A., Mannor, S., Kinane, D., Gunopulos, D.: Heterogeneous stream processing and crowdsourcing for urban traffic management. In: EDBT (2014)
Boutsis, I., Kalogeraki, V.: On task assignment for real-time reliable crowdsourcing. In: ICDCS, pp. 1–10, Madrid, Spain, June 2014
Schnizler, F., Liebig, T., Marmor, S., Souto, G., Bothe, S., Stange, H.: Heterogeneous stream processing for disaster detection and alarming. In: 2014 IEEE International Conference on Big Data (Big Data), pp. 914–923. IEEE (2014)
Zacheilas, N., Kalogeraki, V., Zygouras, N., Panagiotou, N., Gunopulos, D.: Elastic complex event processing exploiting prediction. In: Big Data. IEEE, USA, October 2015
Acknowledgments
This research has been financed by the European Union through the FP7 ERC IDEAS 308019 NGHCS project, the Horizon2020 688380 VaVeL project and a Yahoo Faculty award.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2016 Springer International Publishing AG
About this paper
Cite this paper
Panagiotou, N. et al. (2016). INSIGHT: Dynamic Traffic Management Using Heterogeneous Urban Data. In: Berendt, B., et al. Machine Learning and Knowledge Discovery in Databases. ECML PKDD 2016. Lecture Notes in Computer Science(), vol 9853. Springer, Cham. https://doi.org/10.1007/978-3-319-46131-1_5
Download citation
DOI: https://doi.org/10.1007/978-3-319-46131-1_5
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-46130-4
Online ISBN: 978-3-319-46131-1
eBook Packages: Computer ScienceComputer Science (R0)