Titulo Estágio
Anomaly detection in transportation systems through multi-source traffic data mining
Áreas de especialidade
Sistemas Inteligentes
Local do Estágio
DEI-FCTUC
Enquadramento
Due to the increase in motorization and population growth, transportation systems are put under great strain, resulting in increased traffic congestion, air pollution, and fuel consumption. This results in millions of working hours lost in traffic jams and billions wasted due to increased fuel consumption as well as material damage and medical and work loss costs due to accidents.
For these reasons, designing efficient and safe transportation systems has become one of the challenges of our time. By using wireless communications and sensing technologies, we can increase the traffic flow, decrease the emissions, and prevent accidents. Increasing the utilization of the roadway by just a fraction would result in billions saved in terms of both workers' productivity and fuel costs. Similarly, by improving the safety at critical locations (highways and intersections), thousands of lives can be saved.
Objetivo
Intelligent Transportation Systems (ITS) applications can be enabled through a spectrum of devices. There are several applications focused on using either vehicle sensors and inter-vehicle communication, or mobile devices (e.g., smart-phones). However, there is scarce work that looks at combining the data collected from different sources.
With this thesis we aim at using insights gained from multi-source data analysis and modeling to devise an application for detection of anomalous events (e.g. sudden stops, traffic jams, etc). By analyzing the underlying characteristics of ITS systems such as mobility patterns, spatial relationships, and physical surroundings, this application shall make use of machine-learning techniques to provide detection of anomalous events.
By employing data-mining algorithms, we aim at extracting pertinent events from gathered data (e.g., speed variations to detect when a vehicle breaks down) and combine them with sensor information from mobile devices (e.g., mining the information from the accelerometer to distinguish between sudden stops and usual traffic dynamics) to properly detect anomalous events.
Plano de Trabalhos - Semestre 1
- State-of-art of methodologies & applications for intelligent transportation systems
- Study and characterization of sources of data for anomaly detection applications
- Study of data representation and machine-learning methodologies for classification of data
- Intermediate thesis report writing
Plano de Trabalhos - Semestre 2
- Implementation of the identified machine-learning solutions for detection of anomalous events
- Validation: evaluation and testing of the proposed solution
- Writing of a scientific article
- Writing of dissertation
Condições
Possibility of a paid internship during the 2nd semester at [url=http://uk.nec.com/en_GB/emea/about/neclab_eu/]NEC laboratories Heidelberg[/url], Germany.
Observações
Survey article on sensing platforms and applications for vehicular networks available here.
There are several open data sources that will server as the basis for this work, such as, for example, for the cities of Dublin, London, Manchester, and England as a whole. Whenever needed, these shall be complemented by synthetic traces generated with the SUMO simulator.
Orientador
João Vilela (DEI), Mate Boban (NEC), Bernardete Ribeiro (DEI)
jpvilela@dei.uc.pt 📩