Data collection and cross-domain representation models for trajectory analysis

Due to the vast amount of available tracking sensors in recent years, high-frequency and high-volume streams of data are generated every day. Such tracking sensors include but are not limited to vessel, airplane or vehicle tracking data, drones, smartwatches and smart bands as well as cameras and earth observation sensors. Despite the overabundance of data generated by the tracking devices, there are still cases in which the trajectory of a moving object (e.g., vessel or human) has gaps, errors, or is unavailable. This research proposal aims to advance trajectory mining from a multimodal perspective, addressing challenges that might arise from single- or multi-source tracking data, across two main aspects: i) Data collection and creation: effort will be allocated to create open-access datasets from multiple data sources, acting as a basis for multi-sensor trajectory mining research. ii) Cross-domain representation models for trajectories: Transformation techniques are necessary to convert raw data into formats suitable for Artificial Intelligence/Machine Learning (AI/ML) applications. Hence, graph-based representation models inspired by other domains within the Information and Communication Technology (ICT) will be applied for trajectory analysis.

Faculty Supervisor:

Aris Leivadeas

Student:

Partner:

National Technical University of Athens

Discipline:

Computer science

Sector:

Artificial Intelligence; Information and Communications Technology; Transportation (excluding aerospace)

University:

École de technologie supérieure

Program: