Thesis Title
Leveraging Pre-trained Models for Multi-Task Trajectory Analysis in Urban Transportation

Research Description
The research focuses on urban trajectory representation learning, aiming to develop a pre-trained self-supervised learning model that unifies the representation learning of diverse forms of urban trajectories and generalize to multiple downstream trajectory analysis tasks. The pre-trained model is expected to extract robust and generalized representations from large-scale, unlabelled, low, and irregularly sampled trajectory data, while integrating diverse geospatial knowledge to enhance trajectory semantics.

Supervisors
Prof Flora Salim, University of New South Wales