With the availability of large-scale GPS trajectory data, automatically detecting anomalous trajectories has become a critical concern in many real-world scenarios. For example, taxi drivers may take unnecessary detours to overcharge tourists. The main challenge of this task comes from the complex patterns of traveling routes in both spatial and temporal dimensions. We aim to develop deep generative models that can automatically discover normal traveling routes in different time periods and detect anomalous trajectory online based on the time and the route passed.

Duration and Type

  • Preferable 1 semester, but can be also done in 2 semesters
  • Honours project, other postgraduate project (MSc, MProfStuds, …)


  • Background knowledge of maths, statistics and probability theory
  • CS and machine learning skills
  • Programming experience in Tensorflow or Pytorch

Supervisor and contact

  • Kaiqi Zhao
  • Send CV and transcript by email to Kaiqi Zhao