P2: Trajectory Forecasting

• Applicant: Professorin Dr. Anne Driemel
• Subject Area: Theoretical Computer Science ; Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing
• Term: since 2021
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 313421352

Anticipation of movement is a challenging problem when the movement is irregular, diverse, has many degrees of freedom, and the recorded trajectories are missing data. An additional challenge is posed by the fact that often the movement is captured in a continuous environment with a granularity that is different from the desired granularity of the anticipation. Computational methods that anticipate movement therefore need to generalize from the observed data in several different aspects, namely in the space and the time dimension, and potentially also across several observations of the same or similar movements.In this project, we aim to address these challenges by transferring methods from the area of geometric algorithms and shape matching. In doing so, we desire to achieve provable guarantees under realistic input assumptions which have been successfully used in other contexts in the past. The main goal is to develop provably efficient and approximately accurate computational methods that facilitate the anticipation of movement based on past observations of the moving object or a group of moving objects. On a technical level, we approach the problem by developing new data structures for searching among trajectories using partial queries, by developing new algorithms for segmenting and clustering trajectories into frequent patterns, and by developing new hierarchical representations of clustered structures derived from sets of trajectories. The focus is on computational efficiency and provable correctness within specified approximation errors and probabilistic error bounds.

• DFG Programme: Research Units
• Subproject of: FOR 2535: Anticipating Human Behavior