Representing Spatial Trajectories as Distributions
This addresses trajectory prediction for applications like autonomous vehicles or mobility analysis, but appears incremental as it builds on existing representation learning approaches.
The paper tackles the problem of representing spatial trajectories with uncertainty by modeling partial observations as probability distributions in a learned latent space, enabling interpolation, extrapolation, and attribute modification. Experiments demonstrate advantages over baselines in prediction tasks.
We introduce a representation learning framework for spatial trajectories. We represent partial observations of trajectories as probability distributions in a learned latent space, which characterize the uncertainty about unobserved parts of the trajectory. Our framework allows us to obtain samples from a trajectory for any continuous point in time, both interpolating and extrapolating. Our flexible approach supports directly modifying specific attributes of a trajectory, such as its pace, as well as combining different partial observations into single representations. Experiments show our method's advantage over baselines in prediction tasks.