Gaussian Random Vector Fields in Trajectory Modelling

Abstract

This paper proposes the use of Gaussian random vector fields as a generative model to describe a set of observed trajectories in a 2-dimensional space. The observed trajectories are sequences of points in space sampled from continuous trajectories that are assumed to have been generated by an underlying velocity field. Given the observed velocities connecting the trajectory points, a vector field is obtained by condition- ing a Gaussian random vector field. Some results obtained in simulation are presented.