Destination-Directed Trajectory Modeling and Prediction Using Conditionally Markov Sequences

Reza Rezaie, X. Rong Li

In some problems there is information about the destination of a moving object. An example is an airliner flying from an origin to a destination. Such problems have three main components: an origin, a destination, and motion in between. To emphasize that the motion trajectories end up at the destination, we call them \textit{destination-directed trajectories}. The Markov sequence is not flexible enough to model such trajectories. Given an initial density and an evolution law, the future of a Markov sequence is determined probabilistically. One class of conditionally Markov (CM) sequences, called the $CM_L$ sequence (including the Markov sequence as a special case), has the following main components: a joint endpoint density (i.e., an initial density and a final density conditioned on the initial) and a Markov-like evolution law. This paper proposes using the $CM_L$ sequence for modeling destination-directed trajectories. It is demonstrated how the $CM_L$ sequence enjoys several desirable properties for destination-directed trajectory modeling. Some simulations of trajectory modeling and prediction are presented for illustration.

Knowledge Graph



Sign up or login to leave a comment