About this document

Motion by Divyesh is a document available to read on EtoBox.

This paper proposes a new model called GNeVA (Goal-based Neural Variational Agent) for interpretable and generalizable motion prediction. GNeVA predicts motion in two steps: 1) Estimating the spatial distribution of long-term destinations using a variational mixture of Gaussians, and 2) Completing the trajectory from the current location to the destinations. This allows GNeVA to model the uncertainty in multi-modal driver behaviors. The model identifies a causal structure where map features determine goal l

Author

Divyesh

Language

EN