Add to Chrome
Add to Firefox
Add to EdgeOn-Demand Trajectory Predictions for Interaction Aware Highway Driving
Sep 11, 2019
Highway driving places significant demands on human drivers and autonomous vehicles (AVs) alike due to high speeds and the complex interactions in dense traffic. Merging onto the highway poses additional challenges by limiting the amount of time available for decision-making. Predicting others' trajectories accurately and quickly is crucial to safely executing these maneuvers. Many existing prediction methods based on neural networks have focused on modeling interactions to achieve better accuracy while assuming the existence of observation windows over 3s long. This paper proposes a novel probabilistic model for trajectory prediction that performs competitively with as little as 400ms of observations. The proposed method fits a low-dimensional car-following model to observed behavior and introduces nonconvex regularization terms that enforce realistic driving behaviors in the predictions. The resulting inference procedure allows for realtime forecasts up to 10s into the future while accounting for interactions between vehicles. Experiments on dense traffic in the NGSIM dataset demonstrate that the proposed method achieves state-of-the-art performance with both highly constrained and more traditional observation windows.
