Low Latency Trajectory Predictions for Interaction Aware Highway Driving
Cyrus Anderson · Ram Vasudevan · Matthew Johnson-Roberson
TL;DR
A probabilistic car-following model predicts nearby vehicles' trajectories from just 400 ms of observation, matching state-of-the-art accuracy on dense NGSIM highway traffic.
BRIEF
Merging onto a highway leaves an autonomous vehicle very little time to decide, yet many trajectory predictors need three seconds or more of observed motion before they work well. This work starts from a deterministic car-following model and makes it probabilistic by treating its parameters as unknown random variables, with added regularization. A real-time inference procedure estimates those parameters from a short observation. On dense NGSIM traffic, it performs competitively with only 400 ms of observations and reaches state-of-the-art results with longer windows too.
Abstract
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 execute maneuvers safely. 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 3 s long. This letter proposes a novel probabilistic model for trajectory prediction that performs competitively with as little as 400 ms of observations. The proposed model extends a deterministic car-following model to the probabilistic setting by treating model parameters as unknown random variables and introducing regularization terms. A realtime inference procedure is derived to estimate the parameters from observations in this new model. 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.