Uncertainty-aware Short-term Motion Prediction of Traffic Actors for Autonomous Driving

TL;DR

This paper presents a deep learning approach for short-term traffic actor motion prediction in autonomous driving, incorporating uncertainty estimation and validated through real-world experiments and onboard testing.

Contribution

It introduces a novel raster-based deep learning model that predicts future traffic actor states while explicitly modeling uncertainty.

Findings

The approach improves prediction accuracy over baseline models.

Uncertainty estimates enhance decision-making safety.

Successful onboard testing demonstrates practical viability.

Abstract

We address one of the crucial aspects necessary for safe and efficient operations of autonomous vehicles, namely predicting future state of traffic actors in the autonomous vehicle's surroundings. We introduce a deep learning-based approach that takes into account a current world state and produces raster images of each actor's vicinity. The rasters are then used as inputs to deep convolutional models to infer future movement of actors while also accounting for and capturing inherent uncertainty of the prediction task. Extensive experiments on real-world data strongly suggest benefits of the proposed approach. Moreover, following completion of the offline tests the system was successfully tested onboard self-driving vehicles.