Deep Predictive Models for Collision Risk Assessment in Autonomous Driving

TL;DR

This paper presents a deep predictive model for collision risk assessment in autonomous driving, utilizing multi-modal, temporal, and uncertainty-aware data to predict accidents from video streams.

Contribution

It introduces a Bayesian Convolutional LSTM model that incorporates temporal, multi-modal, and uncertainty information for improved collision risk prediction in autonomous vehicles.

Findings

Model predicts accidents with reasonable accuracy.

Multi-camera inputs improve prediction performance.

Incorporating uncertainty enhances decision-making.

Abstract

In this paper, we investigate a predictive approach for collision risk assessment in autonomous and assisted driving. A deep predictive model is trained to anticipate imminent accidents from traditional video streams. In particular, the model learns to identify cues in RGB images that are predictive of hazardous upcoming situations. In contrast to previous work, our approach incorporates (a) temporal information during decision making, (b) multi-modal information about the environment, as well as the proprioceptive state and steering actions of the controlled vehicle, and (c) information about the uncertainty inherent to the task. To this end, we discuss Deep Predictive Models and present an implementation using a Bayesian Convolutional LSTM. Experiments in a simple simulation environment show that the approach can learn to predict impending accidents with reasonable accuracy, especially when multiple cameras are used as input sources.