Recurrent Neural Networks for Driver Activity Anticipation via Sensory-Fusion Architecture

TL;DR

This paper presents a deep learning architecture using RNNs with LSTM units for early anticipation of driving maneuvers by fusing multiple sensory inputs, significantly improving prediction accuracy over previous methods.

Contribution

The authors introduce a novel sensory-fusion RNN architecture with a new loss layer for early and accurate driver activity anticipation in real-world driving scenarios.

Findings

Achieved 90.5% precision in maneuver anticipation

Achieved 87.4% recall in maneuver anticipation

Significantly outperformed previous state-of-the-art methods

Abstract

Anticipating the future actions of a human is a widely studied problem in robotics that requires spatio-temporal reasoning. In this work we propose a deep learning approach for anticipation in sensory-rich robotics applications. We introduce a sensory-fusion architecture which jointly learns to anticipate and fuse information from multiple sensory streams. Our architecture consists of Recurrent Neural Networks (RNNs) that use Long Short-Term Memory (LSTM) units to capture long temporal dependencies. We train our architecture in a sequence-to-sequence prediction manner, and it explicitly learns to predict the future given only a partial temporal context. We further introduce a novel loss layer for anticipation which prevents over-fitting and encourages early anticipation. We use our architecture to anticipate driving maneuvers several seconds before they happen on a natural driving data set of 1180 miles. The context for maneuver anticipation comes from multiple sensors installed on the vehicle. Our approach shows significant improvement over the state-of-the-art in maneuver anticipation by increasing the precision from 77.4% to 90.5% and recall from 71.2% to 87.4%.