Passive and Active Learning of Driver Behavior from Electric Vehicles

TL;DR

This paper compares passive and active learning methods for modeling driver behavior in electric vehicles, evaluating neural network architectures and sampling strategies to improve energy consumption prediction with less labeled data.

Contribution

It introduces the use of non-recurrent models like self-attention and convolutional neural networks with JRP for driver behavior modeling and assesses active learning techniques to reduce annotation effort.

Findings

Self-attention models perform well but do not outperform recurrent models at tested window lengths.

JRP does not significantly improve model performance.

Active sampling methods can reduce annotation effort compared to random sampling.

Abstract

Modeling driver behavior provides several advantages in the automotive industry, including prediction of electric vehicle energy consumption. Studies have shown that aggressive driving can consume up to 30% more energy than moderate driving, in certain driving scenarios. Machine learning methods are widely used for driver behavior classification, which, however, may yield some challenges such as sequence modeling on long time windows and lack of labeled data due to expensive annotation. To address the first challenge, passive learning of driver behavior, we investigate non-recurrent architectures such as self-attention models and convolutional neural networks with joint recurrence plots (JRP), and compare them with recurrent models. We find that self-attention models yield good performance, while JRP does not exhibit any significant improvement. However, with the window lengths of 5 and 10 seconds used in our study, none of the non-recurrent models outperform the recurrent models. To address the second challenge, we investigate several active learning methods with different informativeness measures. We evaluate uncertainty sampling, as well as more advanced methods, such as query by committee and active deep dropout. Our experiments demonstrate that some active sampling techniques can outperform random sampling, and therefore decrease the effort needed for annotation.