Passive and Active Learning of Driver Behavior from Electric Vehicles

TL;DR

This work tackles two challenges in driver behavior modeling for electric vehicles: efficient sequence modeling on short time windows and the cost of annotating driving data. It compares passive learning methods using non-recurrent architectures such as self-attention and CNNs with joint recurrence plots against recurrent LSTM models, and evaluates active learning strategies including uncertainty sampling, query-by-committee, and active deep dropout. Across two datasets and 5–10 second windows, LSTM generally delivers the best performance, while non-recurrent approaches offer mixed results and require more compute for self-attention. The study demonstrates that active learning can reduce labeling effort, with uncertainty sampling frequently outperforming random sampling, guiding practical deployment in ADAS and EV energy-consumption estimation workflows.

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.

Figures (2)

• Figure 1: Comparison of uncertainty sampling methods with random sampling on the (a) 1D-CNN and (b) LSTM models.
• Figure 2: Comparison of query by committee methods with random sampling on the (a) 1D-CNN and (b) LSTM models.