EditFollower: Tunable Car Following Models for Customizable Adaptive Cruise Control Systems

TL;DR

The paper tackles the challenge of fixed-parameter ACC by introducing Editable Behavior Generation (EBG), a data-driven car-following framework that enables tunable driving discourtesy. By encoding three discourtesy calculations—acceleration-based, jerk-based, and speed-based—into LSTM and Transformer architectures, EBG can synthesize realistic FV trajectories conditioned on a desired courtesy level. Experiments on HighD and Waymo show state-of-the-art reductions in spacing and speed errors and strong alignment with target courtesy, demonstrating controllability of driving style. The work offers a path toward socially aware ACC systems that adapt to individual driver preferences, improving acceptance and safety in real-world deployments.

Abstract

In the realm of driving technologies, fully autonomous vehicles have not been widely adopted yet, making advanced driver assistance systems (ADAS) crucial for enhancing driving experiences. Adaptive Cruise Control (ACC) emerges as a pivotal component of ADAS. However, current ACC systems often employ fixed settings, failing to intuitively capture drivers' social preferences and leading to potential function disengagement. To overcome these limitations, we propose the Editable Behavior Generation (EBG) model, a data-driven car-following model that allows for adjusting driving discourtesy levels. The framework integrates diverse courtesy calculation methods into long short-term memory (LSTM) and Transformer architectures, offering a comprehensive approach to capture nuanced driving dynamics. By integrating various discourtesy values during the training process, our model generates realistic agent trajectories with different levels of courtesy in car-following behavior. Experimental results on the HighD and Waymo datasets showcase a reduction in Mean Squared Error (MSE) of spacing and MSE of speed compared to baselines, establishing style controllability. To the best of our knowledge, this work represents the first data-driven car-following model capable of dynamically adjusting discourtesy levels. Our model provides valuable insights for the development of ACC systems that take into account drivers' social preferences.

Figures (6)

• Figure 1: Overview: simulating tunable car following behaviors. The model utilizes real-world car-following data to synthesize a simulated following vehicle's future trajectory based on an input discourtesy value.
• Figure 2: LSTM_IDM Car-following model architecture. Observed time series data, including FV speed, relative speed, and spacing are fed into LSTM cells to capture the temporal dependencies. The LSTM model then uses these extracted features to estimate IDM parameters.
• Figure 3: HighD dataset.
• Figure 4: Waymo dataset.
• Figure 5: Distribution changes when adjusting discourtesy levels. The probability density distribution of the time gap between LV and FV shifts to the left when the discourtesy level changes from 50% to 150%, indicating more aggressive driving behaviors.