Learn2Drive: A neural network-based framework for socially compliant automated vehicle control

TL;DR

Learn2Drive addresses socially aware automated vehicle control in mixed traffic by integrating a neural LSTM-based AV controller with physics-informed car-following dynamics and a Social Value Orientation (SVO) utility. The approach defines a mixed-traffic model where HVs follow physical car-following laws and the AV optimizes a $U = \cos(\phi) U_{self} + \sin(\phi) U_{collective}$ with $\phi \in [0, \frac{\pi}{2}]$, enabling egoistic to altruistic behaviors. An adaptive learning framework combines $\mathcal{L} = \alpha \mathcal{L}_{prediction} + \beta \mathcal{L}_{cost} + \gamma \mathcal{L}_{constraint}$ with $\mathcal{L}_{prediction} = \int_{0}^{T} (a_{pred}(t) - a_{true}(t))^2 dt$ and utility terms, guiding the AV to act as a mobile traffic regulator. Validation on the Arizona Ring Experiments Dataset shows that increasing social emphasis ($\phi$ toward $\frac{\pi}{2}$) yields notable improvements in follower speeds and traffic stability, at the cost of higher AV energy expenditure, demonstrating a practical energy-traffic efficiency trade-off and the potential for scalable socially compliant control in larger networks with further work.

Abstract

This study introduces a novel control framework for adaptive cruise control (ACC) in automated driving, leveraging Long Short-Term Memory (LSTM) networks and physics-informed constraints. As automated vehicles (AVs) adopt advanced features like ACC, transportation systems are becoming increasingly intelligent and efficient. However, existing AV control strategies primarily focus on optimizing the performance of individual vehicles or platoons, often neglecting their interactions with human-driven vehicles (HVs) and the broader impact on traffic flow. This oversight can exacerbate congestion and reduce overall system efficiency. To address this critical research gap, we propose a neural network-based, socially compliant AV control framework that incorporates social value orientation (SVO). This framework enables AVs to account for their influence on HVs and traffic dynamics. By leveraging AVs as mobile traffic regulators, the proposed approach promotes adaptive driving behaviors that reduce congestion, improve traffic efficiency, and lower energy consumption. Within this framework, we define utility functions for both AVs and HVs, which are optimized based on the SVO of each AV to balance its own control objectives with broader traffic flow considerations. Numerical results demonstrate the effectiveness of the proposed method in adapting to varying traffic conditions, thereby enhancing system-wide efficiency. Specifically, when the AV's control mode shifts from prioritizing energy consumption to optimizing traffic flow efficiency, vehicles in the following platoon experience at least a 58.99% increase in individual energy consumption alongside at least a 38.39% improvement in individual average speed, indicating significant enhancements in traffic dynamics.

Figures (5)

• Figure 1: Illustration of traffic flow with AVs and HVs, highlighting the role of a single AV as a key coordinator to guide traffic movement, reduce congestion, and enhance system efficiency.
• Figure 2: Partial trajectories of speed and inter-vehicle spacing for the first five vehicles in the ARED dataset, illustrating the platoon's car-following dynamics.
• Figure 3: Trajectories of speed and inter-vehicle spacing for a five-vehicle platoon, with real data for the first and second vehicles and calibrated IDM for the third, fourth, and fifth vehicles, validating the accuracy of the IDM in replicating realistic car-following dynamics.
• Figure 4: Trajectories of speed and spacing under different $\phi$ values, illustrating the impact of social preference settings on traffic flow dynamics. At $\phi = 0$, $v_2$ adopts a conservative strategy, lagging behind $v_1$ dynamics (1.0-2.0 m/s), while $\phi = \pi/4$ and $\phi = \pi/2$ show improved matching and smoother spacing, reflecting a shift from individual to system optimization based on SVO styles: egoistic ($\phi = 0$), balanced ($\phi = \pi/4$), and altruistic ($\phi = \pi/2$).
• Figure 5: Energy consumption and average speed under different $\phi$ values, highlighting the trade-offs in the Learn2Drive framework. $v_2$ increased energy cost at $\phi = \pi/4$ and $\phi = \pi/2$ correlates with enhanced speed gains for followers, reflecting SVO styles: egoistic ($\phi = 0$), balanced ($\phi = \pi/4$), and altruistic ($\phi = \pi/2$).