Smoothing traffic flow through automated vehicle control with optimal parameter selection
Shian Wang, Jose Acedo Aguilar, Miguel Velez-Reyes
TL;DR
This paper tackles stop-and-go traffic in mixed AV/HV flows by designing a class of additive AV controllers that require only local measurements and do not rely on equilibrium-speed assumptions. The authors propose tracking a virtual speed profile $\tilde{v}_i = v_{i-1} + \mathcal{I}(s_i, \Delta v_i)$, with $\mathcal{I}$ constrained by monotonicity, safety, differentiability, and boundedness, and provide an explicit example $\mathcal{I}(s_i, \Delta v_i) = \beta \arctan(\gamma s_i \Delta v_i)$. They develop a gradient-based method to select optimal parameters $\beta$ and $\gamma$ to minimize $J = \frac{1}{2}\int_0^{t_f} (v_i - v_{i-1})^2 dt$, ensuring AV safety via a derived upper bound on $\beta$; simulations with IDM HVs and OVVR AVs show substantial reductions in traffic oscillations (ASV) and modest energy savings (FC), especially at higher AV penetration, while maintaining safety. Although a recently proposed TS-TRC controller can offer slightly better performance by assuming knowledge of the equilibrium speed, TS-OPS emphasizes implementability by avoiding this assumption. Overall, the approach demonstrates practical traffic smoothing through localized AV control and systematic parameter optimization, with clear implications for scalable AV deployment and energy efficiency.
Abstract
Stop-and-go traffic waves are known for reducing the efficiency of transportation systems by increasing traffic oscillations and energy consumption. In this study, we develop an approach to synthesize a class of additive feedback controllers for automated vehicles (AVs) to smooth nonlinear mixed traffic flow, including both AVs and human-driven vehicles (HVs). Unlike recent explicit AV controllers that rely on strict assumptions such as time-varying equilibrium traffic speed, our proposed AV controller requires only local traffic information, such as inter-vehicle spacing and relative speed, which are readily available through AV onboard sensors. Essentially, it allows a controlled AV to track a subtler version of the perturbed speed profile resulting from its preceding vehicle, thereby enabling smoother traffic flow. Additionally, we provide a method for selecting the optimal control parameters to achieve traffic-smoothing effects efficiently. These unique features of the developed AV controller ensure much higher implementability. We demonstrate the effectiveness of the proposed approach through simulations of two distinct traffic scenarios with varying levels of oscillation. The results show that AVs using the proposed controller are capable of effectively reducing traffic oscillations and lowering vehicle fuel consumption by up to 46.78\% and 2.74\%, respectively, for a platoon of 10 vehicles. The traffic-smoothing effect of the controller is more pronounced at higher penetration rates of AVs. While the performance of the proposed approach is slightly less superior to that of the most recent additive AV controller, it offers greater implementability and provides an efficient method for selecting optimal control parameters.