Optimal Safe Sequencing and Motion Control for Mixed Traffic Roundabouts

TL;DR

The paper addresses safe and efficient navigation of mixed HDV/CAV traffic through a single-lane roundabout by developing the Optimal Safe Sequencing (OSS) framework, which combines a Safe Sequence (SS) policy with an MPC-CLBF controller. The SS policy robustly restricts CAV merging ahead of HDVs to maintain safety without modeling HDV dynamics, while the MPC-CLBF component jointly optimizes sequencing and longitudinal motion to balance travel time, energy, and comfort. Key contributions include a method to generate feasible, HDV-robust sequences, a CLBF-based transformation for safety constraints, and an MPC horizon that mitigates myopic control, all demonstrated across varying traffic demands and CAV penetration. Results show substantial improvements in energy efficiency and safety metrics with OSS, at a modest increase in travel time, highlighting the method’s potential for real-time deployment in mixed-traffic environments as CAV penetration grows.

Abstract

This paper develops an Optimal Safe Sequencing (OSS) control framework for Connected and Automated Vehicles (CAVs) navigating a single-lane roundabout in mixed traffic, where both CAVs and Human-Driven Vehicles (HDVs) coexist. The framework jointly optimizes vehicle sequencing and motion control to minimize travel time, energy consumption, and discomfort while ensuring speed-dependent safety guarantees and adhering to velocity and acceleration constraints. This is achieved by integrating (a) a Safe Sequencing (SS) policy that ensures merging safety without requiring any knowledge of HDV behavior, and (b) a Model Predictive Control with Control Lyapunov Barrier Functions (MPC-CLBF) framework, which optimizes CAV motion control while mitigating infeasibility and myopic control issues common in the use of Control Barrier Functions (CBFs) to provide safety guarantees. Simulation results across various traffic demands, CAV penetration rates, and control parameters demonstrate the framework's effectiveness and stability.

Figures (7)

• Figure 1: A roundabout with 3 entries
• Figure 2: Illustration of $z_{i,i_{p}}$ definition for $i=1$, $i_p=0$
• Figure 3: Illustration of MPC framework
• Figure 4: Performance comparison for different policies under CAV penetration rate 0.8
• Figure 5: Comparison of trajectories for BS and SS policy: (a) is the partial roundabout state under BS policy at time 57.2s, 61.8s, 65.4s; (b) is the partial roundabout state under SS policy at time 55.6s, 57.7s, 60.1s. In both plots, red agents indicates CAV while blue agents indicates HDV. Agents with star shape indicates violation of safety constraint. (c) is the velocity trajectories under both policies for vehicle 22 and 24 respectively.

Theorems & Definitions (1)

• Remark 1