Bi-Level Control of Weaving Sections in Mixed Traffic Environments with Connected and Automated Vehicles

TL;DR

This work addresses the challenge of improving traffic operation in highway weaving sections under mixed traffic by introducing a bi-level coordination scheme that couples a roadside upper-level DRL controller with per-vehicle MPC lower-level controllers. The upper level determines global MPC weightings across the weaving section, while the lower level computes ego accelerations, steering, and lane changes, informed by HV trajectory predictions. An HV trajectory predictor based on EvolveGCN, augmented with GIN, CBAM, and CNN components, handles the dynamic topology of HVs. Simulation results show that global coordination (Bi-Level-G) consistently outperforms baselines, significantly reducing capacity drops and increasing space-mean speed and exit flow, with greater gains at higher CAV penetrations. The approach demonstrates strong potential for scalable, coordinated CAV deployment in mixed-traffic weaving scenarios and lays groundwork for extension to other maneuvers and more general conditions.

Abstract

Connected and automated vehicles (CAVs) can be beneficial for improving the operation of highway bottlenecks such as weaving sections. This paper proposes a bi-level control approach based on an upper-level deep reinforcement learning controller and a lower-level model predictive controller to coordinate the lane-changings of a mixed fleet of CAVs and human-driven vehicles (HVs) in weaving sections. The upper level represents a roadside controller that collects vehicular information from the entire weaving section and determines the control weights used in the lower-level controller. The lower level is implemented within each CAV, which takes the control weights from the upper-level controller and generates the acceleration and steering angle for individual CAVs based on the local situation. The lower-level controller further incorporates an HV trajectory predictor, which is capable of handling the dynamic topology of vehicles in weaving scenarios with intensive mandatory lane changes. The case study inspired by a real weaving section in Basel, Switzerland, shows that our method consistently outperforms state-of-the-art benchmarks.

Figures (8)

• Figure 1: Framework of the global coordinated bi-level controller in weaving sections.
• Figure 2: Illustration of circle decomposition.
• Figure 3: Framework of the trajectory prediction algorithm.
• Figure 4: Principle diagram of EvolveGCN layer. The green vehicle represents SV and blue vehicles indicates SV's surrounding HVs within the communication range.
• Figure 5: Queuing diagrams with background flow of 4800 veh/h and fundamental diagrams under different CAV penetration rates. Notice that VD represents the virtual departure curve at the exit of the weaving section, which is obtained from the simulated traffic demand based on the empirical data.