Multi-lane freeway traffic control with connected and automated vehicles

Highway bottlenecks caused by traffic incidents frequently trigger a systemic breakdown of traffic flow, leading to the capacity drop phenomenon and severe congestion. To mitigate this inefficiency, this thesis proposes a Sequential Model Predictive Control (MPC) strategy that uses a fleet of Connected and Autonomous Vehicles (CAVs) to actively regulate upstream traffic flow. The physical plant is simulated using a macroscopic Cell Transmission Model (CTM) integrated with a customized incentive-based lateral flow logic. Operating over a finite prediction horizon, the MPC dynamically optimizes the speed of the CAV fleet to trying maintaining bottleneck density below the critical threshold, thereby preserving maximum discharge throughput. To comprehensively evaluate the controller’s robustness, a rigorous sensitivity analysis was conducted during a simulated 60-minute lane-closure incident. The experiments varied both the spatial control horizon, CAV activation distances from 5 to 20 km upstream, and the fleet sizes, ranging from 6 to 10 vehicles. The simulation results demonstrate that the uncontrolled capacity drop imposes a systemic penalty of 363 vehicle-hours of delay. The optimal control configuration—deploying 10 CAVs 5 kilometers upstream of the incident—successfully recovered 94.7% of this delay, effectively neutralizing the capacity drop and forcing the network to operate almost at its theoretical physical limit. Furthermore, the analysis reveals a fundamental spatiotemporal trade-off in fleet deployment: activating control further upstream significantly increases the temporal footprint of each CAV, allowing smaller fleets to partially mitigate the incident. Conversely, proximity activation requires larger fleets to sustain temporal coverage but achieves higher absolute efficiency by minimizing initialization latency. Finally, to validate the real-world feasibility of the proposed system, a probabilistic demand model and topological Monte Carlo simulation were conducted. The analysis proves that the Sequential MPC framework does not require near-total autonomous adoption; instead, it achieves over 99% spatial reliability for forming the moving bottleneck at a modest CAV market penetration rate of just 6.3%. These findings provide highly practical, deployment-ready strategies for future traffic management centers operating under realistic near-term conditions.