A Mixed Platoon Guided Control Model for Efficient and Safer Merging in Mixed Traffic With Low Penetration of Autonomous Vehicles

Traditional ramp merging relies on drivers' observation and cooperation, which creates efficiency bottlenecks and accident hotspots. Existing mixed-traffic strategies offer improvements, many remain limited by requirements for high AV penetration and strong driver cooperation. This paper introduces the Mixed Platoon Guided Control (MPGC) model, a hierarchical AV-guided merging framework that combines protocol, prediction, and optimisation within a rolling MPC structure. MPGC applies a platoon-interleaving mechanism where AV leaders manage most merging actions, allowing human-driven followers to simply follow without complex decision-making. The framework integrates a dual-rule protocol for dynamic traffic, a data-driven prediction layer accounting for human driving heterogeneity, and a trajectory optimisation layer based on merging time window search. SUMO simulations demonstrate that at 10% AV penetration, MPGC increases throughput by up to 33% and 12% compared to FIFO and RM, respectively. For safety, MPGC achieves an 86% reduction in risky time-to-collision (TTC) events compared to FIFO, reaching a safety level comparable to RM. The system also remains robust under human driving heterogeneity, incomplete platoon formation, and V2X failures, showing good scalability in low-penetration scenarios.