Safer and Smoother Traffic with Knowledge-Guided Self-Learning Platoon Control

1. Executive Summary

The coexistence of human-driven vehicles and autonomous vehicles will remain a reality for decades, creating complex challenges for traffic flow, road safety, and energy efficiency. Human drivers introduce unpredictable actions such as sudden braking, acceleration, or lane changes, while autonomous vehicles rely on consistent algorithms and require stable communication channels to operate effectively. Mixed platoons, where both vehicle types travel together, often suffer from traffic oscillations, increased fuel consumption, and safety risks. The research introduces a knowledge-guided self-learning control strategy that enables connected autonomous vehicles to anticipate human behavior and compensate for wireless communication delays. By integrating classic traffic flow models with reinforcement learning, the system can predict collective human-driven vehicle behavior, adapt to delays, and stabilize platoons under varying conditions. Simulation results demonstrate improvements in fuel efficiency, ride comfort, and traffic stability, achieving zero collisions in merging and diverging scenarios. This strategy provides governments, mobility companies, and investors with a scalable solution for safer, more efficient roads, marking a significant step toward the practical adoption of connected vehicle ecosystems.