Knowledge-guided self-learning control strategy for mixed vehicle platoons with delays.

Area of Science:

• Intelligent Transportation Systems
• Control Engineering
• Artificial Intelligence

Background:

• Autonomous and traditional vehicles will coexist for decades, posing challenges for mixed traffic management.
• Communication delays in connected autonomous vehicles degrade platooning control performance.
• Heterogeneity and randomness of traditional vehicles complicate traffic flow.

Purpose of the Study:

• To propose a knowledge-guided self-learning mixed platoon control strategy.
• To enhance road throughput, fuel consumption, and traffic stability in mixed traffic environments.
• To address challenges posed by communication delays and traditional vehicle behavior.

Main Methods:

• Integrating kinematic wave and Newell's car-following models to predict traditional vehicle behavior.
• Extracting key features like time gap and standstill spacing from traditional vehicles.
• Incorporating previous control instructions into the state representation of the soft actor-critic algorithm to handle delayed information.

Main Results:

• Outperformed existing methods in traffic stability, passenger comfort, and energy consumption.
• Demonstrated significant dampening of traffic oscillations.
• Achieved a zero collision rate in vehicle merging and diverging scenarios.

Conclusions:

• The proposed strategy offers a generalizable and scalable solution for connected autonomous vehicle systems.
• Effectively manages mixed traffic by predicting traditional vehicle trajectories and compensating for communication delays.
• Significantly improves overall traffic efficiency and safety.