Law compliance decision making for autonomous vehicles on highways

Area of Science:

• Robotics
• Artificial Intelligence
• Autonomous Systems

Background:

• Autonomous vehicles (AVs) will soon share roads with human drivers, necessitating adherence to traffic laws.
• Current AVs prioritize safety, potentially causing distrust and accidents in mixed traffic.
• A need exists for AVs that can safely deviate from traffic laws in critical situations.

Purpose of the Study:

• To propose a novel decoupled hierarchical framework for compliance and safety decision-making in autonomous driving.
• To ensure AVs comply with traffic laws under normal conditions while prioritizing safety in emergencies.
• To enhance trust and predictability for human drivers interacting with AVs.

Main Methods:

• A two-layer framework: decision-making and motion planning.
• Decision-making layer: candidate behavior generation, dual-layer assessment, and optimal behavior selection.
• Motion planning layer: potential fields for surrounding vehicles and traffic laws integrated into Model Predictive Control (MPC).

Main Results:

• The framework ensures lane safety and compliance with static traffic laws at the decision-making level.
• The planning layer guarantees trajectory safety and dynamic traffic law compliance under safe conditions.
• Simulations demonstrated the method's effectiveness in ensuring compliance and enabling safe deviations during emergencies.

Conclusions:

• The proposed framework effectively balances traffic law compliance with safety requirements for autonomous vehicles.
• This approach can improve the interaction and trust between autonomous vehicles and human drivers.
• The method provides a robust solution for complex mixed-traffic scenarios.