Enhancing legal driving for autonomous vehicles through law-compliance potential fields

Area of Science:

• Autonomous Driving Systems
• Traffic Law Compliance
• Decision-Making Frameworks

Background:

• Mixed traffic environments necessitate autonomous vehicles (AVs) complying with traffic laws alongside human drivers.
• Current AV decision-making often prioritizes safety over systematic legal compliance, risking violations in complex scenarios.

Purpose of the Study:

• To develop a novel method for enhancing AV traffic law compliance.
• To address the limitations of existing safety-focused decision-making approaches in mixed traffic.

Main Methods:

• A Law-Compliance Potential Fields (LCPF) method is proposed, categorizing traffic law constraints into four potential fields.
• These fields encode vehicle states, environmental elements, and compliance thresholds.
• A novel fusion strategy resolves field-overlap distortions, integrated into a Model Predictive Control (MPC) framework.

Main Results:

• Experimental validation in five representative scenarios, including safety-compliance conflicts, was performed.
• The LCPF method significantly improved AVs' ability to comply with traffic laws.
• Effective balancing of safety and compliance was demonstrated, even in challenging conditions.

Conclusions:

• The proposed LCPF method offers a robust solution for ensuring AVs adhere to traffic laws.
• This approach is crucial for safe and legal AV operation in mixed traffic environments.
• Future work can explore further refinements and real-world deployment.