TraInterSim: Adaptive and Planning-Aware Hybrid-Driven Traffic Intersection Simulation.

Area of Science:

• Traffic simulation
• Computational modeling
• Artificial intelligence in transportation

Background:

• Traditional traffic simulation methods struggle with complex intersection dynamics due to undefined lanes.
• Data-driven approaches require extensive high-quality datasets and significant parameter tuning for intersection scenarios.

Purpose of the Study:

• To develop an adaptive and planning-aware hybrid-driven method for realistic traffic intersection simulation.
• To address the limitations of existing methods in handling agent convergence and diverse motion planning at intersections.

Main Methods:

• Introduced TraInterSim, a hybrid-driven approach combining an optimization-based data-driven scheme with a velocity continuity model.
• Integrated planning-aware collision avoidance, considering agent motion plans and relative distances.
• Implemented automatic parameter adjustment to maintain individual agent flexibility.

Main Results:

• TraInterSim generates realistic behaviors for heterogeneous agents in various intersection scenarios.
• The method achieves interactive simulation rates.
• Validated effectiveness and rationality through extensive experiments and user studies.

Conclusions:

• The proposed TraInterSim method offers a robust solution for simulating complex traffic intersection dynamics.
• Its adaptive and planning-aware nature enhances realism and efficiency in traffic simulation.