Fuzzy Petri Nets for Traffic Node Reliability.

Area of Science:

• Robotics and Artificial Intelligence
• Control Systems Engineering
• Transportation Engineering

Background:

• Self-driving cars generate massive sensor data, posing challenges for traditional traffic management and validation.
• Current methods struggle with the complexity of interpreting numerous traffic junctions, impacting autonomous vehicle feasibility.
• The reliability of perception systems is critical for the safe operation of self-driving cars.

Purpose of the Study:

• To introduce Fuzzy Petri nets as a novel solution for managing large-scale traffic data in self-driving cars.
• To analyze traffic node safety and reliability using Petri nets and fuzzy logic.
• To demonstrate how Fuzzy Petri nets can improve the efficiency of deep learning perception models.

Main Methods:

• Utilizing modified Fuzzy Petri net procedures to model and analyze traffic node dynamics.
• Applying fuzzy analysis and Petri nets to assess the reliability of traffic nodes.
• Leveraging real traffic databases to inform the fuzzy extension of Petri nets.

Main Results:

• Fuzzy Petri nets provide a manageable model for vast amounts of traffic data.
• The method accurately describes node reliability through its dynamics, crucial for perception.
• A smaller deep learning mesh is required when node reliability is accurately determined.

Conclusions:

• Fuzzy Petri nets present a viable and economical solution for self-driving car data management and validation.
• The developed approach enhances the safety analysis of traffic nodes for autonomous systems.
• This research contributes to the advancement of reliable perception models in self-driving technology.