Abstract
In order to improve the growing serious traffic congestion on the junction of freeway and expressway (junction of road network in short), starting from the on-ramp metering, considering the main line and its on-ramp control requirements, the multi-agent collaborative was introduced. And the relativity of different segments of junction of road network was analyzed. Meanwhile, the correlation of different segments was calculated. With the consistency of control objectives for urbanized segments of expressway, a relative density model of multi-agent consistency was proposed. Then, on basis of multi-agent and fuzzy control theory, a multi-agent ramp cooperative control model based on fuzzy neural network was presented. Urbanized segments of Beijing-Tianjin-Tanggu Expressway were selected to validate the model. The results show that the model was effective. The method can stabilize the density of the urbanized expressway, and relief traffic congestion in the area.
Introduction
Traffic congestion at the junctions of freeways and expressways is a significant challenge for modern urban infrastructure. Traditional on-ramp metering strategies typically regulate traffic flow at individual entry points in isolation. However, traffic flow is inherently interdependent; congestion at one ramp frequently propagates to adjacent segments. To address these complexities, this research proposes a collaborative control system that treats road segments as intelligent agents working toward a shared objective of network-wide consistency.
Methodology
Multi-Agent Consistency Model
The expressway is partitioned into discrete sections, each managed by an autonomous agent. To facilitate cooperation, the system employs correlation functions to quantify the influence of a segment on its neighbors (up to six ramps in either direction). This data informs a "Relative Density" model, ensuring that agents prioritize the stability of the main line over localized flow maximization.
Fuzzy Neural Network (FNN) Architecture
Each agent utilizes a five-layer Fuzzy Neural Network to determine optimal ramp control rates. This hybrid approach combines the interpretability of fuzzy logic with the adaptive learning capabilities of neural networks. The layers facilitate the transition from raw traffic data to categorical linguistic variables (e.g., "Heavy Traffic"), apply a set of 49 fuzzy rules based on expert knowledge, and finally produce a precise control output.
Case Study: Beijing-Tianjin-Tanggu Expressway
The proposed model was validated using real-world traffic data from urbanized segments of the Beijing-Tianjin-Tanggu Expressway. The study focused on a specific corridor comprising segments labeled 13001 through 13007, located between the Fenzhongsi and Shibalidian Interchanges.
Results and Discussion
Simulation results in MATLAB demonstrated the model's efficacy in stabilizing traffic conditions. Key findings include:
- Density Stabilization: Traffic density in critical sections (e.g., 13002 and 13006) was significantly smoothed, preventing the erratic spikes characteristic of uncontrolled flow.
- Speed Recovery: Section 13001 exhibited a marked increase in average velocity compared to scenarios without collaborative control.
- Congestion Mitigation: The system showed a rapid response time, effectively preventing chronic congestion build-up during peak simulation periods in segments 13003, 13004, and 13007.
Conclusion
The multi-agent collaborative control method based on a Fuzzy Neural Network provides a robust solution for managing urban expressway junctions. By fostering consistency across segments, the model successfully relieves congestion and stabilizes traffic density. Future research will aim to scale this architecture to broader regional networks and incorporate microscopic variables such as individual driver behavior.
Summary updated