Abstract
This paper introduces FUZZY-MIX, a novel two-layer fuzzy logic controller designed to manage real-time traffic signals at modern roundabouts. By combining the responsiveness of "phase jumping" with the stability of "phase rotation" and optimizing the system using Particle Swarm Optimization (PSO), the researchers achieved a 20% reduction in vehicle delays compared to standard fuzzy methods and significantly outperformed traditional vehicle-actuated systems.
Introduction: The Roundabout Bottleneck
Modern roundabouts are highly efficient for moderate traffic volumes, but as urban populations swell, they frequently become significant bottlenecks. While traffic signals can mitigate these issues, most existing control methods are either "fixed-time"—failing to adapt to sudden traffic spikes—or designed for standard four-way intersections rather than the unique, complex geometry of a roundabout.
Existing real-time fuzzy controllers generally fall into two categories: FUZZY-TURN, which decides whether to extend the current green light (stable but slow to react), and FUZZY-JUMP, which immediately switches to the most "urgent" direction (responsive but prone to causing inconsistent traffic flows and potential gridlock). The authors sought a "middle way" that captures responsiveness without the chaotic transitions that lead to traffic jams.
Methodology: The FUZZY-MIX Architecture
The core contribution of this research is the FUZZY-MIX architecture, a hierarchical fuzzy system designed to balance stability and agility:
- The Outer Layer (The Strategist): This layer evaluates the "urgency" of different phase subsets (e.g., North-South vs. East-West). It determines which group of lanes should have priority at any given moment.
- The Inner Layer (The Tactician): Once a subset is active, this layer calculates the "extension time" for the current green light, ensuring smooth transitions between specific lanes within that subset.
PSO-Based Optimization
Traditionally, fuzzy "membership functions" (definitions of what constitutes a "long" queue or "short" wait) are set manually by engineers. To remove human bias, the authors utilized Particle Swarm Optimization (PSO)—an algorithm inspired by bird flocking—to automatically tune these functions offline. This ensures the controller is mathematically optimized to minimize both vehicle delays and queue overflows.
Experimental Results
The researchers evaluated the controller using a custom-built simulation environment across 16 distinct traffic conditions, including steady-state and time-varying "rush hour" scenarios. The optimized controller, FUZZY-MIX-OPT, was compared against Vehicle Actuated (VA), FUZZY-TURN, and FUZZY-JUMP methods.
- Average Delay Reduction: FUZZY-MIX-OPT achieved a mean delay of 25.97 seconds, compared to 57.94 seconds for the standard VA method—a reduction of over 50%.
- Queue Management: In 14 out of 16 test cases, the optimized controller resulted in zero undetected vehicles, meaning queues never exceeded sensor range. In contrast, the VA method saw over 10,000 undetected vehicles in the heaviest traffic scenario.
- Efficiency Gain from Optimization: Moving from a standard FUZZY-MIX to the PSO-optimized version reduced vehicle delays by approximately 20%.
Conclusion and Future Work
The study demonstrates that a two-layer fuzzy approach effectively manages roundabout traffic by "jumping" between major directions while "turning" through specific lane phases. However, the authors acknowledge that physical capacity remains a hard limit; in extreme saturation scenarios, signal timing alone cannot prevent all queue overflows. Future research will focus on multi-agent systems to coordinate signals across wider networks of multiple roundabouts and intersections to prevent downstream congestion.
Summary updated