Introduction
Urban traffic management frequently employs perimeter control to maintain the stability of city centers, often referred to as Protected Networks (PNs). By "metering" traffic at the edges of these zones, authorities aim to prevent the interior from reaching a state of gridlock. While the theoretical foundations for determining optimal inflow—based on the Network Fundamental Diagram—are well-established, practical implementation at the signalized intersection level remains a challenge.
Traditional methods often rely on scaling down green light durations based on volume or adding redundant traffic lights, which can lead to imprecise flow regulation and inefficient queue spillbacks. This paper proposes a novel Auction-Based Signalling framework that leverages connected vehicle (CV) technology to bridge the gap between macro-level perimeter control and micro-level intersection management.
Auction-Based Signalling Framework
The core of the proposed system is a Sealed Bid, Second Price (SBSP) auction integrated with a flow-metering budget. In this model, each traffic signal phase acts as a "wallet" that aggregates bids from approaching connected vehicles. Vehicles submit bids based on their individual Value of Time (VOT) and an impatience function that increases the longer they wait at the intersection.
As shown in the intersection layout, specific movements are designated as "gated inflows" entering the protected network. The intersection manager tracks the cumulative number of vehicles passing through these gated movements within a set budget period (e.g., 5 minutes). Once the pre-defined maximum allowable flow count is reached, the movements associated with that inflow (such as movements 3 and 4 in Phase 3 and Phase 4) are excluded from the auction for the remainder of the period. This "precision through exclusion" ensures the gate closes exactly when the budget is spent, offering tighter regulation than traditional green-time adjustments.
Experimental Methodology
The researchers evaluated the system using microscopic traffic simulation in AIMSUN. The test environment consisted of a synthetic four-way intersection with 750-meter arms, each featuring four lanes. To test the system's sensitivity and robustness, demand was set at 800 vehicles per hour (veh/hr) per arm, while the hourly inflow limits for the gated movements were varied from highly restrictive (100 veh/hr) to essentially unconstrained (700 veh/hr).
Key Results and Performance Analysis
The auction-based method demonstrated significant improvements in both flow accuracy and operational efficiency compared to a standard volume-based benchmark.
Flow Accuracy and Precision
One of the primary advantages of the auction-based approach is its precision. For most tested limits, the delivered inflow remained within 30–40 veh/hr of the target. In contrast, the volume-based benchmark consistently overshot its targets, particularly at less restrictive limits where driver behavior and saturation flow variability made simple green-time scaling less effective.
Queue Management and Delay Reduction
The system also showed a superior ability to manage queue formation during the "Perimeter Active" control window.
As illustrated in the queue length comparison, the auction-based method allows queues to grow predictably and dissipate rapidly once the control period ends. Despite enforcing the same strict flow constraints, the auction-based approach generally resulted in lower mean vehicle delays. This efficiency stems from the system's ability to prioritize "high-value" users—those with higher urgency—even while overall flow is being restricted. Furthermore, at high flow limits where gating is unnecessary, the system reverts to a high-efficiency state that outperforms fixed-time control.
Limitations and Future Directions
While promising, the study identifies several areas for future research:
- Mixed Movement Phases: Currently, if a signal phase serves both a gated and a non-gated movement, the entire phase may be excluded, potentially penalizing vehicles not entering the protected zone.
- Connected Vehicle Penetration: The model assumes a high penetration of CVs. Real-world deployment would require strategies to account for legacy, non-connected vehicles.
- Transition Dynamics: The current study used abrupt "step-changes" in flow limits. Future iterations could explore gradual limit adjustments to prevent sharp flow surges at the start of new budget cycles.
By combining market-driven prioritization with strict vehicle-count budgets, this auction-based framework offers a more precise and equitable tool for urban traffic gating.
Summary updated