🤖 AI

A QUBO-Inspired Computational Framework for Airport Landside Bottleneck Diagnosis and Dynamic Dispatch Optimization

This paper proposes a QUBO-inspired computational framework for diagnosing airport landside bottlenecks and optimizing dynamic dispatch, demonstrating through case studies at Shanghai Pudong and Hangzhou Xiaoshan airports that the method effectively reduces passenger queues and adapts to varying congestion patterns and uncertainties compared to traditional model predictive control.

Original authors: Wuming Lei, Xiaobin Li, Mingyan Sun, Jianing Long, Yulin Tong, Yanbin Gao

Published 2026-08-11

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Wuming Lei, Xiaobin Li, Mingyan Sun, Jianing Long, Yulin Tong, Yanbin Gao

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a busy airport not just as a place for planes, but as a giant, living organism where thousands of people and vehicles are constantly trying to move in and out. The "landside" is the ground-level part of this organism—the area where passengers step off the plane and try to squeeze into taxis, ride-share cars, buses, or their own private vehicles. Think of it like a massive, chaotic dance floor where the music is the arrival of flights, and the dancers are people and cars all trying to find a partner and exit the room at the same time. When too many people show up at once (a "peak"), the dance floor gets crowded, people get stuck in lines, and cars get backed up on the roads leading away from the terminal. This isn't just an annoyance; it's a complex traffic puzzle. Scientists who study this use tools like "queuing theory" (math that predicts how lines form) and "optimization" (finding the best way to arrange things). The big question is: when the chaos hits, where exactly is the jam? Is it because there aren't enough cars? Is it because the pickup spots are too small? Or is it because the roads leading away are already full? Figuring this out is crucial because if you try to fix the wrong part of the problem, you might just make the traffic worse.

This paper introduces a new, clever computer framework designed to act like a detective for these airport traffic jams. The researchers, working with data from Shanghai Pudong and Hangzhou Xiaoshan airports, built a digital simulation that runs in five-minute chunks to see how passengers, vehicles, and roads interact. They created a system to diagnose exactly what's causing the bottleneck—whether it's a lack of cars, a shortage of pickup spots, or a saturated road. Then, they tested two different "traffic controllers" to see which one could clear the lines best. One controller was a standard, rule-based method called Model Predictive Control (MPC), which plans a few steps ahead. The other was a more experimental, "QUBO-inspired" method (a fancy term for a type of math problem-solving technique often used in quantum computing, but here simulated on a regular computer) that uses a search strategy similar to "simulated annealing"—think of it as shaking a box of puzzle pieces until they fall into the perfect shape.

The results of their simulations were quite revealing. They found that the two airports had very different problems. At Shanghai Pudong, the main issue was that the roads leading away from the airport were getting completely clogged; the traffic was so heavy that the roads couldn't handle the flow. In contrast, at Hangzhou Xiaoshan, the roads were fine, but the specific spots where passengers meet their drivers (the "berths") were the bottleneck. When they tested their new QUBO-inspired controller against the standard one, the new method did a better job of reducing the number of people stuck in line. In the worst-case "strong peak" scenario at Shanghai Pudong, the new method reduced the final line of waiting passengers from 3,445 down to 2,477. At Hangzhou Xiaoshan, it dropped the line from 2,053 to 1,482. The study suggests that while the standard controller is good and easy to understand, this new, more flexible search method can squeeze out more efficiency, especially when the traffic patterns are complex. However, the authors are careful to note that these are results from computer simulations, not real-world tests on actual airports yet. They also showed that their new method remained effective even when they added random "noise" or changes to the data, suggesting it's a robust tool for the future of airport management.

Technical Summary: A QUBO-Inspired Computational Framework for Airport Landside Bottleneck Diagnosis and Dynamic Dispatch Optimization

Problem Statement
Airport landside traffic centers represent a complex, tightly coupled system where passenger arrivals interact with diverse vehicle modes (taxis, ride-hailing, private cars, buses), pickup berths, storage areas, and terminal-area roadways. During peak periods, these interactions often lead to coupled congestion across passenger queues, vehicle queues, and access roads. A critical challenge is that congestion can stem from different binding resources—such as insufficient vehicle supply, inadequate berth service, limited storage, or downstream road saturation. Analyzing these constraints in isolation can lead to misleading operational conclusions. Furthermore, emerging mobility services have altered access patterns, necessitating adaptive management strategies that can diagnose the active bottleneck and optimize dispatch policies dynamically.

Methodology
The study proposes an integrated computational framework that unifies demand construction, state propagation, bottleneck diagnosis, and dispatch optimization. The core components are:

Integrated State Model: A five-minute time-step model simulates the landside environment. It tracks passenger queues ( $Q_p$ ), vehicle queues ( $Q_v$ ), and road queues ( $Q_r$ ) for various transport modes. The model incorporates:
- Service Constraints: Terminal-side service is limited by vehicle availability, facility capacity (berths/curbs), and passenger availability.
- Roadway Discharge: Modeled as a shared downstream constraint where loaded vehicles from all modes compete for road capacity ( $C_r$ ). The system uses a mechanism to release carried-over queues first, then proportionally share remaining capacity, ensuring realized saturation does not exceed capacity while recording demand-side pressure.
- Impedance: Uses the Bureau of Public Roads (BPR) volume-delay function to approximate speed degradation based on volume-capacity ratios.
Bottleneck Diagnosis: The framework identifies four distinct bottleneck types: vehicle supply, berth facility, demand-limited, and road saturation. Diagnosis relies on:
- Bottleneck Frequency: The share of time-mode cells where a specific resource is the binding constraint.
- Severity: Accumulated positive passenger-queue increments associated with the bottleneck.
- Shadow-Price Leverage: Measuring the reduction in total weighted delay when a resource is perturbed by 10%.
- Composite Congestion Severity Index (CSI): A metric combining road over-saturation intensity, peak passenger backlog, and average passenger waiting relative to thresholds.
Dynamic Dispatch Schemes: Two optimization methods are compared under consistent inputs:
- Finite-Action Model Predictive Control (MPC): Evaluates candidate actions (berth allocation, release factors, modal guidance) over a six-step horizon. It minimizes a cost function including queues, waiting times, congestion, and fairness.
- QUBO-Inspired Simulated Annealing: Discretizes dispatch decisions into binary choices (berth allocation, release factors, modal shifts). The operational cost is formulated as an energy function within a Quadratic Unconstrained Binary Optimization (QUBO) framework. Simulated annealing is used to search this energy landscape. This approach is "quantum-inspired" but does not require quantum hardware.

Key Contributions
The paper makes three primary contributions:

Unified State Modeling: Formulation of a passenger-vehicle-facility-road state model at a five-minute resolution where all queues and capacities are updated within a single simulation loop.
Diagnostic Layer: Development of a diagnostic mechanism that distinguishes between vehicle-supply, berth-facility, demand-limited, and road-saturation bottlenecks, allowing for targeted policy evaluation.
Comparative Dispatch Analysis: A direct comparison of finite-action MPC and QUBO-inspired simulated annealing. The study reports interpretable indicators for queue reduction, waiting times, saturation levels, and robustness under uncertainty.

Experimental Results
The framework was tested using Shanghai Pudong International Airport (a large hub with high road pressure) and Hangzhou Xiaoshan International Airport (a regional hub where berth capacity is critical) under strong-peak baseline scenarios.

Bottleneck Diagnosis:
- Shanghai Pudong: Dominated by road saturation. Road demand saturation reached 3.892, accounting for 63.9% of bottleneck cells.
- Hangzhou Xiaoshan: Dominated by berth facility pressure. Ride-hailing service intensity reached 6.049, with berth bottlenecks accounting for 56.9% of cells. Road saturation remained below 1.0.
Dispatch Performance (Strong-Peak Baseline):
- Shanghai Pudong: The QUBO-inspired method reduced the final passenger queue from 3,445 to 2,477 passengers (a reduction of ~28%). Average passenger waiting decreased from 39.74 to 25.71 minutes. MPC also reduced queues but less effectively than the QUBO method in this scenario.
- Hangzhou Xiaoshan: The QUBO-inspired method reduced the final passenger queue from 2,053 to 1,482 passengers. Average passenger waiting dropped from 51.71 to 28.26 minutes. MPC reduced waiting times but failed to significantly reduce the final queue, consistent with a berth-constrained system where service capacity is the hard limit.
Robustness:
- Under demand scaling (multipliers up to 1.3) and random noise (coefficient of variation = 0.3), the QUBO-inspired method consistently maintained lower final passenger queues compared to both the "no dispatch" baseline and MPC across both airports.

Significance and Claims
The paper claims that reformulating airport landside congestion as an integrated computational problem allows for the identification of distinct dominant bottlenecks that vary by airport context. The study demonstrates that:

Context Matters: Shanghai Pudong requires strategies balancing service with downstream road protection, whereas Hangzhou Xiaoshan benefits more directly from improving pickup service capacity and passenger-vehicle matching.
Methodological Complementarity: QUBO-inspired simulated annealing is shown to be effective at exploring larger discrete action spaces to achieve significant queue reductions. Conversely, MPC offers a more transparent and interpretable rolling-horizon policy. The authors suggest these methods are complementary: MPC for structured control and QUBO-inspired search for combinatorial optimization.
Robustness: The proposed framework retains queue-reduction benefits even under significant perturbations in demand, supply, and service conditions.

The authors note limitations, including the reliance on inferred facility parameters and the use of BPR-based speed outputs as pressure indicators rather than measured field speeds. They suggest future work should integrate real-time trajectory data and stochastic passenger choice models for further validation.

Technical Summary: A QUBO-Inspired Computational Framework for Airport Landside Bottleneck Diagnosis and Dynamic Dispatch Optimization

More like this