Urban traffic signals serve as the critical "valves" of a city's transportation network. Historically, signal timing has been optimized for a single objective: minimizing vehicle delay or maximizing throughput. However, as environmental concerns become central to urban planning, there is an urgent need to integrate air quality into traffic management strategies. A recent study published in Transportation Research Part C proposes a sophisticated mathematical framework to achieve this, balancing traffic efficiency with vehicle emission reduction.
The Challenge of Modeling Emissions
Optimizing for emissions is significantly more complex than optimizing for delay. Emissions are highly sensitive to microscopic vehicle behaviors—specifically, the cycles of braking, idling, and acceleration characteristic of "stop-and-go" traffic. Modeling these dynamics across a large urban network typically results in non-convex, non-linear mathematical problems. Such problems are computationally expensive and often impossible to solve in real-time using standard hardware.
The researchers sought a "middle ground": a model detailed enough to capture the essential physics of emissions but simple enough to be solved efficiently through linear programming.
Methodological Framework
The proposed approach rests on three innovative pillars that bridge the gap between traffic engineering and mathematical optimization.
1. Link-Based Kinematic Wave Model (LKWM)
To maintain computational efficiency, the authors utilize a link-based version of the classic Lighthill-Whitham-Richards (LWR) model. Unlike cell-based models that divide roads into numerous small segments, the LKWM captures critical phenomena like "spillback"—where a queue at one intersection blocks the previous one—using a significantly smaller number of variables.
As shown in the intersection diagram above, the model tracks flows ($I_i$) and turning proportions ($\alpha_{i,j}$) to maintain a continuous representation of traffic dynamics across the network.
2. The Occupancy-Emission Correlation
A key breakthrough in this research is the discovery of a statistically valid macroscopic relationship. Through 42,000 simulations using modal emission models, the authors found that the Aggregate Emission Rate (AER) of a road link is highly correlated with its Vehicle Occupancy (LO). This means that by knowing the number of vehicles on a road segment, one can accurately estimate the total pollution produced without needing to track every individual car's trajectory.
3. Robust Optimization (RO)
Because the relationship between occupancy and emissions contains inherent "noise" or uncertainty, the authors employed Robust Optimization. This technique accounts for errors in emission estimates by creating an "uncertainty set." Using duality theory, they reformulated these uncertain, non-linear constraints into a Mixed Integer Linear Program (MILP).
The table above outlines the primary variables used in this MILP formulation, including entry/exit flows ($\bar{q}, \hat{q}$), supply/demand ($S, D$), and the binary signal control variable ($u$). This linear structure allows commercial solvers like CPLEX to find optimal solutions rapidly.
Key Insights and Results
The study validated the MILP approach using a hypothetical 10-link network with four intersections. Several critical insights emerged:
- Efficiency vs. Environment: In light traffic, the model reduced total emissions by nearly 2% with a negligible 0.02% increase in traffic delay. This suggests that "green" timing plans can be implemented with almost no perceived cost to drivers.
- Targeted Mitigation: In heavy traffic scenarios, the model demonstrated the ability to reduce emissions on specific, highly-polluted links by approximately 15%.
- Environmental Equity: The framework allows planners to set "pollution ceilings" for individual links. This prevents the optimization process from inadvertently concentrating all network pollution into a single neighborhood, ensuring a fairer distribution of air quality impacts.
Conclusion and Future Directions
The primary contribution of this work is making emission-aware traffic control computationally tractable. By transforming non-linear emission curves into linear "robust counterparts," the authors have paved the way for more sustainable urban traffic management.
Future work will focus on expanding the model to include different pollutants (such as NOx or CO2) and accounting for dynamic routing, where drivers might change their paths in response to new signal timings. Ultimately, this linearization approach serves as a foundational step toward real-time, adaptive "green" traffic control systems.
Summary updated