كنفرانس بين المللی مهندسی صنايع و سيستم ها

صفحه اصلی / دوازدهمین كنفرانس بين المللی مهندسی صنايع و سيستم ها

Demand-Weighted Signal Timing for Over-Saturated Intersections via Integrated Discrete-Event Simulation and Convex Optimization: A Case Study of Ahmadabad Intersection, Mashhad

نویسندگان :

Ali Nokar¹

1- دانشگاه فردوسی مشهد

کلمات کلیدی :

Discrete-Event Simulation (DES)،Cellular Automaton،Latin Hypercube Sampling (LHS)،Traffic Signal Optimization،Convex Quadratic Programming (QP)،Urban Intersection Control

چکیده :

Traffic congestion at major urban signalized intersections often results in queue spillback, operational imbalance, and excessive delays under static signal timing plans. This paper presents an integrated simulation-optimization framework for improving the performance of the Ahmadabad Intersection in Mashhad, Iran, a major four-way junction operating under a fixed 5-stage signal control scheme with a 150-s cycle length. A continuous-time Discrete-Event Simulation (DES) incorporating a Cellular Automaton (CA) movement model was developed in Python to represent microscopic vehicle movements along straight and left-turn trajectories. Baseline experiments under stochastic Poisson arrivals and Bernoulli movement probabilities revealed severe performance heterogeneity, including unbounded queue growth on the Ahmadabad West and East approaches. To investigate the effects of traffic demand and turning movements on system performance, a multi-level sensitivity analysis combining Full Factorial Design, Latin Hypercube Sampling (LHS), and response surface visualization techniques was conducted, identifying strong nonlinear interactions among signal phases and the movement streams with the greatest influence on network-wide delay. Based on the simulation outputs, a Response Surface Metamodel (RSM) with quadratic delay functions exhibiting high predictive accuracy (R²> 0.95) was constructed and embedded within a Convex Quadratic Programming (QP) model formulated in Pyomo. The optimization minimized demand-weighted system delay subject to practical green-time and cycle-length constraints. The proposed framework rapidly converged to an optimal solution, reducing total system delay to 1335.93 veh·s while reallocating green times toward high-impact traffic streams, eliminating structural bottlenecks, and restoring operational balance without requiring physical infrastructure modifications. The results demonstrate that the proposed simulation-optimization framework provides an effective and computationally efficient decision-support tool for data-driven traffic signal management and intersection performance enhancement.