A Monte Carlo Simulation Procedure to Search for the Most-likely Optimal Offsets on Arterials Using Cycle-by-Cycle Green Usage Reports

More and more towns in the states have installed central traffic managing software (ATMS) to manage their signal systems. ATMS’s not only enable traffic engineers to remotely watch real-time traffic or access local controllers but also enable them to collect more data than ever. How to better utilize these data to improve the performance of traffic signals has been a topic receiving wide interest in the signal community. The data collected in ATMS’s can be classified into two types, traffic counts/occupancies via detectors and green usage information via local actuated controllers. Unlike the previous researches and practices which mostly focused on how to better use the detector data, this paper is intended to explore how to use the green usage information in an ATMS to design optimal offsets of coordination. Although many factors could affect the effectiveness of offsets, the optimization of offsets usually starts with computing main-line link travel times. In actuated coordination, the main-line greens may start earlier than what are programmed because uncoordinated phases could gap out and return unused green back to the main-line. This phenomenon is referred to as “early-return-to-green”. When it occurs, it makes the programmed coordination less effective. In this paper, the authors considered the main-line greens random variants ranging from the programmed maximum greens to the whole cycle length. In this random scheme, the optimal offsets are also random and should vary cycle by cycle. Given that it is still uncommon to adjust the offsets cycle by cycle in major signal controllers, the objective of offset optimization in this paper is to find optimal values that can maximize bandwidth and therefore minimize delay with the largest likelihood. The authors first defined this new concept as Most-likely Optimal Offsets, then used cycle-by-cycle green usage reports and a Monte Carlo simulation model to determine the most-likely optimal offsets. The cycle-by-cycle green usage reports is a typical function of major ATMS systems to provide the distributions of random main-line greens. It serves as the basis to infer the optimal offset distributions and thus allow for identifying the most likely optimal offsets. Three intersections on Payne Road in Scarborough, Maine were selected to test a set of new offsets inferred with this method. A before-and-after comparison in simulation revealed that the new offsets could significantly reduce the travel times on arterials with 95% confidence level compared to the offsets optimized with SYNCHRO 7 when the early-return-to-green frequently occurs. The implementation in the field also supports the speculations from simulation.