A Multi-objective Optimization Model at Signalized Junctions

A Multi-objective Optimization Model at Signalized Junctions


W Ma, Citilabs Inc, US


This research proposes a genetic algorithm model which generates a fully optimized junction design. It optimizes not only signal timing but also lane allocation, and meets any objectives specified by users such as construction costs or expected LOS.


In most urbanized settings worldwide, drivers have become accustomed to undesirable congestion and excessive delay. Efficiently operated traffic signals can reduce congestion and bring about significant payoffs in time and energy benefits. Nowadays, it is difficult to widen existing roads or build new roads in urban areas to improve the service of traffic networks. Better utilizing the existing traffic facilities is the only reasonable answer to most of the traffic congestion problems.

Signal optimization is considered as one of the most cost-effective methods to solve existing problems within signalized intersection networks and improve traffic signal operations. There have been considerable amount of relevant research studies reported. Some of the earliest research works focused on the optimization of the two basic signal timing parameters: cycle and splits. With improvements of computation technology, a variety of computer based signal timing software had been widely used. These tools provide optimization capabilities for estimation of more timing parameters, including phase durations and offsets.

However, most of the existing models have been concentrated on signal timing optimization and thus overlook the importance of other constraints. For example, the design of signal timing plan should be complementary to lane allocation pattern. An efficient signal timing plan should be based on a reasonable lane allocation pattern, and vice versa; otherwise, it would not really help to improve the operation service. Recently, the author proposed an optimization model for the integration design of signal timing and lane pattern at signalized junctions. The decision variables include not only signal parameters but also lane allocation variables including number of exclusive lanes and shared properties for each movement.

In this paper, the author extends the previous research and proposes a multi-objective optimization model at signalized junctions. The extended model provides more flexibility to users to meet their different needs. The objectives could be for operation purpose (to minimize delay or to maximize through put), for planning purpose (to minimize construction budget or to meet expected level of service), or for a combination of both. A fully optimized junction design, including control type, cycle length, phase durations, phase sequence, permitted movements, lane allocations, shared pattern, construction budget, and etc, is generated according to the data and objectives provided by users. The problem is formulated and solved by a genetic algorithm-based model. The model is implemented and validated in the Sugar ArcGIS extensions that released by Citilabs.


Association for European Transport