Anticipatory Dynamic Traffic Sensor Location Problems with Connected Vehicle Technologies

Abstract

Technological advancements in wireless systems enable the advancement of traffic operations. When onboard equipment in connected vehicles transmits data to roadside sensors integrated with a traffic signal controller, green phases are redistributed to directly and indirectly reduce network delays. Despite the potential benefits of sensor technologies, the challenges associated with identifying optimal sensor locations for multiple time stages throughout a day with uncertain demand patterns has received little attention. In this paper, we focus on proactively reducing the network delay by controlling traffic signals through an optimized sensor deployment. The framework is based on portable sensors that may be repositioned within day and day to day to new locations such that delay savings over multiple time stages will be maximized. To tackle this multiperiod stochastic problem, dynamic models are proposed, considering the future sensor locations given budget constraints on the sensor costs and relocation costs, and the effect of control is tested on various demand profiles and penetration rates of an urban transportation network. A subproblem decomposed by Lagrangian relaxation enhanced with valid cuts has a better bound, and a variable neighborhood search algorithm quickly finds solutions. Two dynamic models that constrain flexible and restricted relocation, respectively, present higher savings compared to the stationary model without sensor relocation. The flexible relocation model guarantees higher savings than the restricted model by achieving the same maximum savings with fewer sensors. The gap between two dynamic models decreases when more sensors are available.