Sequential Modeling Framework for Optimal Sensor Placement for Multiple Intelligent Transportation System Applications

Abstract

Traffic sensors have been deployed for decades to freeways to meet the requirements of various traffic/transportation applications, most noticeably traffic control and traveler information applications. A unique feature of traffic sensor deployment is that it is a continuous and evolving process. That is, with new applications that emerge, additional sensors are usually required to be deployed to meet new requirements. This process is also sequential in nature and the new deployment has to consider existing sensors. In this paper, we propose a modeling framework to capture this sequential decision-making process for traffic sensor deployment. The framework is based on our recent findings that (1) optimal sensor deployment for a single application can be determined by a staged process or, more formally, a dynamic programming (DP) method and (2) new sensor locations for new applications can be optimally solved by the DP method via considering existing sensors. We illustrate the framework using two applications: ramp metering control and travel time estimation. It is found that the proposed scheme can appropriately capture the decision-making process of traffic sensor deployment and can generate optimal sensor placement at any stage by considering sensors that have already been deployed. The model is tested using global positioning system enabled cell phone data and traffic simulation on a real-world freeway route in the San Francisco Bay Area.