Multi-Source Data Fusion Oriented Freeway Sensor Placement Method Research

Traffic detector placement determines the travel time estimation and prediction precision and is also effected by travel time estimation methodology for research and application. As a critical parameter in freeway traffic management, travel time can provide data support and decision reference for the implementation of traffic guidance and publishing of travel information.According to existed research literature and summary, this research studied freeway detector placement methods for data fusion. This research emphasized cellular handover tech and regarded it as an aspect of detector placing consideration and an innovative breakthrough. With the support of traffic state data provided by cellular handover technology, the research estimated travel time based on BP neural network data fusion method and studied point detector placement with constraints and inference of travel time estimation errors.This thesis took cellular positioning traffic information extraction as an entry and innovation point, introduced the traffic application of cellular positioning technologies and analyzed the emphasis of their system framework and technology flow. Based on technical characteristics, the research compared traditional fixed detector and GPS probe technology with cellular handover to underline its advantages and application future. Meanwhile, this research also referred usual method of simulating GPS probe and proposed a new cellular handover microsimulation method in VISSIM environment and verified the feasibility and effectiveness through a realistic traffic simulation and analysis. The research results stated that with enough sample percentage (here is more than 3%), travel time estimation accuracy based on cellular handover data is more precise and not be influenced by traffic conditions.Then through a traffic simulation and its export data of point data and handover data, this research used a direct method and fusion method to estimate the real freeway section travel time. The research established a data fusion model using BP neural network for travel time estimation in different and changeable traffic state situations to satisfy complex demands and proposed a travel time estimation evaluation index system. According to research demands and data fusion steps, this research separately analyzed and estimated the whole freeway section and each handover link with fusion method, compared with single data source and direct method and deeply analyzed BP fusion method and fusion results. This research claimed that travel time estimation based on BP neural network fusion method is more accurate and stable than that based on single data source, with more advantages in bad traffic condition.Taking the principles and effects of detector placement into consideration, the research established a uniform detector layout strategy and a blind area compacting strategy, with the support of cellular handover as a traffic information collection environment. To determine the detector placement method and strategy, the research took a real freeway section into micro simulation in VISSIM and calculated point traffic data and cellular handover data to verify and derive feasible placement strategy and method, based on the proposed evaluation index and results deep analysis. The research proved that when the spacing of detectors is between 1000 and 1500 meters, travel time estimation is stable and accurate relatively and strategies in this range are reasonable. Meanwhile, detector placement should take in consideration of long handover links and ramp existence to realize estimation benefits maximized with low investment cost.