Exploring the Use of Lidar Data From Autonomous Cars for Estimating Traffic Flow Parameters and Vehicle Trajectories

2017PDFTech ReportDTRT13-G-UTC33Traffic flowLaser radarAlgorithmsVehicle trajectoriesAutonomous intelligent cruise controlIntelligent vehiclesLIDAR3D point cloudKalman Filtermicroscopic traffic flowmacroscopic traffic flowVirginiaMid-Atlantic Transportation Sustainability University Transportation CenterCetin, MecitSazara, CemNezafat, Reza VataniOld Dominion University. Transportation Research InstituteMid-Atlantic Transportation Sustainability University Transportation CenterUnited States. Dept. of Transportation. Office of the Assistant Secretary for Research and TechnologyNTL-OPERATIONS AND TRAFFIC CONTROLS-Traffic FlowUS Transportation CollectionLIDAR has become one of the major enabling technologies for autonomous vehicles. LIDAR sensors generate 3D point cloud data around the instrumented vehicle. These data enable detailed localization of both fixed and moving objects in the surrounding environment. In this project, various aspects of LIDAR data processing are studied with an emphasis on exploiting such data for estimating traffic flow parameters. To achieve that, LIDAR data are collected in the field by an instrumented vehicle under different traffic conditions. Data processing and machine learning algorithms are developed to detect and classify vehicles based on the point cloud data. These algorithms help track ambient vehicles and extract their trajectories. From the extracted trajectories, both microscopic and macroscopic traffic flow parameters can then be estimated. In the case when some vehicle trajectories are partially observed or missing for a short duration, e.g., due to occlusion, this report shows how such trajectories could be completed by employing car following models. To demonstrate the use of trajectory data for estimating traffic flow parameters, a richer dataset with all vehicle trajectories, i.e., the NGSIM data, are utilized. Models are developed to predict traffic density, or the number of unobserved vehicles between two known trajectories, under congested conditions from a sample of trajectories. Under the dense traffic conditions analyzed, the results show that the number of unobserved vehicles between two probes can be predicted with an accuracy of \ub11 vehicle almost always