This session will highlight recent research in data science and data fusion applications to capture freight data elements that have consistently been challenging to measure. Presentations will address new or improved methods to estimate truck empty and loaded trips, truck stop by purpose, and disaggregate multi-modal commodity flows.
Multi-Commodity Port Throughput from Truck GPS and Lock Performance Data Fusion
Magdalena Asborno, U.S. Army Corps of Engineers (USACE)Show Abstract View Presentation
Sarah Hernandez, University of Arkansas, Fayetteville
Taslima Akter, CPCS Transcom
Inland waterway ports serve as critical connections among freight modes along America’s Marine Highway Network, making them key elements of an efficient multimodal freight transportation system. Understanding the capacity and throughput of inland waterway ports by commodity type can support more effective long-range freight planning and travel demand model development. More specifically, such data can be used to estimate multimodal, commodity-based freight fluidity performance measures and to support location selection for waterborne freight transload facilities. State-of-the-practice means of gathering commodity flows such as shipper/carrier surveys and vessel and vehicle movement data are limited in their spatial disaggregation, temporal continuity, and multimodal integration. This work addresses these limitations by developing a Multi-Commodity Assignment Model that fuses mode-specific datasets, i.e. truck Global Positioning System (GPS) and waterborne Lock Performance Monitoring System (LPMS) data. Publicly available LPMS data provides the required commodity dimension, while anonymous truck GPS data allows for spatial disaggregation. A goal programming approach was adopted for multi-objective optimization in which minimal deviation between known and estimated truck flows at each port was sought. The methodology was applied to the Arkansas portion of the McClellan Kerr-Arkansas River Navigation System (MKARNS) which consists of 308 miles of river, divided by locks into 13 segments and served by approximately 43 freight ports. Results were assessed by comparing observed and predicted truck flows at each port. Overall, 84% of ports along the study area had less than 20% difference between observed and predicted truck flows.
Empty Platform Semi-Trailer Classification Using Side-Fire LiDAR Data for Supporting Freight Analysis and Planning
Olcay Sahin, Argonne National LaboratoryShow Abstract View Presentation
Mecit Cetin, Old Dominion University
Ilyas Ustun, DePaul University
Empty truck trips constitute an important aspect of commodity-based freight planning and modelling. But this information is generally not available to State DOTs or Metropolitan Planning Organizations (MPOs) since detecting empty trips is a challenge with traditional vehicle sensors. In this study, we propose a method for detecting empty and loaded platform semi-trailers using data from a multi-array LIDAR sensor. From the LIDAR cloud points, 3D profiles of trucks can be generated, and these profiles allow extracting useful information (e.g. body type, empty and loaded platforms). Since only platform semi-trailers’ load is observable from their 3D profiles, we only consider open platform trailers which constitute 20% of the truck trailer population in the USA. This paper shows how point-cloud data from a 16-beam LIDAR sensor are processed to extract useful information and features to distinguish between empty and loaded platform semi-trailers versus all other major truck body types (e.g. dry van, container, tank, automobile transport, etc.). Several machine learning (ML) models, in particular, K-Nearest Neighbors (KNN), Multilayer Perceptron (MLP), Adaptive Boosting (AdaBoost.M2), and Support Vector Machines (SVM) are implemented on the field data collected on a freeway segment that includes over nine-thousand trucks. The results show that all major semi-trailers and empty platform semi-trailers can be distinguished with very high level of accuracies of 99% and 97% respectively.
Imputation of Truck Rest Stop Events from GPS Data with a Continuous Hidden Markov Model
Mehdi Taghavi, University of QueenslandShow Abstract
Elnaz Irannezhad, Australian Road Research Board (ARRB)
Carlo Prato, University of Queensland
The increasing wealth of GPS truck data has broadened the opportunities for understanding freight logistics activities and enhancing research capabilities to real-world case-studies. A very important piece of information from the planning and regulatory perspective concerns the occurrence and location of rest stops. In this study, we propose a data-driven unsupervised machine learning method to impute truck stop events by using a Continuous Hidden Markov Model (CHMM). Specifically, we estimate the joint probability distribution of a mixture of multivariate Gaussian densities, whose parameters depend on the latent states of a Markov chain. Each density represents a cluster of stops that are identified not only from their spatial proximity, but also from their temporal proximity as the clustering of the rest stops depends on latent states that are conditional on expected times retrieved from the observed data. In this study, we applied the proposed method to a database containing more than 71 million GPS records of Australian trucks, and we particularly aimed to identify rest stops based on a list of features related to the locations and the load of the trucks. The results showed that the CHMM could account for the location proximity for different activities of truck drivers, and they were validated against complementary data on truck loads and land use by using a stratified sampling technique. Validation results indicated that 94.1% of the rest stops were correctly identified, and highlighted the advantage of the proposed approach without any requirement of labelled data, driver logbook or complimentary survey.
Identifying Freight Delivery Stops from GPS Data
Jose Holguin-Veras, Rensselaer Polytechnic Institute (RPI)Show Abstract View Presentation
Trilce Encarnacion, University of Missouri, St. Louis
Sofia Perez-Guzman, Rensselaer Polytechnic Institute (RPI)
Xia Yang, State University of New York (SUNY)
Truck stop identification is crucial to characterize freight deliveries and asses the performance of freight transportation systems. Identifying the stops from raw GPS data is very challenging, particularly in urban freight systems where congested traffic is common. This paper presents a deterministic classification method to identify freight delivery stops from raw GPS data. The method was implemented to identify stops on three disctinct case studies that present a range of traffic conditions in the cities of Barranquilla, Colombia; Dhaka, Bangladesh; and NYC, USA. The results show that the algorithm achieves an average accuracy above 98.6% when identifying truck delivery stops in all cases.