Big Loads, Big Challenges, Big Data: Exploring Oversize/Overweight Flows and Crashes
Blake Moris, Burns and McDonnellShow Abstract
Jia Liang, Kansas State University
Eric Fitzsimmons, Kansas State University
Gregory Newmark, Kansas State University
Oversize and overweight (OSOW) truck movements represent a special challenge for road agencies. These flows support the local economy but also increase damage to infrastructure and raise safety concerns. Haulers of OSOW loads must request permission to use public roadways and many states have implemented automated systems to process these requests and provide route guidance. The electronic records associated with these systems provide an exciting new data source to understand OSOW traffic. This research demonstrates the use of these permit records to understand the industry mix requiring OSOW movements and to innovatively visualize the intensity of those movements using heatmaps. This work finds that much of the variation in the magnitude and spatial distribution of OSOW trips comes from the energy and agriculture sector. This research further explores OSOW crash data to identify hot spots and reveals two locations reporting repeated challenges with OSOW loads. Finally, a logistic regression model of crash severity identifies the factors associated with OSOW crash severity. These factors include overweight rather than oversized loads, traveling on Friday and Saturday, and a series of specific accident events, most notably overturning. Aside from the specific findings, this research offers a straightforward methodology to use big data to handle the big challenges of big trucks. The use of geocoded permit and crash data enable an excellent visual method to quickly identify OSOW demands on roadways.
A Data-Driven Opportunity Identification Engine for Collaborative Freight Logistics Based on a Trailer Capacity Graph
Jianlin Luan, Imperial College LondonShow Abstract View Presentation
Nicolo' Daina, Imperial College London
Kristian Reinau, Aalborg Universitet
Aruna Sivakumar, Imperial College London
John Polak, Imperial College London
A novel data-driven engine for identifying trailer capacity sharing opportunities during shipment planning stages is developed for providing a practical solution for collaborative freight logistics. The engine is based on a novel trailer capacity graph (TCG) that describes the real-time trailer capacity status of all collaborating partners. Compared with previous methods that replying on OD of shipments and trailer trips, the application of a TCG enables this engine to match shipments with both the OD and the route of trailer trips. Moreover, special treatments of a TCG, namely trailer route approximation and route shape simplification, dramatically reduce the computational cost, which makes this engine extendable for serving more than two collaborating partners in real-time. Experiments using this engine based on real-world data provided by two companies first illustrate that users of this engine should be aware that the configuration of the treatments of TCG is a trade-off between computational performance and effectiveness in identifying opportunities. Secondly, experiment results suggest that this two-partners collaboration implies an imbalance favourable towards the smaller operator. However, this unfavourable imbalance for the larger operator is likely to reduce as the number of operators joining the collaboration increases.
Representative Truck Activity Patterns from Anonymous Mobile Sensor Data
Taslima Akter, CPCS TranscomShow Abstract View Presentation
Sarah Hernandez, University of Arkansas, Fayetteville
Freight forecasting models are used to assess infrastructure needs, transportation policy implications, and environmental impacts of the multi-modal freight transportation system. Practical implementations of advanced freight forecasting tools like activity-based and truck touring models are hindered by the unavailability of data necessary to construct such models and the inability to generalize a set of representative patterns from data depicting the complex behaviors of the population. To fill this critical data gap, we developed a methodology to extract unique activity patterns from passively collected truck Global Positioning System (GPS) data. The use of passively collected data overcomes limitations associated with conducting periodic and expensive travel surveys. Salient features, i.e., stop durations, trip lengths, were extracted from around 338 million GPS records through stop identification and map-matching algorithms. Unsupervised machine learning (e.g., K-means clustering) was applied to discern common activity patterns among the records in a way that maintained anonymity of the trucks while providing high-resolution travel profiles. Six unique daily activity patterns were found representing a variety of truck operational types, i.e., long-haul movements with single stop, short-haul home-based movements with multiple stops, and medium-haul home-based movement with one/multiple stops. These activity patterns depict chains of activities and their trajectories over time and space which can serve to calibrate and validate advanced freight forecasting models. With more advanced forecasting models, we will be able to evaluate a wider spectrum of policy and infrastructure scenarios in order to ensure positive efficiency and environmental outcomes.
Characterizing the Movement of Freight Trucks Using Passive GPS Data: A Case Study in the Calgary Region of Canada
Ashok Kinjarapu, University of CalgaryShow Abstract View Presentation
Merkebe Demissie, University of Calgary
Lina Kattan, University of Calgary, Schulich
Robert Duckworth, Alberta Transportation
The movement of trucks represent a significant portion of travel. Surveys have traditionally been used to measure truck movement, but this costly and limited method of data collection typically involves in-person interviews and requires a high workload. This study explores different ways in which passive GPS data can be used to complement traditional data collection methods for obtaining detailed information about the travel behaviours of freight trucks. First, we develop an heuristic-based stop detection model to identify the trucks’ stop activities. Then, we develop another model that uses the identified stops to derive the trucks’ trips. A new methodology is proposed to identify the purpose of each stop, thereby improving the accuracy of the identified trips. Finally, we develop a destination choice model for modeling truck movements in the Calgary region of Alberta, Canada. This model applies a discrete choice modeling technique to distribute truck trips within the Calgary region. We test the utility function in the destination choice model for the inclusion of business establishment data, travel impedances, and other dummy variables that are likely to influence truck demand. The results show that a combination of trucks’ dwelling times and their entropy can be used to classify truck stops by purpose. This study also shows the potential of using passive GPS data to gain additional insights into the characterization of truck movements and for modeling truck trip distribution.
A GPS-Based Shipment Survey Assisted by Machine Learning Algorithms: Survey Design, Survey Platform, and Case Study
Peiyu Jing, Massachusetts Institute of Technology (MIT)Show Abstract View Presentation
Kyungsoo Jeong, National Renewable Energy Laboratory (NREL)
Linlin You, Singapore-MIT Alliance for Research and Technology
Jinping Guan, Massachusetts Institute of Technology (MIT)
Lynette Cheah, Singapore University of Technology and Design
Fang Zhao, Singapore-MIT Alliance for Research and Technology
Moshe Ben-Akiva, Massachusetts Institute of Technology (MIT)
Conventional shipment data collection methods are limited due to high cost, intense labor, and lack of details on shipment paths and stops. In this view, we develop an innovative shipment survey methodology using Future Mobility Sensing (FMS) - Freight to collect high-resolution shipment data at path-based origin-destination level and minimize respondent burden. FMS - Freight is a freight data collection, processing, and visualization platform which leverages sensing technologies and machine learning algorithms to interpret sensing data into travel diaries. We customized the existing FMS - Freight to accommodate the shipment survey. Specifically, we refined the stop detection, mode detection, and activity inference algorithms, revamped user interfaces, and developed a shipment data analysis and visualization tool. This web-based survey first collects the establishment’s business information, outgoing shipment information, historical shipment logs, and then requests tracking shipments with GPS devices, supplemented by a shipment registration survey and verification of shipment travel diaries. For proof-of-concept, we conducted a pilot shipment survey. 6 establishments participated in the pilot and we gathered verified GPS data from 57 shipment trips. The pilot demonstrated the feasibility of the survey design and instrument. This shipment survey has three aspects of significance: 1) It supplements the Commodity Flow Survey and enhances the capabilities to capture freight flows by combining user-verified geolocation data and detailed shipment information; 2) Collected shipment data can fill the significant data gap in the freight planning and modeling sector; 3) For individual establishments, FMS-Freight enables managing shipments in real-time and provides insights to assist decision-making.
Dynamic Pricing via a Web-Based Platform for Commercial Vehicles
Ioanna Pagoni, University of the AegeanShow Abstract View Presentation
Athena Tsirimpa, University of the Aegean
Amalia Polydoropoulou, University of the Aegean
Ioannis Tsouros, University of the Aegean
André Ramos, TIS-Consultores em Transportes
George Proios, University of the Aegean
Dynamic toll pricing can help road operators achieve better utilization of the infrastructure, as well as maximize revenue collection by applying variable tolls based on the congestion level of the highway. It also helps road users make optimal travel choices by adjusting their travel decisions. This paper presents a dynamic pricing web-based platform developed to incentivate the usage of toll roads by commercial vehicles and support the decision making process of the highway operators. The platform calculates toll prices for heavy vehicles applying algorithms based mainly on the Users’ Value-of-Time and the Highways’ Level-of-service offered. The results are provided in an online web-based platform to be used by freight forwarders to plan their trips. The proposed methodology was validated through field research involing the fleet of a major Portuguese freight forwarder in 2017. A number of key performance indicators are calculated to evaluate the effects of the variable tolls on: i) the traffic shift from the urban/national roads to toll highways; ii) the reduction of the freight operator’s cost elements; and iii) the increase of the toll operator’s revenues for the heavy vehicles’ passes. The results demonstrate that the proposed platform contributes to achieving a better utilization of the road network, as well as to maximizing the amount of revenue collected by the toll operator.
Research, Development, and Application of Alternative Methods to Estimating Freight Analysis Framework Out-of-Scope Commodity Flows
Christopher Lindsey, Cambridge SystematicsShow Abstract View Presentation
Krishnan Viswanathan, Cambridge Systematics
Birat Pandey, Federal Highway Administration (FHWA)
Out-of-scope (OOS) commodities are those goods that fall outside the sampling frame of the Commodity Flow Survey (CFS) conducted by the U.S. Census Bureau. They are shipments that originate with establishments that are not covered by the CFS such as farms, households, and foreign establishments. Goods contained in this category include farm-based shipments; fishery shipments; logs; municipal solid waste; construction and demolition debris; retail; services; household and business moves; crude petroleum; natural gas; and foreign trade. OOS commodities comprise a substantial share of the the Freight Analysis Framework version 4.2 (FAF4) and for many state departments of transportation and metropolitan planning organizations, these commodities represent goods and industries that are important to local economic development. Because of this, improvements to the estimation of OOS commodity flows yields benefits to freight planning at the national, state, and local levels. The objective of this research is to evaluate the existing FAF4 methods of integrating CFS OOS data, identify alternative methodological approaches and data, and develop and test alternative methodologies that offer short-term improvements for estimating OOS commodity flows. The review of literature found that other research efforts to model commodity flows offer opportunities to improve aspects of the OOS commodity estimation methods employed by the FAF4. The study then went on to develop an alternative methodology for modeling farm-based shipments of corn based on insights from those research efforts and demonstrate its effectiveness at the national level. Overall, the methodological approach produced strong results.