This session will highlight innovative data sources, methods, and applications for freight planning and operations.
Specified Highway Freight Estimation Based on Multi-Source Data Fusion
Peng Zhang, Ministry of Transport of the People’s Republic of ChinaShow Abstract
Enze Huo (email@example.com), Beihang University
Fangshu Lei, Beijing Transport Institute
Yingping Wang, Ministry of Transport of the People’s Republic of China
Mingchen Gu, Ministere des Transports du Quebec
Xiaolei Ma, Beihang University
The current method for highway freight statistics is introduced, and several critical problems are discussed. To improve the performance of highway freight statistics and meet the increasing demands for specified statistics, using multi-source data from the latest highway informatization projects of China, an automated method to estimate the highway freight with multi-dimensional statistical definition is developed. The specified freight estimation is realized in terms of the transport categories, road level categories, truck categories and highway routes. In addition, compared with that only the commercial trucks are included in current method, all the trucks operating on the road are included. The freight volume is counted in terms of the area in which the freight is completed instead of the registration area of the truck. Finally, a test approach for the proposed method is presented. The result shows that the MAPEs of tons and ton-kilometers are 11.5% and 2.7%, respectively. The proposed method is more reliable and more efficient than existing methods. The method is expected to be a supplement to current highway statistical policy of China as well as a reference for the development of automated highway transport statistics for other countries
Truck Parking Occupancy Prediction: End-to-End Machine Learning Framework
Sebastian Gutmann (firstname.lastname@example.org), Technical University of MunichShow Abstract
Christoph Maget, Bavarian Road Administration
Klaus Bogenberger, Technische Universitat Munchen
For haul truck drivers it is becoming increasingly difficult to find appropriate parking at the end of a shift. Proper, legal and safe overnight parking spots are crucial for truck drivers in order to comply with Hours of Service regulation, reduce fatigue and improve road safety. The lack of parking spaces affects the backbone of the economy, as 70% of all US domestic freight shipments (in terms of value) are transported by trucks. Many research projects exist which provide real time truck parking occupancy information at a given stop, however truck drivers ultimately need to know whether parking spots will be available at a downstream stop at their expected arrival time. We propose an end-to-end machine-learning-based prediction framework that is capable of accurately predicting occupancy. The framework is able to deal with real world data with all its imponderabilities. Our results show that, even during peak times, prediction of truck parking occupancy can be achieved with small mean absolute errors: 1.71 trucks for 30min, 2.60 trucks for 60min, 3.30 trucks for 90min, and 3.98 trucks for 120min forecasts. We can also demonstrate that our machine-learning-based approach outperforms classical prediction techniques. In contrast to other methods, our proposed solution only needs easily obtainable data. Ultimately, any truck occupancy detection system could also provide forecasts by implementing our end-to-end prediction framework.
Suitability of Fusing Vehicle Probe Data and Vessel Data to Contextualize the Multimodal Interaction Impacts on Corridor Mobility – a New Orleans Case Study
Bethany Stich, University of New OrleansShow Abstract
Kirk Zeringue, Louisiana Department of Transportation and Development
Guang Tian, University of New Orleans
Using New Orleans as a case study, this research explores the conflation of vehicle probe data with various vessel datasets to characterize the interactions between container vessels and motor vehicles as it relates to interstate congestion in a port city. The case study investigates the impact of container vessel presence/size and fluctuations in container volumes on roadway congestion. The exploration relies on comparing different conditions using cumulative distribution functions and Innovative Trend Analysis. The results show that fusing vehicle and vessel data is achievable and appropriate, but temporal and data completeness issues exist. The results also show that by joining these modally disparate datasets together and analyzing them as one, additional context is added to discussions related to transportation operations and investment decision-making through either the confirmation or disproval of perceptions or expectations related to container truck traffic on interstates.
Study on the Establishment of Integrated Database on Livestock-Related Vehicles to Prevent the Spread of Livestock Infectious Disease
Heehyeon Jeong, University of SeoulShow Abstract
Jungyeol Hong (email@example.com), University of Seoul
Dongjoo Park, University of Seoul
The outbreak of African swine fever virus has raised global concerns regarding epidemic livestock diseases. Therefore, various studies have attempted to prevent and monitor epidemic livestock diseases. Most of them have emphasized that integrated studies between public health and transportation engineering are essential. However, it has been difficult to obtain big data regarding the mobility of livestock-related vehicles. Thus, it has been challenging to conduct integrated research; additionally, there are no experts simultaneously familiar with both fields. In this study, we matched digital tachograph (DTG) data (including commercial truck coordinate information but no load items) to visit history data by facility based on actual livestock-related vehicle access information, but without considering the specific trajectories of the vehicles. In addition, a social network graph was constructed by analyzing the connections between livestock-related facilities; 20 facilities considered vulnerable to the spread of livestock epidemics were identified through a network centrality analysis. A total of 36,794 trajectories for commercial trucks visiting the facilities were extracted from the DTG data. The integrated database can be used as a significant resource for preventing the spread of livestock epidemics, by pre-monitoring the movements of livestock transport vehicles. In future studies, epidemiological research on infectious diseases and livestock species will be conducted through the derived matching database, and indicators of the spread of infectious diseases can be suggested based on roadway links, for the management of livestock epidemics.
STATEWIDE ANALYSIS OF TRUCK GPS DATA TO UNDERSTAND TRUCK 1 PARKING UTILIZATION
Makarand Gawade, HDRShow Abstract
Arjun Chauhan, HDR
One of the challenges in understanding truck parking shortages is the lack of appropriate data and procedures for analyses. This study applied advanced geospatial and tabular analytics to develop truck parking utilization measures in the state of Florida using raw truck GPS data. The methodology included identifying truck parking locations in Florida, determining all stopped trucks at identified truck parking locations and along roadways and expanding the sample stopped trucks to determine the total number of trucks stopped at all parking locations. The final step involved computing utilization measures of all truck parking locations, identifying the locations with high unauthorized truck parking and discerning the major areas of concerns. Approximately, 300 truck parking locations were identified in the state. The outcomes indicated that utilization and average truck dwell time at public parking locations is lower than the private parking locations. The highest parking utilization is observed from 7 pm to 9 am. Truck parking near urban areas and along major freight corridors matched the freight activity occurring in those areas. Areas where truck parking demand exceeds supply often have unauthorized truck parking, such as trucks parked on the side of highways, on or off-ramps, vacant lots, on local roadways and near freight origins/destinations. Priority areas of concern with high concentrations of unauthorized truck parking areas and over-utilized truck parking locations were determined which can be used to focus the analysis and enable the identification of needs, opportunities, and recommendations.
Using machine learning to predict freight vehicles demand for loading zones in urban environments
Andres Regal Ludowieg, Universidad del PacíficoShow Abstract
Ivan Sanchez-Diaz, Chalmers tekniska hogskola
Lokesh Kalahasthi, Chalmers tekniska hogskola
This paper studies demand for public loading zones in urban environments and seeks to develop a machine learning algorithm to predict their demand. Understanding and predicting demand of loading zones can (i) support better management of loading zones for the public sector and (ii) provide better pre-advice so that transport operators plan their route in an optimal way. The methods used are linear regression analysis and neural networks. The authors used 6 months of parking data from the city of Vic in Spain to calibrate and test the models. The results show that linear regression gives a fair prediction of availability of loading zones, and the probabilistic neural network is the one that provides better results among neural networks.
A Deep Ensemble Neural Network Approach for FHWA Axle-based Vehicle Classification using Advanced Single Inductive Loops
Yiqiao Li (firstname.lastname@example.org), University of California, IrvineShow Abstract
Andre Tok, University of California, Irvine
Stephen Ritchie, University of California, Irvine
The Federal Highway Administration (FHWA) vehicle classification scheme is designed to serve various transportation needs such as pavement design, on-road emission estimation, and transportation planning. Many transportation agencies rely on Weigh-In-Motion and Automatic Vehicle Classification sites to collect these essential vehicle classification count data. However, the spatial coverage of these detection sites across the highway network are limited due to high installation and maintenance costs. One cost-effective approach investigated by researchers has been the use of single inductive loop sensors as an alternative to obtain FHWA vehicle classification data. However, most datasets used to develop such models are skewed since many classes belonging to larger truck configurations are rarely observed in the roadway network. This increases the challenge to accurately classify under-represented classes, even though many of these minority classes may pose disproportionately adverse impacts on pavement infrastructure and the environment. As a consequence, previous models have been unable to adequately classify under-represented classes, and the overall performance of the models are often masked by excellent classification accuracy of majority classes, such as passenger vehicles and five-axle tractor trailers. To resolve the challenge of imbalanced datasets in the FHWA vehicle classification problem, this paper describes a study that developed a bootstrap aggregating (bagging) deep neural network (DNN) model on a truck-focused dataset using single inductive loop signatures. The proposed method significantly improved the model performance on several truck classes, especially minority classes such as Classes 7 and 11 which were overlooked in previous research studies.
Truck Parking Pattern Aggregation and Real-time Availability Prediction by Multi-task Learning
Hao Yang, University of WashingtonShow Abstract
Chenxi Liu, University of Washington
Yifan Zhuang, University of Washington
Wei Sun, University of Washington
Yinhai Wang (email@example.com), University of Washington
The truck industry is an indispensable part of freight transportation in the United States. Nevertheless, the supply of truck parking infrastructure is far from enough to meet the drivers' demands. Lack of parking spaces and real-time parking availability information greatly increase the uncertainty of trips, sometimes even resulting in illegal and potentially dangerous parking or overtime driving. Although extensive research and technologies have been conducted to improve the utilization of parking facilities in urban areas, however, services are still limited and out-of-date for trucks. This paper elaborates on a pilot research of improving truck parking facilities funded by the Washington State Department of Transportation (WSDOT), including slot-based data collection with the Truck Parking Information and Management System (TPIMS), the parking pattern aggregation and similarity analysis, and a multi-task learning neural network for time-variant occupancy prediction. Our proposed framework achieved 5.78%, 5.14%, 4.92%, and 4.24% mean average percentage error (MAPE) for 16, 8, 4, and 2 minutes ahead occupancy prediction, respectively. The promising results of this study would benefit truck drivers on trip scheduling and the government agencies on optimal parking facility operations.
Modelling Long-Haul Truck Route Choice in Ontario
Syed Ubaid Ali, York UniversityShow Abstract
Kevin Gingerich (firstname.lastname@example.org), York University
A route choice model is developed to explain and predict long-haul truck vehicle movements in Ontario. To accomplish this task, an algorithm is first devised to process approximately 58,000 observed trips from GPS data in ArcGIS and establish variable choice sets based on an optimal commonality factor that measures route overlap. Novel implications of the commonality factor add to the existing literature. Route characteristics are next used to estimate a C-logit model. Results indicate that truck drivers are more likely to select routes exhibiting lower minimum travel times, more freeway and Highway 401 usage, more diesel stations, and fewer intersections. The travel time is the most dominant variable based on measurements of elasticity. Two scenarios are tested using the final model to determine routing changes due to increased travel time on Highway 401 and other freeways. Further detailed scenarios can be used to predict long-haul trucking patterns for future transportation planning purposes.
Incremental System for Identifying Prime Routes in Logistics Transportation Services
Ajay Hayagreeve, FourKites India Private LimitedShow Abstract
Jayakarthick Sathyanarayanan, FourKites India Private Limited
Vinodh Krishnaraju, FourKites India Private Limited
Logistics serves as the backbone of supply chain domain and truck transportation services are a crucial component. Transportation Services planning is affected a lot by external factors such as weather, operational reschedules, unexpected hazards in the route and so on. Route preferred during transportation is a critical factor playing a major role in deciding travel time, cost savings and downstream planning. Public map providers route doesn’t suit Logistics Transportation Services in most cases due to truck/trailer size and time constraints. The regular shortest or fastest route is not suitable for large-size vehicles. Also certain industry standards require specific routes to be avoided. In this research we have built a system to identify prime routes from historic data in the North American region. The system comprises of a spatial algorithm that mines the routing patterns from the noisy historical trip data. The target user can query our system for identifying the prime routes in North American Logistics network. We have validated the results on unseen trips from the future time frame and profiled against the map provider routes. The lane is defined at a resolution of city level while building the algorithm. Our System serves multiple stakeholders in Logistics and enables better planning and data-driven decision making.
DISCLAIMER: All information shared in the TRB Annual Meeting Online Program is subject to change without notice. Changes, if necessary, will be updated in the Online Program and this page is the final authority on schedule information.