speaker-abbink (2)

Erwin J.W. Abbink

Managing Consultant Innovation
Netherlands Railways

Crew Management At Netherlands Railways

Recent industry and scientific developments provide railways with a wide range of tools. We focus on Crew Management, including crew scheduling, rostering and real-time dispatching. Based on practical examples, we show that applying these models and algorithms in the railway domain is not trivial, but in the end was successful. This success led to: – A stable relationship between management and train crew. – Ability to perform scenario analyses, e.g., to study adjusting the labor rules.- Shortening the lead-time of the planning process from months to hours. – NS can adjust their service, when severe weather conditions are expected, shortly before the day of operation.- Better real-time rescheduling of crew, reducing the number of cancelled trains and train delays in case of unforeseen disruptions. We describe how we implemented decision support in the organization, how we managed the change process, lessons learned, and ongoing research in the domain of Security teams.


Erwin Abbink holds a MSc in Operations Research and a MSs in Information and Knowledge Technology. Currently, he is a managing consultant innovation at Netherlands Railways. His main focus is on leading the development of Decision Support Systems and Advanced Analytics. He has a broad experience in logistics and disruption management for railway systems. In 2004 he was a finalist in the Daniel H. Wagner Prize for Excellence in Operations Research Practice with the paper: Reinventing Crew Scheduling at Netherlands Railways. In 2008 he was winner of the Franz Edelman Excellence in Practice award with the paper: The New Dutch Timetable: The O.R. Revolution. Based on this work he received a PhD degree on Crew Management in 2014 at the Rotterdam School of Management.


Srinivas Bollapragada

Chief Scientist, Software Sciences & Analytics
General Electric Global Research Center

Rail Network Operations Optimization

Each of the class 1 railroads in North America owns between 20,000 and 60,000 miles of track and runs between 1000 and 5000 trains per day. Moving freight on such complex networks requires meticulous planning and execution. Multiple opportunities exist to realize large savings from improving railroad operations. For example, increasing the average speed of freight trains by one mile per hour saves a class 1 railroad around $200 million per year. Reducing the average dwell time of rail cars in yards by 1% saves the railroad industry $2.2bilion per year. Decreasing fuel consumed by locomotives by 1% saves 70 million gallons of fuel in North America. We will describe some of the products that we had developed that use data and domain knowledge to improve the operations of the rail network.


Dr. Srinivas Bollapragada is the Chief Engineer in the Software Sciences & Analytics organization at General Electric’s Global Research Center in Niskayuna, NY. His research interests are in business applications of analytics and optimization technologies. During the past 20 years at GE Research, he had led the development of numerous innovative, algorithms-based business systems that are in use at several organizations including NBC-Universal, Norfolk Southern, and GE’s Transportation, Capital Services, Energy, Aviation, and Healthcare businesses. Srinivas has published around 30 articles in journals such as Operations Research, Interfaces, European Journal of Operational Research, IIE Transactions etc. He has a Ph.D. in Operations Management from Carnegie Mellon University, and Master’s and Bachelor’s degrees in Electrical Engineering from IISc, Bangalore and IIT Madras respectively. He represents GE Research on the INFORMS Roundtable, had served as the Chair of the Edelman award committee (2008-2010), and the Area Editor of OR Practice (2008-2010). He was named an INFORMS Fellow in 2009 and has been serving as the Editor-in-Chief of the journal, Interfaces since 2011. Srinivas had received the Idelchik (2015), Dushman (2010), and Blodgett (2009) awards for his contributions at GE.


Juan Esteban Calle

Senior Operations Researcher
TDM Transportes

Transportation Network Design In TDM Transportes

TDM Transportes is a Colombian based transportation company. One of the main responsibilities of the Operations Research and Engineering department in the company is to evaluate and explore potential transportation network designs for TDM’s clients. Optimization has been a widespread methodology for continuing studies in transportation network design. However, there are opportunities in which simulation technology might be a massive help to optimization based techniques. There are companies in the market such as Llamasoft and Anylogic which are increasingly promoting simulation as a tool for evaluating supply chain networks performance. This presentation explores a case study in which a ‘simulation’ approach was used to explore possible scenarios for the finished product – sourcing transportation network of Industrias Haceb, the main Colombian Appliances Manufacturing and Distributor Company. The results are: the new network has accounted for saving of around 10% in an operation that costed approximately 2 million dollars a year.


Juan Calle is Senior Operations Research Analyst at TDM Transportes, a Colombian based Transportation Company. Before joining TDM, Juan worked at Decisionware, a Colombian based mathematical programming company. Juan has been teaching Operations Research related courses in leading Latin American Universities. Juan has a Bachelor’s degree in Industrial Engineering and has a Masters in Systems Engineering from the Universidad Nacional de Colombia. He is the cofounder and active member of UNGIDO, the Operation Research Group at Universidad Nacional de Colombia and ASOCIO, the Colombian Operations Research Society.


Ted Gifford

Distinguished Member of Technical Staff
Schneider National Incorporated

A Machine Learning Approach To Capacity Forecasting

One of the foremost challenges facing the freight transportation industry today is a shortage of qualified drivers. To address this challenge some large truckload carriers augment a company-employed driver fleet with Independent-Contractor (IC) drivers who are offered greater flexibility in determining both the freight they select and their work schedules. This paradigm presents a significant challenge to planners who need to forecast IC capacity (number of IC drivers who will work) over a 1-4 day rolling horizon. These forecasts are an integral component of down-stream decisions related to pricing, load acceptance, and scheduling of company drivers.We will present a predictive-analysis approach to this forecasting problem that applies an ensemble of machine learning algorithms to estimate the probability that a given IC driver will elect to take a time-off break in the near horizon. Once individual IC driver probabilities are estimated, we are able to use properties of a resulting Poisson binomial distribution to forecast the number of IC drivers available in various geographic and time-frame subdivisions and the associated estimate error.We will discuss the process of determining the IC driver attributes and historical behavior metrics that serve as predictor variables in our Machine Learning (ML) framework. We then use a two phase approach of clustering (unsupervised learning ) to partition drivers into groups with similar behavior patterns and then apply several ML techniques (supervised learning), including Bayesian networks, CHAID trees, logistic regression, and neural nets; and show how a weighted ensemble of these techniques can improve the predictive power over any individual algorithm. We also describe how this approach has led to substantial improvement in forecasting accuracy over previous ad hoc and time-series methods and has led to significant positive economic impact.We will conclude with a discussion of how this work can to be applied to more general problems of forecasting aggregate availability of individual workers whose demographic attributes and behavior patterns can be identified and correlated to self-governed work profiles.


Ted Gifford is a Distinguished Member of Technical Staff at Schneider National. His technical responsibilities include lead design of models/algorithms for various problems in logistics, transportation networks, revenue management, supply chain design, and capacity planning. He is also engaged in statistical analysis, data mining, and simulation in areas such as operations productivity, safety cost reduction, and vehicle maintenance. His management responsibilities include overseeing strategic partnerships with external research and development groups. In his most recent prior position, he served as Vice President of Engineering & Research, where he lead an organization comprised of groups engaged in supply-chain and transportation engineering, industrial engineering, statistical analysis, operations research modeling, business consulting, and development of decision support software tools. Mr. Gifford has an M.S. in Operations Research from Georgia Institute of Technology, an M.A. in Mathematics from University of California, Berkeley.


Shrikant Jarugumilli

Manager, Operations Research
BNSF Railway

Augmentation Of Strategic Optimization With Tactical Analytics For Better Operational Response In Rail Freight Blocking Network

BNSF Railway is one of North America’s leading freight transportation companies, with a rail network of 32,500 route miles in 28 states and three Canadian provinces. In 2014, BNSF handled more than 10 million shipments. Roughly 19 percent of those were moved in what is known as ‘merchandise’ trains. The merchandise rail shipments move between 2,300 + stations and consist of a group of railcars (referred to as blocks) ranging from 1 to 80 railcars, which are further grouped together to make full-length trains. The objective of forming the blocks is to minimize the number of handlings and miles traveled during the entire trip of a railcar from its origin to destination. Within BNSF, the Service Design team is responsible for defining the blocking and train schedules, a very complex task requiring an in-depth knowledge of the network (and restrictions), traffic profiles, and other federal regulations. While designing the block and train plans, the design team is constantly faced with the decision to find the right tradeoff between excessive handlings, out-of-route miles and good train size. Once the service plan is created, the plan path distance and the shortest path distance between the railcar’s origin and destination may be different due to several business constraints or tradeoff’s stated above. Though, both the blocking plan and the train plan can contribute to this difference in mileage (referred as Circuity), after performing several pilot studies, we have concluded that the blocking plan is a dominant driver among the two. Identifying opportunities for improvements and slightly tweaking both the blocking and train plans can help reduce the Circuity in the network. The Operations Research team working closely with the Service Design team has developed an analytical toolkit, referred to as Circuity Report, which complements and enhances the capabilities of the existing optimization tools. This toolkit is comprised of various modules, graphical and analytical, that provide improved traceability of railcar movement and identify tradeoff opportunities between handlings and out-of-route miles considering the changes to the traffic patterns and the blocking plan. Also, this toolkit provides the capabilities of prescribing the volume shifts on various lanes and drilldowns to trace the Origin-Destination pairs and the characteristics of the impacted waybills. The insights provided by the tool have directly resulted in adjustments to the volumes on existing blocking plans that provide the right trade-off between the handlings and out-of-route miles. In the past, the two teams have developed a series of strategic Transportation Service Planning & Design Tools (Interactive Block and Train Optimizer) that have been recognized with the Daniel H. Wagner Prize for Excellence in Operations Research Practice and have been semi-finalist for the Franz Edelman Award. Impact: Since the implementation of this tool in 2015, we have seen a big improvement in the various key performance indicators captured periodically. This talk will cover: • An introduction to the Merchandise Service Design business process. • An overview of the Analytical Tool Suite: Circuity Report and various modules of actionable reporting. • The positive business impacts that have resulted from the implementation of this tool.


Shrikant Jarugumilli is a Manager in the Operations Research and Advanced Analytics group at BNSF Railway. Currently, he provides analytical consulting and supports the development of decision support systems in the areas of Service Design, Capacity Planning, and Workforce Planning. Prior to BNSF, he received his Ph.D. in Engineering Management from Missouri University of Science and Technology (formerly University of Missouri – Rolla). He also has MS degrees in Applied Mathematics from Missouri University of Science and Technology and Industrial Engineering from Arizona State University. As a graduate student, he worked on research projects funded by Intel Corporation that involved developing automated workforce and capacity planning for semiconductor manufacturing. Over past several years, he has been an active INFORMS member serving as the Chair of the IPC organizing committee (2013 – 2014), Secretary of the Railway Application Section of INFORMS (2011-2013), and student officer at ASU and Missouri S&T INFORMS student chapters. In 2012, he received the Judith Liebman Award and the BNSF Achievement Award for his contributions to the Interactive Train Optimizer.