An Open Collaborative Research project in Zurich is using algorithmic technology to automate train movement and identify impediments to smooth, timely train travel.
Project Leader, Transportation
& Operations Research
Simulated scenarios have an obvious advantage over real world environments: When the Law of Unintended Consequences strikes, nobody gets hurt. Nowhere is this more true than in the railroad industry where a disabled train can spark a cascade of dispatching problems. To help create a more efficient railway system, a team of IBM researchers is working on a smarter transportation project that will help the industry optimize train movement across a nation's entire railway system.
The simulation project – part of an Open Collaborative Research effort to share open source code with university partner Zurich University of Applied Sciences – began in response to the increasing demand for rail transportation in Europe. Railway networks have become more congested, especially in hub stations.
Transportation planners compared two basic mitigation strategies: Building new railway infrastructure versus managing existing networks through smarter transportation solutions, such as algorithms. They asked a key question: What is the best way to evaluate train activity and metrics? The answer: Through simulation.
Modeling networks: an engineering challenge
Modeling train networks is a long-standing engineering challenge. Older train simulations involved studying track layout to get a fix on the effectiveness of various signaling systems. Indeed, some railway functions, such as signaling, are well served by existing physical simulations. The IBM team is using new algorithmic technology to address other network issues, such as network-wide dispatching; performance analysis and visualization, and passenger behavior and other impeding external events.
Dispatching in China
Comparable railway optimization efforts are under way in China, where IBM opened a Global Rail Innovation Center to advance next-gen rail systems.
The IBM team, coordinated by Marco Laumanns, project leader for the Transportation and Operations Research Group at IBM Research - Zurich, points out that the team's focus is on the simulation framework's "logical layer," the interface between the management layer -- which includes a sophisticated train scheduler -- and the physical layer -- which ultimately mimics the real railway network.
"Think of that middle logical layer as the place where data is shoveled back and forth between the other two layers," says Laumanns, who spoke in January about the evolution of railway scheduling optimization techniques at IT13 Rail. The day-long symposium drew experts from IT research, industry and government.
At present, a framework does not exist for a coordinated simulation of a complex railway system, such as the one that crisscrosses Europe. Only aspects of Europe's railway system have been simulated, including train runs and simple dispatching rules.
For commuters, an optimized and simulated railway network would help ensure a more predictable travel experience. For network operators, such a framework would offer up a system-wide picture of commuter and freight trains across countries -- and continents. Moreover, it would enable the seamless integration of existing railway tools.
For IBM -- and for optimization researchers overall -- the open source project will demonstrate the power of analytics and optimization science to develop more responsive control systems in high-speed rail and other logistics industries.