Developing Real-Time Traffic Models to Optimize Urban Transport Systems

Real-time traffic models could soon become feasible by zooming out and studying traffic at the city scale, thanks to ever expanding sensor and computing networks. This should pave the way for smarter traffic management schemes that would vastly expand the capacity of urban infrastructure.

When cities prosper and begin to outgrow their infrastructure, traffic often grinds to a halt. Expanding roads and highways is one common remedy to urban congestion. Recently, Nikolas Geroliminis from EPFL’s Urban Transport Systems Laboratory was awarded a starting grant of 1.25 million euros by the European Research Council to explore a second approach: increasing network passenger capacity through advanced traffic management schemes based on developments in monitoring, logistics, modeling, and control of urban traffic. The grant will be paid out over five years and will finance four PhD students and PostDocs who will work in these key areas.

The ERC project, entitled: “Modeling and controlling traffic congestion and propagation in large-scale urban multimodal networks,” targets to operate traffic in future cities in a holistic way that, until now, has been impossible. It tackles the problem of modeling and optimization in large congested traffic networks with an aggregated realistic representation of traffic dynamics and route choice for multiple modes of transport.

Attempts to accurately model traffic and predict its evolution tend to fall short for a number of reasons. One of them is lack of perfect information. It is impossible to exactly know how many people are on the streets at any given time, where they are headed, and whether or not they follow traffic rules. Psychology is another: drivers act irrationally as they weave their ways through a complex web of streets. And since the entire traffic network is connected, congestion can spread across large portions of cities in ways that are difficult to anticipate.

But as monitoring techniques mature, cities are becoming smarter. ‘Big mobility data’ from multiple sensors (GPS of taxis and buses, loop detectors, Bluetooth devices) provide a unique social observatory, and help to paint a more detailed picture of the traffic situation at any given time, but because of the complexity of the problem there is still a lot of uncertainty. “Mobility will advance through the integration of big data, the understanding of multimodal patterns, the coordination and optimization of urban efficiency for the travel of people and goods”, says Nikolas Geroliminis, whose research group, the Urban Transport Systems Laboratory is member of EPFL’s Transportation Center.

Congestion governance is currently fragmented and uncoordinated, and traditional approaches use more detailed models with a higher degree of unpredictability and complexity that cannot be solved in real time. One way of dealing with this complexity involves splitting the problem into smaller, more manageable regions. But modeling the behavior of each individual vehicle is unrealistic and computationally impossible. “Our recent research shows that we can develop realistic models of congestion without having to know the exact position of every individual vehicle in a city. And we are able to go out and measure all of the parameters used in our models,” he says.

The method aims to dissect the city into regions in which traffic can be described in elegant mathematical terms. While there is no clear relationship between traffic density and traffic flow on individual road segments, this changes when several roads are grouped together in the right way. Doing so leads to the emergence of a relationship that assigns a traffic flow rate to each density. This relationship also helps to determine the critical density at which that group of roads becomes congested. By dividing a city’s road network in this way, computer models of traffic can be simplified to the point that real-time management of traffic in an entire city become feasible.

With a better understanding of when and where traffic jams form, more efficient ways to control traffic can be designed and implemented. According to Geroliminis, these can favor public transport, efficient goods delivery, and car sharing options. By the same token, emergency services such as ambulances would benefit. -- EPFL