Advanced machine learning for traffic light management

Description
This commitment aims at leverage the longstanding partnership between the city of Darmstadt, the TU Darmstadt, and the urban institute® to (i) monitor and evaluate the green phase shifting, and (ii) employ advanced machine learning to recommend an optimal solution for the fully adaptive traffic light management.
 
Sustainable Urban Mobility
 
Integrated Planning, Policy, Regulation and Management

 

In 2011 the road and civil engineering office of the city of Darmstadt and the Technische Universität Darmstadt decided to investigate the questions of how traffic data can be made available to the general public and interested third parties. A first system was deployed 2011 based on the da_sense platform developed by the Telecooperation group under Prof. Max Mühlhäuser and Dr. Immanuel Schweizer. In 2013 the platform was transferred to the Urban Institute as Urban[Traffic!]Pulse. It now provides real-time access to over 880 million traffic data points and additional services in Darmstadt alone.
Today, the city of Darmstadt is planning to invest heavily into adaptive traffic light management to optimize traffic flow and reduce emissions. The planned activity is implemented in two stages: First, traffic lights are coordinate to provide optimal green phase shifting. Second, a fully adaptive traffic light management is implemented. 
While the first stage is comparably inexpensive, the second stage involves heavy investments (even for small test deployments) with unclear benefits. Today’s solutions are, thus, not scalable across or between cities. Fortunately, Darmstadt is uniquely positioned as we can build upon the fiber communication network employed between traffic lights, the large amounts of historical data points, and the data gathered in real time. Thus, we are able to closely monitor and evaluate the outcome of the first stage, i.e., the green phase shifting, on a city-wide level. This will enable us to answer key questions such as: How much energy can be saved? How much can GHG emissions be reduced? What is the best ICT architecture for optimizing traffic and gathering data (centralized vs. de-centralized)? Based on the system architecture and with the knowledge acquired, we will simulate an adaptive traffic light system employing advanced machine learning for almost optimal real-time adaptation. We are not only able to build a first of its kind ML-based traffic light management system, but to evaluate the second stage before it is deployed and the heavy investment is done. 
In summary, this commitment aims at leverage the longstanding partnership between the city of Darmstadt, the TU Darmstadt, and the urban institute® to (i) monitor and evaluate the green phase shifting, and (ii) employ advanced machine learning to recommend an optimal solution for the fully adaptive traffic light management. The commitment is, thus, related to three actions in the operational implementation plan: “Big Data for planning and management” (6.2), “Road systems” (3.5), and “Open up intelligence in urban transport system” (1.4)

 

(i) The partners aim at accelerating the development process and usage of innovative technology to improve the traffic flow. It has been estimated that fighting traffic congestions can reduce GHG emissions by up to 30%. Hence, improving traffic flow directly contributes to the EU sustainability targets. The impact monitor will also help communicating the benefits between citizens, public, and private entities. (ii) Open data is crucial to all partners in the consortium. In fact, the starting point of the collaboration between the city of Darmstadt and the TU Darmstadt was the desire to open the traffic data to the public. And we are committed to keep the complete raw traffic data open. (iii) We will investigate open standards, e.g., DATEX II. (iv) We will re-use the existing traffic detectors (induction loops and cameras) and make use of the data collection framework already set up for past projects. Using machine learning we do gain additional knowledge out of the provided data.

For this proposal the city of Darmstadt cooperate with the TU Darmstadt and the urban institute. As the responsible partner for the traffic infrastructure, the city of Darmstadt is actively engaged and motivated to improve the traffic flow in the city. For the TU Darmstadt the following research labs will participate: Telecooperation lead by Prof. Mühlhäuser, Knowledge Engineering lead by Prof. Fürnkranz, and Transport Planning and Traffic Engineering lead by Prof. Boltze. They will contribute their expertise in ubiquitous computing, machine learning, and traffic engineering respectively. With the urban institute a SME is joining the proposal with their data gathering platform UrbanPulse. This illustrates our commitment to integrate partner from the private and public sector from the start. This will allow us to innovate across the value chain, from research to market. It also shows how we unite actors from both ICT and the mobility sector to offer an innovative, integrated solution.

Is the commitment potentially open to additional partners and if so, in which priority area (please tick the relevant options)?

  • Sustainable Urban Mobility
  • Integrated Infrastructures and Processes across Energy, ICT, and Transport

 

(i) To reach our overall target of improved traffic flow and reduced GHG emission, the proposal makes two substantial contributions. First, implement and monitor green phase shifting at two main traffic light in Darmstadt. Second, implement and simulate adaptive traffc light management with machine learning on real traffic data for all of Darmstadt. (ii) The overall project will commence in 2015. Between 1/2015 and 6/2015 two of the most frequented crossroads will be configured to provide data to subsequent crossroads and for the green phase shifting. Investigating the adaptive traffic management is currently planned from 3/2015 to 11/2015. In 6/2015 we start the monitoring and comparing the outcome of the system with the historic data. In 12/2015 we will provide a recommendation for the investment and the fully adaptive traffic light management. (iii) Monitoring the impact is a core part of the proposal. We will use real-time as well as historical traffic data to report the impact.

Results expected in

  • 2015 Two main crossroads are configured as master to provide data for the green phase shifting. Adaptive traffic light management is simulated. Monitoring and evaluation results will be presented.

 

Management Team

Default profile image
Manager
Default profile image
Manager