Methods for the prediction of switching times of traffic actuated signal controls

• Applicant: Professor Dr.-Ing. Robert Hoyer
• Subject Area: Traffic and Transport Systems, Intelligent and Automated Traffic
• Term: since 2021
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 461855625

In urban road networks with signalled junctions, energy- and emission-saving driving essentially aims at reducing acceleration processes. This supposes that motor vehicle drivers know when to expect the end of the green time. Thus, when approaching a traffic light, it is advisable to coast off. A vehicle-infrastructure communication basically offers the opportunity to bring corresponding information of switching times from the optical signalling systems into the vehicles. However, the problem with traffic actuated signal control systems is that the times of phase changes are variable and must be predicted. Currently known forecasting methods show considerable weaknesses in the reliability of the results. This depends on the adaptive type of traffic dependency. The controllable forecast time horizon also falls short of overall expectations. In addition, the procedures require a considerable effort to supply the necessary data models, which sometimes has to be carried out for each individual traffic signal system. Overall, these methods reach their limits due to the complexity of traffic-dependent signal programs, in particular due to the randomness of individual traffic events. The project takes up this challenge and aims at the development and evaluation of a generally applicable, largely automated and modular procedure for predicting the switching times of traffic signal systems by means of methods of artificial intelligence. The anticipated broad applicability of the prediction method should enable a medium-term implementation in control units and traffic control centres. This in turn should reduce the effort for a medium-term area-wide implementation to the extent that a wider spatial availability and a superior quality of switching time prediction will be achieved. As a consequence, a sufficiently large, but so far unattainable acceptance among drivers is expected. By corresponding changes in driving behaviour, this will lead to a reduction of energy consumption and emissions harmful to the climate and health. The concept for the switching time prognosis process is intended to unite the specific system knowledge for signal program design and traffic signal system operation with the automatic evaluation of historical and current process data and thus enables different types of signal control systems to be integrated. For this purpose, appropriate data models have to be selected and tested. In addition, tests are being carried out to determine the extent to which partial forecast cascades and process data from neighbouring signal control systems play a part in improving the forecast quality. Artificial intelligence methods take on a key role in this context. The results are evaluated with original process data of the City of Kassel and compared to a competitive method.

• DFG Programme: Research Grants