Use of Simulation-based Forecast for Real Time Traffic Management Decision Support: the Case of the Madrid Traffic Centre
A Torday, J Barcelo, G Funes, Transport Simulation Systems, ES
This paper presents a simulation-based traffic forecasting solution for supporting traffic centre operators in choosing the most efficient management strategies in response to a given event. Its benefits are illustrated by the concrete case of Madrid
As it is well known, anticipation is a definitive keyword in the transportation field. Indeed, being able to forecast the evolution of the traffic in a network is a basis on which many traffic management strategies and multiple ITS applications rely. Real time prediction capabilities are therefore becoming a concrete requirement for the management of networks, both for urban and interurban environments.
Numerous techniques for real time forecasting have been developed. However, considering the complexity and the size of todays transportation networks, analytical solutions have been proven in most cases to be inappropriate. On the other hand, traffic simulation is increasingly able to deal with very large networks, former computational limitations no longer providing a barrier to its application in real time.
Dynamic traffic management is clearly the most typical application where real time simulation offers evident advantages. In this case, the forecasting capabilities are used to compare the performances of different management strategies in response to a non recurrent event detected. These scenarios are composed by a set of actions - such as lane closure, rerouting with VMS, speed limits variation or ramp metering - that can be activated manually or under detection triggers. To decrease the operational time, most of the management scenarios are already implemented in the simulation model (within a scenarios catalogue) and are activated when the performance of a particular scenario has to be assessed through simulation.
The process of traffic forecasting by simulation provides may provide large amounts of traffic data as outputs. Thus, in order to keep the computation time reasonably low and offer the operator a synthetic view of the results, a clear definition of indicators for the scenario comparison is necessary. These indicators should provide the operator a multi-objective approach allowing him to understand why a scenario seems to be more adequate than another. This indicator definition is obviously depends on the particularities of the managed network as well as strategic and political considerations. A multi-objective scenario comparison, for example, may point out the scenario offering the lowest global travel time but without passing a fixed noise level in a particular residential neighbourhood and, in any case, avoiding any vehicles queuing in a specific and safety sensitive tunnel.
A real time simulation-based decision support tool is currently under implementation in the Madrid traffic centre. The newly opened M-30 urban highway (composed of a significant number of tunnel sections) is subject to many safety considerations, and many traffic evacuation and rerouting actions may be applied in response to incidents. For this reason, a tool capable of anticipating the consequences of these actions on the neighbouring network over the following critical 15 or 30 minutes is necessary to choose which set of actions on the city can support these safety measures efficiently while minimizing the impact on the rest of the traffic.
For this application, the Madrid City council decided to use the AIMSUN ONLINE solution. This application is fed in real time by traffic measurement in order to be able to deduce the current traffic status on the streets and the actual demand (modelled as an origin-destination matrix). With Control plans changing dynamically during the day, AIMSUN ONLINE also reads the current control plan operated at each network intersection. M-30 safety actuations as well as any other incident previously detected and still existing are automatically (and in some case manually) loaded into the simulation model before starting the parallel simulation runs. Each simulation considers a concrete set of actions (strategy) that might be applied in order to improve the network situation compared to the ?do nothing case?. In this implementation, four different strategies can be simulated simultaneously; however the number of strategies that can be simulated is directly proportional to the numbers of CPUs available on site.
Once simulated, which means between one and two minutes after the initial call, a resumed view of the results is provided to the operator including snapshots of predicted traffic congestions and performance indicators. These results ultimately allow the operator to quickly see, first, if any strategies improve the situation compared to the ?do nothing? case and, if yes, which ones offer the best performance. If the suggested solution is accepted, a single validation click on the screen leads to the field application of the selected strategy. An off-line operated module allows the prediction capabilities of the simulation to be evaluated each day by comparison with real data stored during the day, offering the City Council a measure of confidence in the reliability of the system.
Association for European Transport