Digital Twin Evropská – Technical Brief
The Digital Twin Evropská represents an on-line calibrated digital model of an area around Evropská street in Prague that takes into account individual personal transport, public transport and (to some extent) pedestrians and other vulnerable road users. The system is built around Eclipse SUMO traffic microsimulator, an open-source system that is being developed at German DLR.
The system consists of two key components:
- Data concentrator, a system that is responsible for receiving input data from different sources, and that is capable of basic pre-processing
- Digital twin system, represented by a custom extension to SUMO
Data concentrator (Datastore)
Due to the need to maintain a relatively low data flow to and from the optional cloud storage, data from the detectors are stored in a dedicated database at the premises of the Czech Technical University in Prague in the intermediate phase, from where a subset of them is further distributed. These data are then used to generate traffic simulations, predictions and modelling of traffic scenarios, allowing road managers to better plan and manage traffic.
Digital twin
The basis of the simulation model of the transport digital twin is a virtual copy of the existing transport infrastructure in the modelled area: a microsimulation model contains individual streets, parking lots, public transport interchanges, bus stops, together with information about commercial and cultural centres or residential areas. For initial calibration purposes, this model is linked to off-line (historical) data provided by the concentrator and is calibrated according to this data. The digital twin then uses a statistical modelling approach for continuous calibration of the model that results in changes in the routes of the simulated vehicles and perturbation in vehicle demand.
The digital twin over the off-line data uses a total of three custom extensions for SUMO and one separate computational subsystem responsible for adapting the microsimulation to the measured data. The basic tasks of these modules are as follows:
TraCI flow input — This module first determines which traffic data from specific traffic detectors should be used as inputs and then sequentially retrieves the data from the traffic detectors and generates a traffic demand for the microsimulation. Based on the information about typical trips that start at a given traffic network entry during a given simulation period, the module instructs the microsimulation to instantiate a certain type of vehicle to enter a certain edge of the modelled traffic network.
TraCI detector output — This module represents the "counterpart" to the input module. It continuously aggregates information from the traffic detectors in the simulation so that the temporal resolution of this information matches the temporal resolution of the real measurements. This aggregated data is both stored by the module in the form of XML and CSV files and are also provided to the Path flow estimation and route adaptation system for estimating the most probable paths and traffic flows.
Path flow estimation and route adaptation is a module that uses statistical methods over a state-space model to make a short-term prediction of the traffic evolution in the real world and in the traffic microsimulation model, and based on this prediction determines, in which areas of the traffic model there will be significant deviations from the predicted values and which modelled traffic flows need to be changed so that the short-term prediction of the traffic state in the model corresponds to the short-term prediction of traffic in the real world.
TraCI continuous calibration uses information about typical vehicle routes in the modelled area and the differences between the detected vehicle counts in the microsimulation and in the real area to modify, if necessary, the future vehicle routes in the simulation so that the simulation more faithfully matches the short-term prediction of the traffic condition in the modelled area by the system for estimating the most likely routes and traffic flows described above.
Contacts
Jan Přikryl (jan.prikryl@cvut.cz)