Software
MIGRATION PATTERN ANALYSIS
In collaboration with researchers at the Intelligent Transportation Systems Division and developers at Ericsson Nikola Tesla, a prototype software called “Migration Pattern Analysis” was developed for the laboratory. The mentioned solution is based on anonymized monitoring of mobile users, and enables the following features:
- Origin-destination matrix;
- Heat concentration maps of users;
- Graphic representation of user migration between individual or sets of regions;
All detected mobile users can be filtered by nationality, gender, and OS they use on their device, and data can be obtained during a characteristic day inside and outside the tourist season. The development of the software solution will be implemented in several phases over the next three years. Some of the key features of this tool are:
- Identification of user locations;
- Classification of users by their country;
- Identification of the number of users by sector or specific location;
- Integration with demographics data;
- Search users by the OS of the user’s mobile device;
- User search by gender and age;
The current version includes 4 characteristic days: a typical working day during and outside the tourist season, and one characteristic weekend day during and outside the tourist season.
Software covers several Croatian countys: Istarska, Primorsko-goranska, Ličko-senjska and one part of Karlovačka .
Example of OD migration matrices of mobile users. By selecting an individual cell (or set of cells), it is possible to gain insight into outbound and incoming trips to neighboring cells.
Heat maps. It is possible to get a graphical representation of user concentration by parameters such as age, nationality, OS, etc. It is possible to run the animation over several consecutive heat maps.
Identification of the users per cell.
In 24 hours, it is possible to get a detailed graphical representation of the user’s migration between two cells and group of cells.
Distribution graph of incoming and outgoing foreign users migration. Possible detection of users from foreign countries who were first registered in the Republic of Croatia. Ability to analyze border migration and analyze arrivals and departures of foreign nationals at airports.
A framework for traffic control based on deep reinforcement learning and integrated interface for Vissim traffic simulator
- The framework is used to easily launch deep reinforced learning for traffic management with the Vissim simulator interface. The framework is written based on the Python Keras (backend TensorFlow) environment.
- Ability to convert microscopic traffic flow parameters from a modeled motorway network in Vissim into an “image-alike” figure suitable for deep learning
- The results of a large number of deep learning-based simulations can be saved and structured in Excel format
Example of displaying a structured record of the results of Vissim simulations after performing Variable Speed Limit control based on the deep reinforced learning.