Russian Developers Teached a Neural Network to Optimize Railway Traffic
Russian specialists have developed an intelligent train traffic management system based on neural network (TMS «Prognoz»), which will provide more accurate train schedules in a few seconds without human intervention, — RIA Novosti was told at the NTI Competence Center «Artificial Intelligence» based on the Moscow Institute of Physics and Technology, which acted as the developer of the platform.
The customer of the system was the Signaling Division of the 1520 Group of Companies, the main manufacturer and supplier of computer-based control systems for railway transport in Russia and neighboring countries.
The artificial intelligence modules built into the system make it possible to optimally rebuild the train schedule in seconds, offering the dispatcher a choice of several options optimized with different priorities. The dispatcher only has to choose the most suitable option and confirm the changes.
It is clarified that the same modules are able to calculate the forecast situation, show the dispatcher how events will develop under the current train situation, and how to correct the situation with the forecast shown.
«Currently, the system uses three independent artificial intelligence modules, each of which has its own data processing methods. Based on the results of trial operation, the best method will be selected, or a synergistic processing module will be created on the basis of all three.The site for the trial operation of the innovative product was provided by the Kazakhstan Railways» — added in the centre.
