MIPT experts developed and implemented an automated system for optimizing the “Prognosis” railway transport program. MIPT prepares a new train schedule in emergency situations within three seconds, socialbites.ca was told about this.
Artificial intelligence is already helping to regulate train traffic on the Zhetygen-Altynkol section, which is of strategic importance for the Kazakhstan economy. This section of the railway goes to the Chinese border. The length of the highway is 191 km. The organization of train traffic on this section is fully automatic. At the slightest temporary malfunction, the neural network can generate a prediction about the development of the situation for the coming day, as the delay of one train triggers a chain of events that affects the traffic situation on the entire highway.
To find a way out of an emergency situation (preparing a new traffic plan taking into account repair work and redirecting trains to new rails), the Forecast intelligent system needs no more than three seconds. A professional dispatcher will need much more time to solve problems of similar complexity. However, there are situations where the neural network cannot solve as effectively as an experienced expert.
“If we are talking about the standard program, the expert (grafist) who prepares the traffic programs is given seven days. It needs to create the optimal program within a week: the wishes of all services must be taken into account and changes must be coordinated with responsible persons. To do this, sometimes you have to re-enter the data and recalculate several times. Recalculating the schedule is a long process. The part of the AI that creates the regulatory charts helps the expert greatly. The system takes into account more than 50 parameters and constraints critical for the organization of traffic: from the priority of movement of a particular train to the condition of the roadway on a particular section,” said MIPT Laboratory chief engineer Andrey Novikov on the topic Wave Processes and Control Systems.
On complex sections, the Prediction neural network system works as a clue for the dispatcher, and sections with fewer than 20 trains passing per day will most likely be fully automated in the future. Intensive areas where maneuvering work takes place cannot yet be operated without a professional.
Previously in Russia taught Artificial intelligence detects dolphins by whistling sounds.