The Research Institute of Russian Railways will compare concentration of attention of train drivers and the automatic vision system

Until September 2023, the Research Institute of Russian Railways JSC “NIIAS” must conduct tests for the purpose of finding out who will see obstacles better — the machine or a person

As Pavel Popov, Deputy Director of NIIAS, told at the conference “Functional Safety of Complex Technical Systems Based on Machine Vision and Artificial Intelligence,” which was held on March 17 at the ExpoForum Convention and Exhibition Centre, within the Exhibition “EXPOTECHNOSTRAZH. The Day of Advanced Technologies,” the experiment is planned as part of researching the machine vision system that can automatically detect obstacles on railway tracks.

As Pavel Popov noted, Russian Railways already has the group of 10 drivers interested in the tests. The tests will be conducted at the Shcherbino test ground. “We will record in the protocol who saw better — a human or the system,” he briefly explained the essence of the experiment.

According to the expert, even today, in informal observations, the results shown by the system and the human are comparable in terms of the range of obstacle perception. For implementing this system, AI is used, which, according to Pavel Popov, has become “the key to understanding sensor data.” The most urgent problem of the development is the infinite number of scenarios of the situation related to the railway that needs to be put in the system so that it could respond correctly. At the current stage, as part of the tests, the researchers recorded the situation when the system saw an obstacle that did not exist in reality, but, slowing down a little and driving closer, recognized the error.

More

The Research Institute of Russian Railways will compare concentration of attention of train drivers and the automatic vision system

Until September 2023, the Research Institute of Russian Railways JSC “NIIAS” must conduct tests for the purpose of finding out who will see obstacles better — the machine or a person

As Pavel Popov, Deputy Director of NIIAS, told at the conference “Functional Safety of Complex Technical Systems Based on Machine Vision and Artificial Intelligence,” which was held on March 17 at the ExpoForum Convention and Exhibition Centre, within the Exhibition “EXPOTECHNOSTRAZH. The Day of Advanced Technologies,” the experiment is planned as part of researching the machine vision system that can automatically detect obstacles on railway tracks.

As Pavel Popov noted, Russian Railways already has the group of 10 drivers interested in the tests. The tests will be conducted at the Shcherbino test ground. “We will record in the protocol who saw better — a human or the system,” he briefly explained the essence of the experiment.

According to the expert, even today, in informal observations, the results shown by the system and the human are comparable in terms of the range of obstacle perception. For implementing this system, AI is used, which, according to Pavel Popov, has become “the key to understanding sensor data.” The most urgent problem of the development is the infinite number of scenarios of the situation related to the railway that needs to be put in the system so that it could respond correctly. At the current stage, as part of the tests, the researchers recorded the situation when the system saw an obstacle that did not exist in reality, but, slowing down a little and driving closer, recognized the error.

More