“STOP”: safer construction sites with AI

From GCF and Kiwitron, an intelligent device against accidents

Months of engineering and field training to learn to discern between human figures and "false positives". Now the experimental debut on the tracks, with the ambition to impose a "stop" to investment accidents in railway construction sites.

“Stop” – Supervised by Trained Optical Protection is, significantly, the chosen name for the unprecedented risk mitigation device that for the first time applies artificial intelligence to the non-negotiable issue of safety in railway track construction and maintenance sites. It is the result of a collaboration between the Research and Development Department of GCF S.p.A. and Kiwitron, a Bologna-based company that has been developing high-tech devices combining artificial intelligence, software, hardware, and robotics since 2019 with the aim of increasing safety levels in all industrial sectors.

"It is an ambitious project - explains Ilario Febi, engineer at GCF's R&D Department - that required overcoming the challenges posed by the application of computer vision systems to the particular environmental conditions of railway construction sites".

For experimental training, the system - consisting of one or more AI optical sensors connected to an on-board monitor and light and sound signaling devices - was installed on trains and machinery typically used in railway track laying, renewal, and maintenance operations: from the powerful Mark and Krauss-Maffei locomotives, to the Matisa portal trains, to the ballast regulators and magnetic trolleys.

On these vehicles, "Stop" had to be trained over the months to operate both day and night, at operational and cruising speeds, in all weather conditions and, above all, to recognize and distinguish human figures present "in the danger zone" - even from behind or crouching - and to signal their presence to operators in the cabin without the possibility of confusion with "false positives" in the monitoring area: objects, signage, materials on the track.

"One of the main advantages - explains Carlo Cavallari, technician at GCF's R&D Department - is that the system does not require people to use special signaling devices, such as additional battery-powered PPE that could be burdensome and intrusive".

Simply thanks to artificial intelligence, STOP is able to recognize any human figure, calculate the distance and activate the required signaling protocols when the figure breaches the safety perimeter of the convoy - on the monitor, via a LED light alarm device, and with an intense acoustic signal audible inside and outside the convoy - and consequently, activate the stop of the rolling stock.

"The Kiwi Eye system - comment with satisfaction Andrea Filippini and Alberto Bertaglia, respectively CEO and Business Development & Strategy Manager of Kiwitron - is making a difference even in the railway sector, demonstrating its effectiveness in drastically reducing accidents during rail maintenance operations".

A noteworthy feature of "STOP" is that it is an AI optical system that respects privacy (GDPR), is equipped with internal computational capacity and is engineered in advance to comply with the EU ACT criteria on non-self-evolving AI.

Now, after the advanced experimental phase and preliminary testing, the project enters a new step with the large-scale adoption of "STOP" on GCF and GEFER vehicles used for railway track construction.
"It is a highly customizable device - anticipates ing. Febi - and being a solution based on computer vision, it can be updated with new algorithms and detection techniques, continuously improving its capabilities and efficiency".

{ampz:ampzshortcode}