NVIDIA Jetson-based embedded system for railway application
Posted: 31 January 2023 | Syslogic AG | No comments yet
The AI Railway Computer RML A3 is specifically designed for rolling stock applications and is based on the NVIDIA Jetson AGX Xavier.
As a ‘Preferred Partner’ in the NVIDIA® Partner Network, Syslogic is offering AI-assisted embedded computers.
The company’s AI Railway computers are the first NVIDIA Jetson-based systems certified and approved for the railway industry. They are particularly suitable for use in railway vehicles, but are also used for trackside or video surveillance applications (closed-circuit television; CCTV).
NVIDIA Jetson is a leading platform for artificial intelligence ‘at the edge’. Accordingly, developers across industries are using NVIDIA Jetson Developer Kits as the base for their AI applications in areas such as sensor fusion or inferencing. However, series production in railway technology requires robust hardware that is qualified for these types of applications. These are exactly the kinds of solutions Syslogic offers, supporting customers in using NVIDIA Jetson technology in rolling stock applications.
For the Railway Computer RML A3, Syslogic combines the NVIDIA Jetson module with its own carrier board and a robust housing. All the electronics are designed for reliable long-term railroad operation. The device is passively cooled and features M12 screw-on connectors. Correspondingly, the AI Railway Computer RML A3 is certified according to EN50155. Thus the Railway Computer is suitable for operation in temperatures ranging from –40 to +70 degrees Celsius (–40 to +158 degrees Fahrenheit). It also complies with the EN45545-2 HL3 fire protection standard and is shock and vibration resistant in accordance with EN61373.
The new AI Railway Computer RML A3 features an integrated power supply with a DC input voltage range of 9V to 45V. This allows the Railway PC to be directly connected to the vehicle power supply.
As a computer vision unit, the railway computer performs tasks such as object recognition, environmental perception, intelligent video analytics, or control of autonomous vehicles. Four PoE Gbit interfaces (Power over Ethernet) allow four sensors such as GigE Vision or IP cameras to be connected without requiring their own power supply. The PoE ports are configured as x-coded M12 connectors. Furthermore, digital I/Os, analog inputs, CAN or RS-232/422/485 interfaces can be added.