AI-enabled computer is designed for rail applications
Based on the Nvidia Jetson AGX Xavier industrial module, the AVA-RAGX is a compact, rugged computer from Adlink. The Nvidia industrial module operates at 32T operations per second (TOPS), sufficient compute density for edge applications in harsh, space-restricted environments says the company.
The AI-enabled AVA-RAGX expands Adlink’s EN 50155-compliant AI enabled video analytics (AVA) portfolio which allows developers to optimise configurations for rail projects, using AI at the edge.
The AVA-RAGX is suitable for both onboard and wayside railway applications, for example in railway intrusion detection, train station video surveillance, onboard passenger security, and railroad hazard detection.
“AI is accelerating the digital transformation of the railway industry to improve operational efficiency, deliver smarter and safer customer services, and create new business opportunities,” observed Eric Kao, general manager of Adlink’s networking communication and public sector business unit. AI computing helps railway companies focus on differentiating and transforming their end applications, he added. AVA-RAGX demonstrates Adlink’s Elite status partnership with Nvidia to accelerate product development with access to Nvidia’s latest GPU technologies and technical support, he added.
Adlink is an Nvidia Quadro Embedded Partner, OEM Preferred Partner and Jetson Elite Partner. This status means the company has Nvidia’s permission to build chip-down GPU solutions for ruggedised embedded markets. In addition, it can also leverage access to advanced processing technologies and technical support, driving innovative, open standards-based heterogeneous computing for AI and machine learning applications.
AVA-RAGX can be used for real time video and graphics analysis applications for increasingly complex railway operations. It includes an eight-core Nvidia Carmel CPU and a 512-core NVIDIA Volta GPU, combines AI, machine learning, edge computing and IoT technologies. The rugged, fanless computer is certified to EN 50155.
It is also compact, measuring 288 x 190 x 72mm, and targets space-restricted environments, typical in edge railway applications with small enclosures. The AVA-RAGX also provides a full complement of I/O ports, including four M12 Gbit ports and wireless connectivity (5G / LTE and Wi-Fi), as well as four digital input and four output 24 to 110V DC with 1.5kV isolation.
When integrated with an advanced camera system, radar, lidar, GNSSC and AI software stack, the AVA-RAGX can provide information about the train’s surroundings and exact position. One challenge particular to the rail industry addressed by the AVA-RAGX is the long braking distance of the train. The computer can detect emerging hazards far ahead to increase warning time when connected to appropriate sensors, says Adlink. The hazard detection system can improve operational efficiency and safety and helps prevent accidents, especially when trains run at night or in adverse weather conditions that reduce visibility, adds the company.
When installed on an inspection train, the trackside equipment fault detection system processes images captured from wayside equipment, such as the pantograph and track, in real time. Driven by Nvidia’s GPU accelerated computing, the system can effectively identify potential equipment faults at a train speed of up to 75mph (120km/h).
Incorporating AI at the edge into legacy CCTV networks enhances passenger travel conditions, says Adlink. Instead of relying on CCTV playback footage, the AVA-RAGX can perform real time monitoring, for example to check unattended luggage, seat occupancy and passenger counts. When deployed in station surveillance systems, the AVA-RAGX can analyse the real time video stream received from station platforms to detect suspicious or aggressive human behaviour, crowd movement and even face mask-wearing.
