Depending on who you're conversing with, the "Edge" of the IoT may have different meanings. For the sake of this article, we are going to define it as that last computer that sits between the sensors (where the data is collected) and the Cloud. It's very close to where the data is obtained and where the action must occur.
Edge computing enables real-time or near-real-time processing, a reduction in the amount of data that's transferred, and also offers offline computing capabilities. This approach is vital for applications that require low latency, such as autonomous vehicles, industrial automation, and remote monitoring, where quick decision-making is essential. It also encompasses AI, which is an application area unto itself.
Edge computing enhances security and privacy by limiting the transfer of sensitive data to the Cloud, and it offers scalability and customization options to meet specific use cases. It complements Cloud computing, creating a holistic IoT ecosystem where Edge computing handles local, immediate tasks, while the Cloud manages resource-intensive, global analytics and long-term data storage. This distributed architecture optimizes IoT performance, responsiveness, and efficiency.
In most cases/applications, the "needs" at the Edge are similar, unless you have an extreme application with needs that are highly customized. In the more mainstream case, the needs are generally satisfied by high-performance computing, in a compact fanless package. This would be the case whether it's destined for an industrial setting or for commercial (smart retail) use. When AI enters the equation, the requirements are a little more stringent.
Edge computers built with NVIDIA's Jetson Orin make a compelling choice, for a host of reasons, starting with the platform's powerful AI capabilities. This suits it for applications that require advanced machine learning, computer vision, and AI inferencing at the Edge, all crucial for tasks like facial and object recognition, natural language processing, and autonomous navigation.
NVIDIA® provides comprehensive software development kits and libraries, including CUDA and TensorRT, making it easier for the developer to build and optimize AI models for Edge-based deployments.
The Jetson Orin is designed to be energy-efficient, which is crucial for any Edge device, regardless of where it's running, but more so for those operating in environments with limited power resources. When it comes to connectivity, the Jetson Orin platform provides a range of options, including high-speed interfaces like PCIe, Ethernet, and USB, enabling seamless integration with a variety of sensors and peripherals. Again, this is a critical feature for diverse edge applications.
In industrial environments, it's important to design with components that can operate in harsh environmental conditions, which is a key characteristic of the Jetson Orin. It's also scalable, meaning that it's available with a range of varying performance levels. Hence, developers can choose the right device for their specific application requirements. In addition, NVIDIA offers long-term support for its Jetson platforms, ensuring that devices remain relevant and supported for several years, which is crucial for industries with longer product life cycles.
One industrial Edge AI computer that's powered by the NVIDIA Jetson Orin NX is the MiTAC MA1. Aside from the incorporation of the Jetson Orin, the compact industrial computer operates in a wide temperature range, from -25°C to +60°C, and still doesn't require a fan. The case measures just 90 by 125 by 65 mm.
To maximize its versatility, the MA1 boasts an array of I/O, which includes HDMI (up to 4K at a resolution of 60 Hz); dual RJ-45 1-Gb Ethernet LAN' two USB 3.2 Gen 2 Type A ports; one D-Sub9 RS-232 (four-wire) interface; high-speed M.2 2280 PCIe; four 4 NVMe connections; and support for 5G/LTE and WiFi-6E.
Specific applications that appear to be a target for the MA1 include public safety, for disaster response, anomaly detection, and infrastructure protection; smart retail, for asset protection, autonomous shopping, and customer and store analytics; and traffic management, to provide real-time alerts, incident detection, and speed estimation.
The MA1 can also operate as an edge AI computer for retail applications, such as those that require video input and analytics at the output. As such, the MA1 provides centralized control and management of those IP cameras, making it easier to configure, monitor, and maintain the surveillance system within the retail store. The MA1 then aggregates data captured by the IP cameras, enabling comprehensive insight into customer behavior and product information. This data is essential for both theft prevention and an enhanced shopping experience for the customer. The scalability of the MA1 allows it to accommodate additional IP cameras or expand the scope of the retail surveillance system as needed.
From a software perspective, developers can hit the ground running. That ability comes thanks to NVIDIA's SDK and Linux support package. The former, JetPack 5.1.2, is a production quality release and brings support for the Jetson AGX Orin industrial module. It includes Jetson Linux 35.4.1 BSP with Linux Kernel 5.10, an Ubuntu 20.04 based root file system, a UEFI based bootloader, and OP-TEE as Trusted Execution Environment. The NVIDIA Jetson Linux driver package includes a Linux kernel, the UEFI bootloader, NVIDIA drivers, flashing utilities, and a sample filesystem based on Ubuntu.
In summary, MA1 plays a pivotal role in this smart retail application, facilitating the connection and efficient operation of IP cameras with the powerful AI capabilities of NVIDIA Jetson Orin. Its roles encompass central control, data aggregation, and network connectivity, all contributing to the overall effectiveness of the system in theft prevention and self-checkout while reducing backend workload and latency.
Regardless of what your Edge computing application is, it's likely MiTYAC's MA1 can be configured to fit your needs. The company has built a reputation for its global tech support through an FAE network, and experienced R&D team. If necessary, the industrial platforms can be white-labeled for specific customers.