Application Highlight: Aetina AIB-MN32 Enables AI Robotics Solutions

September 16, 2024

Sponsored Story

Robotics are one of the most important use cases for AI in Embedded Computing, and the applications and implications are only beginning to be identified.

But we’re not going to talk about the future applications of AI and robotics, right now. Instead, it’s important to look at the applications that are already in the field and working today.

The key verticals that are leveraging robotics right now are manufacturing, warehouse, security patrolling, and increasingly, environmental monitoring.

Most of these applications are evolving, too. Even in manufacturing, robots are becoming more mobile, working in several locations throughout facilities, or even roving around and through areas. This requires low-latency processing and connectivity, high-fidelity sensing and computer vision solutions, and well-trained AI. That’s only the beginning, but it’s a strong place to start.

Let’s look at a specific case.  

Application Use Case: Daxbot

Dax is an AI Service Robot designed for urban spaces and as of now, it has been optimized for three applications: delivery, security patrol, and gathering right-of-way ADA compliance and engineering data.

Dax robots are interactive urban units that have a “neck” and use facial expressions to communicate so that people are set at ease and can more easily determine what they are doing. The company built them initially to handle short-range deliveries that are not cost-effective for people to make, but their functions made them ideal for their other applications like how Dax is gathering ADA compliance data to make accessible sidewalks more possible.

On the software side, DaxOS was built to operate smoothly in uncontrolled environments. The custom OS favors Node over Python or C++ and includes features like multi-chip voting and a centralized data store. It contains safety-related code in the master so that critical functions don’t rely on a node that could die while the master is left running.

This represents a departure from the usual OS architecture for robotics. Traditionally, robots run on top-down operating systems that rely on complex algorithms to complete definable tasks. Instead, Dax uses an Aetina AIB-MN32, powered by NVIDIA Jetson Orin, to prioritize and complete tasks in uncontrolled environments, on the fly in real-time. This means that DaxOS programmed via Node and using multi-chip voting helps Daxbot make decisions in any uncontrolled environment, and complete undefined tasks as needed to arbitrate multiple (and potentially conflicting) goals.

Aetina AIB-MN32

The Aetina AIB-MN32 is a premier board that’s designed for Edge AI Embedded Computing. It’s powered by the NVIDIA Jetson Orin and is built to be ready for any AI application, including robotics in uncontrolled or unpredictable environments.

Daxbot chose the Aetina AIB-MN32 because the AI Edge processing capabilities made all the desired functions possible, including the expansion slots that support both Wi-Fi and LTE modules for reliable connectivity, which they said they could not find from other suppliers. Aetina also designed an audio board for Dax along with a wide range of other peripherals to create a robust and specialized compute that is designed to allow Dax to solve problems in a wide range of contexts, especially real-time computer vision and neural network tasks.

The NVIDIA camera/vision hardware is the most essential function for Daxbot because the navigation and interaction strategies all use stereo cameras rather than LIDAR. The Orin-powered compute enables the many complicated but necessary tasks to be executed in real-time, the company said, and the integrated GPU and hardware accelerators unlock many AI inference tasks.

The Aetina AIB-MN32 is a feature-rich all-in-one Edge AI inference engine for robotics. It includes:

• NVIDIA Jetson Orin NX 8GB
• 12 to 24 VDC input
• Operating temperatures from -25°C to +80°C (-13°F ~ +176°F)
• 70 TOPS
• 1024 Core NVIDIA Ampere, with 32 Tensor Cores
• 6-core Arm Cortex - A78AE v8.2 64-bit 1.5MB L2 + 4MB L3
• 8GB 128-bit LPDDR5 102.4 GB/s
• 1 x M.2 B-Key/E-Key/M-Key (NVMe 128GB built-in)
• 2 x RJ-45 GbE Ports
• Display 1 x HDMI 2.0 Type A
• Audio Line-out / Line-in / Mic (optional with daughter board)
• RTC With supercapacitor, battery (optional)
• Camera Input 1 x 8-Lane MIPI Expansion Connector (120-Pin)
• LAN 2 x RJ-45 GbE Ports
• USB 2 x USB 3.2 Gen2 Type A (supports up to 10Gbps shared)
• 1 x OTG Type-C
• I/O Interfaces 5 x GPIO, 1 x SPI, 1 x I2S, 3 x I2C, 1 x UART, 1 x UART (Debug Only),
• 1 x RS-232, 1 x CAN (Isolation; support CAN FD), 1 x RS-422/485 (2-in1), 1 x microSIM Card Slot
• Vibration 1 Grms, IEC 60068-2-64, random, 5 ~ 500 Hz, 1 hr/axis
• Shock 10 G, IEC 60068-2-27, half sine, 11 ms duration

Additional Resources:

• Product Page:  https://www.aetina.com/products-detail.php?i=551
• Datasheet: https://www.aetina.com/dl.php?i=1178
• Success Story: https://www.aetina.com/about-news-detail.php?i=1008