The growing accessibility of artificial intelligence (AI) for industrial edge applications presents system developers with new time-to-market pressures to stay ahead of the competition. Likewise, device manufacturers must decide what edge AI functionality will successfully differentiate their products, and whether that involves audio- or video-based automation or intelligent human-machine interface (HMI) capabilities. To this end, engineering teams benefit from practical, modular solutions that reduce development complexity while allowing for later upgrades as AI workloads—and chipsets—evolve. This blog will discuss typical edge AI use cases in industrial automation before introducing energy-efficient, AI-ready SMARC-based hardware solutions from SECO. It will then describe how SECO’s wider software and development ecosystems support rapid, scalable deployment.
Edge AI is reshaping industrial automation. By placing intelligence directly within machines, controllers, and gateways, original equipment manufacturers (OEMs) and factory operators can provide lower-latency inference, consistent safety awareness, and stronger data security. Here, the cloud’s value shifts from processing to ecosystem orchestration through fleet management and aggregated analytics for remote diagnostics.
For intelligent machines and human-machine interfaces (HMIs), edge AI provides numerous safety and security benefits. AI audio systems can detect operator distress calls for automatic equipment shutdown and provide operator recognition for user-specific functionality or controlled access. Similarly, AI vision systems can monitor personal protective equipment (PPE) compliance and unsafe zone occupation to reduce incidents on factory floors.
However, AI frameworks, toolchains, and hardware platforms are continuously evolving, which presents challenges for developers deploying AI at the edge. Securing competitive advantage also demands rapid time-to-market.
Standards-based architecture and commercial off-the-shelf (COTS) solutions help reduce complexity and risk while enabling platform reuse across product lifecycles. For industrial edge AI, modular platforms with integrated compute, AI acceleration, and industrial I/O cut development time by providing validated architectures on which to base products. By accelerating integration, simplifying scalability, and shortening deployment cycles, these platforms allow OEM teams to focus on delivering application-specific functionality, differentiated performance, and value to customers.
Open COM platforms for industrial AI at the edge
For industrial automation equipment such as assembly line cameras, autonomous mobile robots, and smart controllers, system-on-modules (SOMs) can provide high-performance AI processing in a small physical footprint. Their compact size and carrier-based mounting simplify installation in existing cabinets and HMIs, making SOMs ideal for retrofits and space-constrained applications. Low thermal design power (TDP) is equally critical in these environments. Many SOMs offer efficient operation with high performance-per-watt (PPW) that supports passive or simplified cooling strategies to reduce maintenance requirements.
As industrial edge AI projects move from prototype to production, OEMs and system developers need platforms that reduce integration effort yet preserve upgrade flexibility. SECO’s SOM-SMARC-QCS6490 and SOM-SMARC-QCS5430 address this by combining Qualcomm Dragonwing heterogeneous computing with SMARC, an open computer-on-module (COM) standard with a narrow power envelope and a 50 x 82 mm footprint. By leveraging open standards, developers can upgrade AI performance with new chipsets throughout a product lifecycle while minimizing carrier redesign effort.
For AI processing at the edge, heterogeneous architecture matters. CPUs manage control logic and orchestration, GPUs accelerate graphics and vision pipelines, and NPUs execute AI inference efficiently at the edge. The compute architectures of the SOM-SMARC-QCS6490 and SOM-SMARC-QCS5430 are compared in Table 1. By reducing cloud dependence for AI processing, devices can provide faster responses while maintaining operational continuity during network interruptions.
| Module | Chipset focus | Heterogeneous compute | Example Industrial Edge AI Workloads |
|---|
| SOM-SMARC-QCS6490 | Higher-performance edge AI | Multi-core CPU, GPU, and embedded NPU (up to 12.15 INT8 TOPS) | Multi-camera vision inspection, operator recognition, video analytics, intelligent HMI, predictive maintenance, and AI-assisted robotics |
| SOM-SMARC-QCS5430 | Cost- and power-optimized edge intelligence | Multi-core CPU, GPU, and embedded NPU (up to 9 INT8 TOPS) | Audio analytics, smart HMIs, anomaly detection, condition monitoring, single-camera automation, and gateway AI |
Although the chipsets differ between SOMs, the SMARC standard defines the interfaces, preserving modular upgrades and easing migration as AI workloads evolve. Key industrial HMI and connectivity options include:
- MIPI CSI camera interfaces, supporting operator recognition, visual inspection, safety-oriented AI vision.
- Display interfaces for building sophisticated multi-screen HMIs.
- Audio interfaces, enabling voice interaction and acoustic anomaly detection.
- Ethernet for direct machine data exchange and edge-to-cloud communications.
SECO also integrates soldered LPDDR5-6400 memory to improve inference speed and operational reliability in vibration-prone industrial environments. For rapid evaluation of these platforms, SECO’s DEV-KIT-SMARC provides a ready-to-use carrier with physical connectors for prototyping on the bench or integrating with existing equipment.
Reducing integration timelines through ready-made HMI platforms
In contrast to SOMs, fully realized HMI platforms remove hardware development from the critical path, allowing developers to focus on developing AI applications, defining workflow logic, and creating user-specific functionality. SECO’s Modular Vision 10.1 QCS6490 and Modular Vision 15.6 QCS6490 platforms address this need by handling display integration, mechanical design, interface validation, and more. Here, the Qualcomm Dragonwing QCS6490 chipset enables both HMI platforms to deliver power-efficient, accelerated computing for real-time vision, AI inference, and demanding edge workloads. By sharing the QCS6490 platform with its SMARC solutions, SECO enables developers to scale edge AI functionality across product lines via common software stacks.
With industrial-grade construction that supports operation in production-floor environments, SECO’s Modular Vision reduces time-to-market for a wide range of AI-driven HMI use cases. Slimline 10.1” and 15.6” panel-mount display options offer multi-touch input and gesture support with full HD resolution and high brightness to support detailed machine control and monitoring workloads. Additionally, the rear panel provides 9 to 32 V power input and robust connectivity options for versatile integration with surrounding equipment, sensors, and peripherals.
Accelerating software development with Clea and SECO App Hub
Hardware is only part of successful edge AI deployment. SECO’s comprehensive Clea software suite simplifies initial development, scaling, and lifecycle management when using SECO or third-party hardware. Clea OS presents developers with a modular, Yocto-based operating system (OS) that supports custom industrial software, offering improved maintainability and security by reducing unnecessary components. At the deployment level, Clea Astarte and Clea Edgehog provide edge-to-cloud data orchestration and fleet management infrastructure to support scalable long-term deployments through device visibility and over-the-air (OTA) updates.
To further accelerate edge AI deployment, SECO App Hub completes the development ecosystem by offering ready-to-deploy, pre-validated AI applications optimized for SECO’s Qualcomm Dragonwing-based platforms. Examples for industrial automation and HMI use cases include:
- PoseNet for multi-person pose detection, enabling worker motion tracking and safety analytics.
- Emergency safety control via audio detection for acoustic event monitoring.
- Face recognition for industrial HMI, providing contextual operator interactions and troubleshooting support.
Conclusion
As industrial edge AI adoption accelerates, developers face growing pressure to reduce time-to-market despite an exponential increase in product complexity. In combination with validated software solutions, SOMs like SECO’s Dragonwing-based SMARC portfolio provide a scalable foundation for long-term success. By combining ready-to-implement, integrated hardware systems with modular upgrade paths and software support, SECO empowers OEMs to accelerate deployment while remaining better prepared for future demands.
Visit seco.com to explore how SECO’s AI-ready, integrated modular solutions can enable rapid, scalable deployments across a wide range of applications.