Changes Brought by Edge AI to Smart Retail and Factories

Abstract

Edge AI is rapidly emerging in industrial application fields such as smart retail, smart factories/manufacturing, and smart cities/transportation. It provides smarter business insights to achieve higher productivity, lower operational costs, and a better understanding of human behavior. For example, retailers are deploying edge AI-based systems to better understand customer behavior, employee behavior, manage in-store shelves, and develop virtual CRM (Customer Relationship Management) to increase sales, customer loyalty, and work efficiency. All of this is based on fully utilizing in-store IP cameras or AI based on edge servers. Video IP cameras are not only used for theft/incident management but are increasingly used for business insights. Similarly, smart factories are also leveraging edge intelligence for optical inspection of goods, understanding the status of machines and workflows, thus providing predictive maintenance and improving production efficiency. All these are cases of edge AI applications that not only help make low-latency, real-time decisions but also protect the privacy and security of enterprise data.

Edge inference requires developing and running deep learning networks for multiple concurrent use cases, which places high demands on processing power, requiring very large memory bandwidth, low latency, and low power consumption. Typical edge solutions have high power consumption requirements, and the challenge is to meet processing bandwidth and latency requirements without increasing power consumption and costs. Edge AI also needs to collect large amounts of data and store it locally for AI inference. Another key factor in using AI is device security, including firmware, AI models, and data. Industrial IoT devices can be used in highly distributed environments without enterprise-level IT infrastructure. This requires establishing trust in devices in real-time and continuously to ensure the integrity of the data and insights collected by these devices.

Micron is a leader in the industrial market, having dedicated 25 years to developing solutions for industrial customers. For edge AI use cases, Micron offers a range of industrial-grade, high-performance, low-power memory such as GDDR6, LPDDR4, and DDR4. Micron also provides industrial-grade DDR4 modules that can withstand extreme temperatures, specifically designed for deployment in edge servers in harsh environments such as oil and gas, transportation, and energy.

● Micron has developed industry-leading industrial-grade high-density storage solutions: the i300 series 128GB – 1TB microSD cards designed for edge storage for IP cameras; the 2100AI series 64GB – 1TB BGA and m.2 NVMe PCIe high-performance solid-state drives. These products are equipped with industry-standard data protection features.

● To address cybersecurity challenges, Micron Technology has developed the Authenta™ flash memory series, which embeds security elements into the flash memory to simplify the implementation of platform hardening. The trust root embedded in Authenta flash can be used for various security-related use cases, such as secure supply chains, remote firmware updates, and establishing secure communication and trust in devices and data. This functionality is supported by Micron® Authenta Key Management Services (KMS), which can activate and deliver many security services required for smart edge devices based on the cloud.