Arm Cortex-M52 to bring AI to low-cost IoT devices

A small processor with enhanced Artificial Intelligence (AI) capabilities for low-cost Internet of Things (IoT) devices is the latest addition to the Cortex-Mda Arm portfolio.

According to Arm, the new Cortex-M52 meets the growing demand from partners for more Machine Learning (ML) capabilities and simpler development flows so that they can innovate and scale up quickly.

The Arm Cortex-M52 is designed for Internet of Things applications with Artificial Intelligence (known as AIoT), i.e. those that require better performance in digital signal processing (DSP) and ML without the additional costs of dedicated accelerators. With integrated Arm Helium technology, the Cortex-M52 can provide a significant increase in performance in these applications running on devices with limited power sources, making it possible to deploy ML inference algorithms.

Arm Helium technology has already been successfully implemented in products used in network endpoints, but the Cortex-M52 now allows Arm’s partners to bring this capability to low-cost, power-constrained IoT devices. According to Arm, the Cortex-M52 is the smallest and most cost-effective processor with integrated Arm Helium technology.

Arm claims that the Cortex-M52 will pave the way for a wide range of AIoT applications that can provide richer and faster experiences with enhanced voice and vision systems, for example in the areas of industrial control, predictive maintenance and wearable sensor fusion. Arm’s blog mentions an example of the use of AI with the main stages of a predictive maintenance algorithm.

According to Arm, the Cortex-M52 can offer performance gains compared to previous generations of Cortex-M of up to 5.6x for ML and up to 2.7x for DSP.

In addition, the Cortex-M52 offers the flexibility to adopt different configurations and performance levels, providing more resources without having to adopt a separate processing unit. This means cost and surface savings associated with the processor.

Previously, to achieve the ML and DSP performance that the Cortex-M52 offers today, it would have been necessary to combine a CPU, a DSP and an NPU, which meant that, in addition to more hardware space, developers would need to write, debug and tune code for chips using separate sets of tools (compilers, debuggers, etc.). Now, with Cortex-M52, AI capabilities are being brought under a single architecture and in a single set of tools. This guarantees a unified development flow for traditional DSP and ML workloads.

The Cortex-M52 is the latest member of the Armv8.1-M Cortex-M family of low-power processors, alongside the more powerful Cortex-M55 and Cortex-M85 models.

AIoT market and the Brazilian version

According to Fortune Business Insights, the global Artificial Intelligence market for Internet of Things (AIoT) devices is expected to be valued at US$ 35.6 billion in 2023 and reach US$ 253.8 billion in 2030, which represents a compound annual growth rate of 32.4 per cent over the forecast period.

According to the study, accelerated innovations in the health sector during the Covid-19 pandemic helped drive the growth of the AIoT market. In the coming years, industrial automation and robotics are expected to increase demand for AIoT services.

In addition, IoT is expected to play an important role in transforming digital homes and creating sustainable communities and smart cities, contributing to the management of energy and water systems. However, the study warns that the growth of the AIoT market may face critical challenges related to safeguarding systems against cyber threats and ensuring the privacy of user data.

In this market, there is already a supplier, a Brazilian version. CPQD, an institution set up in 1976 as a Telebrás R&D centre to support public policies in the area of telecommunications and which is now a private law foundation focused on innovation in the area of ICTs, launched an AIoT platform last September, called Pailot, with an initial focus on companies that are looking to digitalisation for greater operational efficiency and new levels towards Industry 4.0.

“The combination of IoT and AI brings important resources that favour the convergence and integration of industries’ Information Technology and Operational Technology areas,” says Rafael Pacheco, CPQD’s product manager. “With the Pailot platform, the data collected by the IoT infrastructure on the shop floor, for example, is transformed into valuable information for both the operational area and the IT department, guaranteeing intelligence and the possibility of applications such as anomaly detection, pattern recognition, prediction and recommendation models and even decisions for autonomous intervention,” he adds.

The new Pailot platform collects and organises data, creates business rules and integration pipelines (low code), manages devices and access and presents personalised dashboards. It offers analysis tools with machine learning capabilities, as well as a Data Lake capable of storing images, audio and streaming, and open APIs for integration with other types of software and applications.