IoT Development Guide

Upgraded microcontrollers now come with small, essentially self-contained neural networks, but could be costly to implement. TinyML enables organizations to deploy streamlined AI for IoT without purchasing specialized new hardware.

Both TinyML and AI-specialized chips were expected to make great strides this year, although COVID-19’s impact on the supply chain has presented a key caveat.

As embedded engineers seek smarter ways to ingest, analyze and leverage data, experts expect the new generation of smarter hardware to eclipse many existing IoT devices.

The synergies of AI and IoT are now making serious headway in the automotive, industrial and medical sectors, according to Omdia’s Josh Builta.

A new performance benchmark for Tiny-ML inference applications was launched at the start of this year to compare embedded devices with a footprint of 100kB or less on a level playing field.

Emerging chip architectures inspired by the biological brain could soon help developers remove bottlenecks associated with existing edge semiconductor designs.

The new wave of AI chips has been built from the ground up for applications like driving monitoring systems and automated drones, where data processing must happen immediately.

With RISC-V, IoT practitioners have a more scalable and malleable option for edge chip development outside of the proprietary ecosystem of Arm.

Here’s another look at how the RISC-V Foundation has shifted the landscape for IoT computing cores, as software developers seek power efficiency and customization at the edge.

Semi makers big and small are vying to make AI at the edge a reality. But, scoping out the best applications and moving machine learning models successfully to silicon can be daunting.

Embedded devices at the edge of vast global systems remain an art form. Power constraints persist, and latency demands are stringent – measured in milliseconds.

Making AI at the edge a reality is challenging, but the area is increasingly hard for enterprise developers to ignore. Resource-intensive industrial applications are likely to lead the charge for deployment, enabling factories to contextualize data as it flows into multiple sensor endpoints.

Away from edge chips, engineers need faster interconnect speeds to facilitate computational exchanges both within IoT networks and to the cloud. The industry is working hard on 112 Gbps speeds, with 224 Gbps to follow by 2024/5.

The open-source silicon world welcomed a new contender to the field in January 2019. Wave Computing unveiled plans to open source its MIPS instruction set architecture – or ISA – enabling chip makers and developers to access 32- or 64-bit versions without licensing fees.

Containers enable developers to test and run applications with greater speed and agility, without reconfiguring the environment based on dependency variations. Here’s an IoTWT exclusive on the U.S Department of Defense’s campaign to use containers for its mission-critical systems.

AWS Greengrass extends AWS cloud capabilities–including AWS compute, messaging, data caching and sync–to local devices. This enables them to securely share that information, and act on the data they generate, while using the AWS cloud for management, analytics and storage.

Low-code and no-code platforms can accelerate complex development in cases where operational technology personnel have skills that cannot be easily conveyed to data engineers. The latest versions of leading low-code software enable easy integration with major IoT platforms, including IBM Watson and Amazon Web Services.

Docker is an OS-level virtualization platform that can provide advantages when bringing containers to the edge. Here’s a model, using Raspberry Pi for horsepower, to help you learn how to quickly run Docker command line scripts.

Veteran systems architect and consultant Dan Sullivan believes Rust is an emerging programming language whose time has come for IoT development. The open-source script, developed by the creators of Mozilla Firefox, shares syntax similarities to C and C++, but also incorporates functional elements and development abstractions.

This 23-page report sheds light on the shortcomings of established product development processes in key areas. It posits their advantages to be captured by switching to an integrated, cloud-based design platform.

In a landscape where IoT’s high adoption rate is undermined by poorly executed products, this white paper sets out possible root causes of development failures.

A confusing array of solutions for IoT use cases make it challenging to start development on the right foot. This in-depth exclusive guide looks at considerations including open-source, avoiding lock-in to public clouds and the need for scalable solutions, without spiraling licensing costs.

The old way of thinking about enterprise and industrial IoT often meant portraying the synergies as some sort of secret weapon over the competition. But a newer model of open source IoT has forged a mantra built on co-development and open innovation, rather than secrecy.

Some 74% of enterprises plan on using open-source technologies solely or in tandem with proprietary technologies for their IoT projects, and 39% will rely on open source alone.

As noted earlier in this guide, many of the advantages of open-source software could be enlisted by IoT hardware designers thanks to the RISC-V architecture. While RISC-V is changing the game for IoT processors, other hardware vendors like IBM – with its Red Hat acquisition – are getting into the open-source OS game.

Back in 2018, Microsoft made inroads into open source by making its Azure Edge runtime available on an open-source license. The decision was intended to bring forward edge development resources for IoT engineers who may have been deterred from building for such environments.

If you’re building a hybrid cloud today, it’s likelier than not that you are using a proprietary platform, like Azure Arc or AWS Outposts. Yet open source hybrid cloud solutions are quietly holding their own, providing an alternative for organizations wary of committing to a proprietary platform for setting up and managing a hybrid cloud.

While companies have expanded their use of open source, most are not giving back to the foundational ecosystems -- and the security of these platforms continues to falter.

In the robotics sector, social media and advertising giant Facebook has released a platform for building robots with computer vision and natural language processing capabilities. The system – dubbed Droidlet – enables a single intelligent agent to be deployed across diverging robotic hardware systems.

For many IoT developers, time-series data analysis is a new frontier. But, as more data tasks are brought to the edge, many need new skills to exploit data and mix different subtypes in time series dashboards, having traditionally outsourced such tasks to remote locations with more computer processing power.

The need for embedded data capabilities has resulted in data management tools attracting more resources in the IoT planning phase, as illustrated in this feature article.

In a similar vein, this session from the virtual Industrial IoT World 2020 event discusses Boeing’s experience of enlisting data analytics to solve operational challenges in the air transport sector.

Enterprises are eager to use IoT to understand operations, but are constrained by the need to contextualize the monolithic data deluge flowing in from disparate locations. Data analysts will be looking to move the paradigm forward to host domain datasets – particularly from edge locations – in a more consumable way.

A recent data management, storage and disaster recovery event foresaw future technologies that will vanquish today’s data processing and management roadblocks.

Here’s how a Chilean mountain-top observatory managed to collate masses of visualized data of the universe. The system relied on time-series data technology from InfluxDB offered under an open-source license.

The arrival of hybrid cloud architectures that disperse functions between edge and remote systems has opened the market for “unified analytics” from vendors like Hewlett Packard Enterprise. Unified analytics aims to help enterprises make sense of their data regardless of its location.