How MCUs will advance endpoint intelligence over the next decade

Article By : Carmelo Sansone, Renesas Electronics

Here are seven predictions of more sophisticated MCU applications in the next decade.

Microcontrollers (MCUs) are increasingly advancing endpoint intelligence. So what can we expect from the MCU over next decade? We identify seven key predictions of how these chips might transform our lives even further with increasingly sophisticated applications.

The microprocessor, at the heart of every computer and smartphone, always has been the smarter, faster cousin of the hardworking, low-power microcontroller in embedded systems.

The unsung microcontroller, however, is now gaining new respect as designers take advantage of the system-on-chip component’s ability to employ machine-learning (ML) algorithms at the “edge,” meaning locally, close to the data source rather than “in the cloud.”

The microcontroller has recently excelled at such chores as predictive maintenance for industrial equipment, and voice processing in consumer products. Using microcontrollers at the edge produces quick results and eliminates the need for systems to send data to a power-hungry data center for processing. Nor do such systems require specialized application processors, such as those in autonomous cars for navigation or in drones for surveillance. That’s because today’s microcontrollers are sophisticated enough to run ML models on a small power budget.

The current advantages of microcontrollers are only the beginning of a new trend toward even lower-cost, lower-power MCUs with artificial intelligence and machine learning built into sensors – such as a smart accelerometer that can differentiate between, say, the sound of a glass window breaking and a crying infant.

Microcontrollers at the edge reduce energy use and eliminate latency from the need to send data to a power-hungry data center for processing. In fact, today’s sophisticated microcontrollers can sometimes replace more costly specialized application processors in physical devices.

Unlike a traditional computer learning model, an ML model is trained on a specific set of data for a particular situation. Once those data are captured, the model running inside an ML-enabled controller in an edge device can – in real time – recognize the input, analyze it, and decide on a result.

MCU application trends in the next decade: our predictions

Here are seven predictions for ever-more sophisticated MCU applications we might see during the next decade. The applications and models described are just a start, but they point to the potential for MCUs to improve our lives:

Transportation: embedded MCUs in cameras may gather visual data for analyzing traffic patterns to optimize public transit routes, and still others to capture data points in support of autonomous vehicle navigation. Large numbers of smart MCUs could be deployed for monitoring a city’s infrastructure and traffic conditions in real time. Such data could be used to predict the likelihood of an accident and alert relevant parties before it happens. They also could be used to identify the optimal traffic patterns based on historical data, then send specific vehicles on the most efficient routes to their destinations while avoiding congestion and other road hazards.

Logistics: MCUs in shipping containers could learn to predict when the next shipment will arrive at its final destination based on historical data about typical transit times for that port; if it detects an anomaly in the data stream, it will flag an alert to notify trucking firms, distributors, wholesalers and retailers.

Healthcare: MCUs may be used in the human body to create a network that can warn of potential health problems. They could even administer medications.

Building automation: smart construction materials in buildings could monitor structural integrity thanks to embedded MCUs, aiding human inspectors conducting routine reviews, or emergency workers after a disaster such as an earthquake. Monitoring could be accomplished using embedded microcontrollers that gather data such as GPS location, accelerometer data, temperature and humidity, bending in metal objects, ultraviolet light intensity to detect cracks in concrete walls, or pressure sensors to detect leaks in pipelines around a building.

Smarter homes: microcontrollers, already common in home-appliance and home-automation applications, may no longer need to send and receive commands from a central control or data center. Instead, they could generate commands themselves after gathering data from sensors in connected devices.

Infrastructure: microcontrollers will be used to monitor power lines and other utility infrastructure to detect anomalies before they result in catastrophic damage. Examples could be automated inspection after major storms such as the recent Hurricane Ida, or alerts about sparks from power transformers that could ignite wildfires in drought-stricken locations.

Industrial: microcontrollers could be used to optimize factory operational processes. For example, MCUs could monitor the performance of industrial equipment in real time and adjust the operation parameters of that equipment to improve product quality and reduce waste.

Over the next several years, engineers will design systems and structures that can accommodate such new capabilities. The improvements to MCU hardware enabling these trends are exciting but would not be practical without security that protects the entire data flow to preserve sensitive information. Robust safety and security technologies are imperative to protect mission-critical deployments from core to endpoints.

Of course, technologies like the internet of things (IoT) already employ microcontrollers in edge processing, transforming the innovation landscape. The ongoing challenge for semiconductor companies is to make MCU chips capable of processing even more sophisticated ML algorithms, while continuing to shrink hardware size and power consumption. This is the current trend, and yet we are only the beginning.

By being flexible, small, inexpensive and integrated, MCUs will continue to advance endpoint intelligence for years to come.

This article was originally published on Embedded.

Carmelo Sansone is director of strategic business development at Renesas Electronics in Milpitas, California.


Leave a comment