The Internet of Things (IoT) and Artificial Intelligence (AI) are transforming industries and society. They make it possible to automate routine activities while unlocking previously unattainable insights and functionality. Edge computing – processing data where it is created, rather than remotely in a data center – offers a greener, smarter solution.
Moving AI to the edge
Sensors have undergone extraordinary proliferation since the beginning of the 21st Century. Thanks to IoT, connected, smart sensors are all around us. In 2030 it is estimated that the billions of sensors on IoT devices could be responsible for 30% of Internet data traffic, thanks to the widespread deployment of 5G. This would significantly increase the carbon impact of AI. IoT and AI applications have traditionally been deployed in the cloud because they require the computing resources of data centers to turn data into insights and action. As the number of applications grows, we need to reduce their dependency on power-hungry cloud computing.
Edge computing provides a solution. It involves moving some portion of the storage and compute resources out of the data center, closer to where the data is being generated. In this way, applications and devices collect and analyze data on their own. This approach is gaining momentum as it offers meaningful innovation for applications as varied as consumer products, building management, predictive maintenance in the industry, autonomy for vehicles, and much more. It lowers energy consumption, better protects personal data, reduces latency, and allows decision autonomy at the point of use for increased control, learning, and intelligence.
Many ways that edge AI can contribute to a smarter and greener industry
There are many ways in which edge-based computing is helping to make different industry sectors smarter and greener. Here are a few examples.
- Condition monitoring and predictive maintenance in factories can make operations smarter and more energy efficient. Sensors provide regular updates on the operating condition of machines to determine when they need servicing or if certain component parts need to be replaced. This reduces downtime and ensures machines run at optimal energy efficiency.
- Edge AI will also be critical for the next generation of Collaborative robots (cobots) designed to operate in real-time in the same workspace as humans to improve efficiency while ensuring their safety.
- Smart cities can use networks of millions of intelligent sensors and IoT nodes to improve monitoring, manage resources, assist citizens, and improve logistics with self-driven drones and vehicles.
- Driving is becoming safer, greener, and more connected. The multitude of sensors distributed over next-generation cars requires local AI so that the vehicle can react quickly to potentially life-threatening situations. In addition, edge AI can be used to optimize the battery management system, as well as adjust the vehicle to the driver’s style, to ensure the vehicle is driven in an energy-efficient manner.
- Bringing automation into the agricultural sector can help increase productivity and lower the environmental impact. Smart farm vehicles and machines will contribute to sustainability strategies by making it possible to use less water, fertilizer, and pesticides. Sensors coupled with edge AI allow the distribution of appropriate amounts of water or chemical substances to individual plants.
- Medical and healthcare are progressing to more personalized, always-on monitoring and the possibility to access medical services from the comfort of the home. Sensors and edge AI play a key role here also in the management of data privacy.
All these examples generate vast volumes of sensor data that would be highly energy and bandwidth-inefficient to send to send to the cloud for processing – as well as generating data protection and latency issues. Edge computing with AI provides a way to make them possible in a sustainable manner.
Fast-tracking edge computing
Moving data processing for AI from the cloud to the edge requires the deployment of a range of semiconductor innovations. These include ultra-low power technologies and system approaches as well as the introduction of specific hardware accelerators in Neural Processing Units (NPUs) and In-Memory Computing (IMC) solutions. These highly-efficient computational hardware technologies supporting AI and machine learning are already transforming at-the-edge data processing and are pushing the envelope on intelligent functionality and IoT network scalability. Combined, they reduce system-level power and bandwidth requirements, while increasing further the computational efficiency of new-generation microcontrollers for edge devices.
STMicroelectronics is focused on optimizing computing efficiency combined with low power consumption and embedded security techniques to protect the gathered data, its processing and the resulting actions. Such developments are required to attain performance for competitiveness and wide customer acceptance.
ST is already strongly involved in the deployment of AI at the edge in applications across the automotive, smart industry, and IoT. STM32 family of 32-bit Arm Cortex microcontrollers includes low-power versions that are well suited for edge-based operation. To make it straightforward for those less experienced in AI, NANOEDGE AI Studio integrates machine learning libraries and automatically finds and configures the best fit for any specific use case. For those working directly with AI models, STM32Cube.AI allows machine learning engineers to import quickly and optimize their AI solutions, exploiting advanced features, like custom layers and deeply quantized neural network solutions. The constructed library can then be loaded onto the microcontroller and it is ready to start handling inference workloads.
ST started addressing AI in sensors with the introduction of the LSM6DSOX with a Machine Learning core in 2019. Progress on empowering sensor AI-related flexibility continues with the introduction of the Intelligent Sensor Processing Unit (ISPU).
The ISPU combines signal processing and AI algorithms onto sensors which enables innovative products to sense, process, and take actions, while substantially saving space and cutting power by up to 80%.
The number of interconnected, miniaturized objects and systems able to sense, process, and take action is growing fast. ST is helping to drive this transformation with sensors, smart embedded actuators, connectivity, security, and embedded AI. These solutions are geared to offer a toolset for industry, automotive, health, agriculture, and many other sectors to continue to innovate while contributing to their decarbonization efforts.
