How can artificial intelligence fundamentally change the Internet of Things by enhancing battery performance?

Can artificial intelligence help alleviate the dilemma of IoT sensors? The Internet of Things (IoT) is slowly changing the way we collect data and live. It also enables daily necessities to share wireless connections with other devices within the organization. This technology is breaking innovation and concepts in all industrial fields. According to Analytics Insight, the compound annual growth rate of the Internet of Things in terms of the market is 10.6%. It is expected to expand from $653.6 billion in 2019 to $1080.4 billion in 2023.

IoT sensors accumulate a large amount of data through device connections via the Internet. Use artificial intelligence and machine learning tools to analyze and evaluate this data. The collected sensor data can be location, sound or humidity, as well as different measurement values of the machine. Once data insights are obtained, they can be used for comparison, calculation, prediction, and verification with existing data, and corresponding follow-up actions can be taken. Artificial intelligence also helps store large amounts of data processed by IoT devices. In addition, artificial intelligence and the Internet of Things can jointly promote the development of interconnected intelligent machines, which share information with each other and make wise decisions without any human intervention.

IoT devices and sensors are becoming increasingly common. They also help businesses create continuous data streams from terminal devices, while using less energy, waste, and budget, thereby minimizing the company's reliance on data scientists and analysts. However, IoT sensors and devices have resources available as long as they operate under power. If there is no fixed power supply, they cannot collect and transmit data.

To address this issue, researchers from the University of Pittsburgh project propose applying artificial intelligence to extend the lifespan of sensors deployed in the Internet of Things. This system will help reduce the energy consumption of IoT sensors and alleviate battery life issues. In this project, researchers use mounted sensors that are driven by energy obtained from the environment to trigger the main sensors. The backpack mounted sensor will operate unattended and be trained to use artificial intelligence algorithms to only send an on signal to the main equipment when specific event conditions are met.

The chief researcher of the study, Hu Jingtong, Associate Professor of Electronic and Computer Engineering at the Swanson School of Engineering at the University of Pittsburgh, explained that one of the main challenges in running artificial intelligence algorithms using the energy obtained from the environment is that the energy generated by the environment is intermittent. He added, "Just like a laptop, if the sensor loses power, you will lose data, so we hope to help artificial intelligence algorithms make accurate decisions, even in intermittent power outages."

Professor Hu and his team plan to develop a method to save electricity from remote sensing equipment by utilizing energy harvesting technologies such as solar, thermal, or wind energy in the environment. Then, they will add a second small sensor that can trigger a more robust device, saving energy and allowing users to replace batteries more frequently. Smaller sensors should be powered by energy obtained from the environment. This smaller sensor will operate unattended, and with the help of artificial intelligence, it can be trained to recognize patterns and send signals to larger devices to turn on during specific events; Thus serving as a watchdog.

According to a blog article on the National Science Foundation (NSF) website, Hu and his team outlined three tasks that will lay the foundation for intermittent incremental inference in IoT devices based on energy harvesting technology. These are:

1. Develop novel power tracking sensing compression, online pruning, and adaptive algorithms to ensure efficient deployment of multi outlet DNNs on intermittently powered devices.

2. Develop a new Multi Exit Statistical and Incremental Neural Network (MESI-NN) to further reduce waiting time and improve accuracy and energy efficiency.

3. Design a new neural architecture search algorithm that can automatically search for the best MESI-NN architecture. This project will be evaluated using real systems and applications, such as image classification, keyword recognition, and activity recognition.

The main objective of this proposal is to extend the lifespan of sensors and devices deployed in remote areas, which will greatly benefit various consumer, commercial, scientific, and national security applications. IoT devices are also used for monitoring and predicting natural disasters. For example, sensor technology is currently used to observe the gases released by active volcanoes in some remote areas of the Earth.

Prev: UCSD researchers use AI technology to discover new DNA promoter codes