Northeastern University researchers have embedded intelligence capabilities in a wireless sensor tag that has the potential to revolutionize the Internet of Things and AI.
Your home’s thermostat relies on a sensor to determine when to switch the heat or air conditioning on or off. These wireless sensors are at the core of the so-called “internet of things,” enabling smart devices to collect and share data.
But most sensors still depend on toxic lithium batteries for power. So far, they haven’t been able to make accurate decisions on their own, especially when processing complex information like identifying multiple threats before triggering an alarm.
Borrowing from condensed matter physics, researchers at Northeastern University have embedded intelligence capabilities in a wireless sensor tag, a development that has the potential to revolutionize the Internet of Things and artificial intelligence.
“This is very promising technology because the sensors can be manufactured very easily, they do not constitute a burden to the environment and they do not require any periodic maintenance,” says Cristian Cassella, associate professor of electrical and computer engineering at Northeastern and a co-author of the research.
The new research was published in the journal Nature Electronics.
Most wireless sensors rely on energy from nearby radio waves or light, which can be inconsistent, Cassella says. They also lack the ability to process the signals they detect or perform computations before sending data back to the reader.
The passive wireless sensor tag Cassella helped develop can perform real-time computations on multiple parameters based on its immediate environment. This innovation allows for smarter, more efficient decision-making in networks of wireless sensors, reducing the need to rely on limited cloud resources, Cassella says.
Using the Ising model — a concept developed in physics and recently exploited in quantum computing — the researchers have developed a passive wireless sensor that can make decisions the way the human brain does. Called SPIN (Sensing Parametric Ising Node), the component makes more accurate decisions because it is capable of responding to multiple data sources simultaneously.
“The innovation is in the fact that SPIN can sense more reliably and make better decisions,” Cassella says. “It can do things that no other passive wireless sensor could do before, which could lead to the reduction of gas emissions and energy consumption in buildings and power plants and to reductions of wastage along the cold-chains.”
About 96 billion sensors will be required to operate internet-connected devices by the end of 2025, Cassella says. Embedding intelligence in wireless passive sensors will allow AI and machine learning algorithms to process unprecedented volumes of wireless sensing data, he says.
Researchers built a prototype that accurately detects changes in temperature, but they envision future prototypes to detect humidity, light and the structural integrity of buildings and bridges. In the future, the sensor tags might be able to identify the presence of humans or more complex patterns in the environment like harmful chemicals.
“Each single passive wireless sensor ‘computes and makes decisions’ based on locally sensed parameters,” Cassella says. “The central node will not need deep use of cloud resources and will have a much clearer and more accurate view of what is going on.”