IoT Devices Could Become 40% More Energy Efficient With This AI-Based Approach

How it’s working: Backscattering allows devices to modulate and reflect incoming wireless signals back to the transmitter, similar to how RFID chips and contactless payment cards work, harvesting energy from the reader. This allows IoT devices to achieve battery-free communication.

A team of researchers from Pusan ​​National University in South Korea has developed a new method based on the same technology that is about 40 percent more energy efficient than currently used techniques.

Backscattering as a concept isn’t new. But the team’s innovation is using AI to optimize the system and make it much more efficient for low-power applications like IoT sensors. Their approach involves using machine learning to precisely model the optimal “reflectivity coefficients,” which determine how much of a wireless signal will be reflected.

Traditionally, calculating these rates has relied on simulations that don’t quite match real-world conditions, making it difficult to achieve low bit error rates and high data rates. But researchers have overcome this by using a technique called “transfer learning,” in which an AI model is first trained on one task and then refined using data from the real target task.

To make this work, they pre-trained an artificial neural network on simulated input voltages to understand the behavior of the modulation circuits under different voltage conditions. They then further trained this pre-trained model using real experimental data, allowing it to accurately predict the reflection coefficients for their specific hardware.

With these fine-tuned AI models, the team was able to optimize their 4-QAM and 16-QAM modulation schemes for maximum efficiency. QAM stands for Quadrature Amplitude Modulation, a scheme widely used in Wi-Fi communications systems. Their prototype system uses less than 0.6 milliwatts during transmission—a fraction of the power required by conventional wireless radios.

The system also includes a 2×2 MIMO antenna to improve signal reception. When tested in the 5.7-5.8 GHz range, it achieved a spectral efficiency of 2 bits/second/hertz using 4-QAM modulation.

“The combination of accurate circuit modeling, advanced modulation techniques, and polarization diversity, tested in wireless environments, provides a holistic approach to solving challenges in the ISC and IoT domains,” said Professor Sangkil Kim, who led the study.

This, say the Busan researchers, provides the basis for reliable, ultra-low-power backscatter systems that could find applications in consumer electronics, healthcare monitoring, smart city infrastructure, environmental measurements and radar communications.

Their research results were published in an article in the IEEE Internet of Things Journal.