In the realm of energy-efficient communication technologies, researchers Talha Akyildiz and Hessam Mahdavifar from Northeastern University have made significant strides. Their work, published in the IEEE Transactions on Wireless Communications, focuses on enhancing ambient backscatter communication (AmBC) systems, which enable battery-free connectivity by harnessing existing radio frequency (RF) signals.
Ambient backscatter communication is a promising technology for the energy sector, particularly for Internet of Things (IoT) applications where devices need to operate with minimal power consumption. These devices, often referred to as “tags,” modulate existing RF signals to communicate without the need for their own power source. However, detecting multiple tags in such systems is challenging due to strong interference from the direct RF signal, extremely weak backscatter signals, and the vast number of possible tag states.
The researchers first established a benchmark by deriving analytical performance bounds for an optimal detection method, the multi-hypothesis likelihood ratio test (LRT), under the assumption of perfect channel state information (CSI). However, obtaining and tracking perfect CSI in AmBC systems is impractical because the RF source is uncooperative and the tags are low-power passive devices.
To address this, Akyildiz and Mahdavifar proposed two deep learning frameworks that relax the CSI requirement while remaining agnostic to the modulation scheme used by the tags. The first framework, EmbedNet, is an end-to-end prototypical network that maps the covariance features of the received signal directly to the states of multiple tags. The second framework, ChanEstNet, is a hybrid scheme where a convolutional neural network estimates the effective channel coefficients from pilot symbols and then passes this information to a conventional LRT for interpretable multi-hypothesis detection.
Simulations conducted over diverse ambient sources and system configurations demonstrated that the proposed methods substantially reduce the bit error rate, closely track the LRT benchmark, and significantly outperform energy detection baselines, especially as the number of tags increases. These advancements could lead to more reliable and efficient communication in energy-harvesting IoT networks, facilitating better monitoring and control in smart grids, industrial automation, and environmental sensing.
The research was published in the IEEE Transactions on Wireless Communications, a prestigious journal in the field of wireless communication technologies.
This article is based on research available at arXiv.