IMDEA Networks work secures the Best Paper Award at ACM MobiHoc 2023

Research carried out in collaboration with Princeton University and University of Brescia

30 October 2023

In an exciting development that promises to reshape the landscape of massive IoT networks, a team of researchers from IMDEA Networks, Princeton University, and the University of Brescia, received the prestigious Best Paper Award at the ACM MobiHoc 2023 conference, held in Washington from October 23-26.

The award-winning paper, titled “Scalable Multi-Modal Learning for Cross-Link Channel Prediction in Massive IoT Networks,” authored by Kun Woo Cho (Princeton University), Marco Cominelli (University of Brescia/CNIT), Francesco Gringoli (University of Brescia/CNIT), Joerg Widmer (IMDEA Networks), and Kyle Jamieson (Princeton University), introduces a groundbreaking solution to address the burgeoning demand for efficient uplink traffic in the realm of massive-scale IoT sensor networks.

These networks, vital for the realization of the Internet of Things, are expanding rapidly, especially in densely deployed areas. However, accommodating this increasing demand for network resources has proven challenging due to the need for accurate Channel State Information (CSI). Obtaining this information typically incurs significant overhead, leading to a reduction in network throughput. Moreover, as the number of clients grows, the overhead scales accordingly, posing a unique challenge in the context of massive IoT sensor networks.

The traditional approach of predicting the channel response within the same frequency band on a given link has limitations. This award-winning paper introduces a game-changing concept: Cross-Link Channel Prediction (CLCP). CLCP leverages multi-view representation learning to forecast the channel response of numerous users across distinct, nearby links, without relying on dedicated channel sounding or additional pilot signals. Instead, it harnesses existing transmissions, enhancing efficiency while reducing overhead.

The researchers have successfully implemented CLCP for two different Wi-Fi versions, namely 802.11n and 802.11ax, with the latter being the leading candidate for future IoT networks. To demonstrate the viability of CLCP, the team conducted extensive evaluations in two large-scale indoor scenarios, encompassing both line-of-sight and non-line-of-sight transmissions, involving up to 144 different 802.11ax users and four channel bandwidths ranging from 20 MHz to 160 MHz.

The results are nothing short of astounding. CLCP outperforms baseline solutions, delivering a remarkable 2× throughput gain over baseline WiFi standards, and even surpasses existing prediction-based scheduling algorithms with a significant 30% throughput increase. This paper not only showcases the promise of CLCP in addressing the challenges of IoT network scalability but also paves the way for a new era of unparalleled efficiency in the world of ultra-dense IoT. As IoT continues to transform our world, this revolutionary research is poised to play a pivotal role in shaping its future.