Learning-Based Receivers for Next-Generation MIMO Wireless Systems: Joint Channel Estimation and Equalization with Efficient Neural Architectures
Join with us the next seminar in the CTTC “Learning-Based Receivers for Next-Generation MIMO Wireless Systems: Joint Channel Estimation and Equalization with Efficient Neural Architectures”.
- Speaker: Dr. Jonathan Aguiar Soares from Universidade Estadual de Campinas
- Date & Time: Wednesday, July 1, 2026 | 10:00 AM CET
- Location: CTTC Auditorium (Building B6)
- Available online
- Duration: Approximately 45 minutes and Q&A
About the Seminar
Next-generation wireless systems are expected to operate under increasingly demanding propagation, mobility, bandwidth, and hardware constraints, making the design of robust and efficient physical-layer receivers a central research challenge. In this context, learning-based signal processing has emerged as a promising approach to complement model-driven communication algorithms, particularly when channel conditions, interference patterns, or hardware impairments are difficult to characterize analytically.
This talk will present ongoing research on neural-network-aided receivers for MIMO wireless communication systems, with emphasis on joint channel estimation and equalization. The presentation will discuss how data-driven and model-aware architectures can be used to improve receiver performance under dynamic channel conditions, while also considering practical constraints such as training data generation, generalization across scenarios, computational complexity, and suitability for real-time implementation. Particular attention will be given to efficient neural architectures for physical-layer inference and to the role of experimental validation in bridging simulation-based studies and deployable wireless prototypes.
The seminar will also connect these topics to broader research directions in AI-native wireless communications, including XL-MIMO, reconfigurable intelligent surfaces, near-field propagation, over-the-air experimentation, digital twins, and GPU/HPC-supported prototyping for 5G/6G systems. The goal is to motivate a discussion on how learning-based receiver design can evolve from offline performance evaluation toward reproducible, experimentally validated, and implementation-aware communication systems.
Jonathan Aguiar Soares, born in Santa Maria, Rio Grande do Sul, Brazil, in February 1991, is a researcher specializing in telecommunications and electrical engineering. He earned his Ph.D. in Electrical Engineering from the State University of Campinas (UNICAMP) in 2024, following a Master’s degree at UNICAMP in 2021 and a Bachelor’s from the Pontifical Catholic University of Rio Grande do Sul (PUCRS) in 2019.
Jonathan’s research focuses on communication systems, applying machine learning techniques to enhance channel estimation, decoding, and signal processing for MIMO and optical networks, with an emphasis on complex-valued neural networks (CVNNs). His work spans wireless systems, optical communication, and 5G/6G networks, earning recognition in academia and industry, particularly for semi-supervised learning methods in massive MIMO systems.
He has authored impactful journal and conference publications, including articles in IEEE Wireless Communications Letters and the Journal of Lightwave Technology. His contributions include developing phase-transmittance radial basis function neural networks and soft-failure localization methods integrating machine learning with software-defined networking. Many of these innovations have practical applications, with some patented and adopted by the industry.
Publications like “Semi-Supervised ML-Based Joint Channel Estimation and Decoding for m-MIMO With Gaussian Inference Learning” and “Deep Phase-Transmittance RBF Neural Network for Beamforming With Multiple Users” demonstrate his ability to link theory with practice. He has presented on subcarrier-level processing, parameter selection, and CVNN complexities at leading conferences, such as ICMLCN and LATINCOM.
Jonathan’s work has resulted in patents for boosting data transmission rates in optical and wireless systems, as well as neural network-based channel estimation and decoding. With over 180 citations and an h-index of 7 on Google Scholar, his research is becoming a reference in CVNNs for communications.
Currently, Jonathan is developing hardware-accelerated solutions for wireless communications using CVNNs on FPGA platforms, enabling real-time training and inference for 5G and beyond. His focus includes scalable MIMO and beamforming systems and dynamic adaptation to channel conditions.
His research aims to advance AI-driven physical layers for 6G, FPGA-based adaptive systems, and cell-free massive MIMO, integrating wireless and optical networks to enhance scalability and reliability.