Posts

A new preprint entitled Model-based learning for joint channel estimation and hybrid MIMO precoding just appeared on arXiv, and it has been accepted to SPAWC. In this paper, iterative algorithms are transformed into neural networks to adress jointly the tasks of channel estimation and precoding.

Our paper entitled Model-based learning for multi-antenna multi-frequency location-to-channel mapping has just been accepted to the IEEE Journal of Selected Topics on Signal Processing ( Editor link). It shows that a physically-inspired neural network can faithfully reconstruct channel responses in a given environment with a reasonable number of learnable weights.

A new preprint entitled Benefits of Mutual Coupling in Dynamic Metasurface Antennas for Optimizing Wireless Communications – Theory and Experimental Validation just appeared on arXiv. This is a joint work with Philipp del Hougne and his team, in which a physical model is used to analyze the impact of mutual coupling on both the strength and simplicity of the configuration/radiation pattern mapping.

A few papers have been accepted in the last few weeks: Efficient Frequency Selective Surface Analysis via End-to-End Model-Based Learning has been accepted to EuCAP 2025. Unsupervised Learning for Gain-Phase Impairment Calibration in ISAC Systems has been accepted to IEEE ICASSP 2025.

A new preprint entitled Physically Parameterized Differentiable MUSIC for DoA Estimation with Uncalibrated Arrays just appeared on arXiv. It is a collaboration with Henk Wymeersch’s team at Chalmers University (Baptiste Chatelier works currently there) which shows that antenna impairments can be learnt while using a subspace method for source localization.

A new preprint entitled Efficient Frequency Selective Surface Analysis via End-to-End Model-Based Learning just appeared on arXiv. It shows that model-based learning can be used to predict an FSS response efficiently from its geometric parameters, taking advantage of an equivalent circuit representation.

A new preprint entitled Unsupervised Learning for Gain-Phase Impairment Calibration in ISAC Systems just appeared on arXiv. It is a collaboration with Henk Wymeersch’s team at Chalmers University which shows that gain and phase impairments can be learnt in a unsupervised manner in the context of ISAC transceivers.

Our paper entitled Optimal blind focusing on perturbation-inducing targets in sub-unitary complex media has just been accepted to laser and photonics reviews. This is a joint work with Philipp del Hougne and his team, in which a physical model is used to focus waves onto a non-collaborating target, which amounts to perform conjugate beamforming without any pilot exchange.

A new preprint entitled Model-based learning for multi-antenna multi-frequency location-to-channel mapping just appeared on arXiv. It shows that a physically-inspired neural network can faithfully reconstruct channel responses in a given environment with a reasonable number of learnable weights.

Our preprint entitled Optimizing Multicarrier Multiantenna Systems for LoS Channel Charting has just been accepted to IEEE Transactions on Wireless Communications. Congratulations to Taha for this solid piece of work!