Research
Physically consistent information theory: bringing back circuits and electromagnetics
MIMO communication systems support more antennas, with wider bandwidths, higher frequencies, and more complicated array configurations. The classical theory used to analyze such systems makes several simplifying assumptions that may not hold in these more extreme regions of operation. The WSIL and external collaborators are progressing in developing information and communication theory based on physically consistent models. This results in research that leverages circuits and electromagnetics to introduce new approaches for analyzing systems, devising performance bounds, and designing algorithms.
Tri-Hybrid MIMO architecture
The next evolution in the hybrid MIMO architecture is the tri-hybrid architecture where beamforming is done using a combination of digital, analog, and reconfigurable antennas. Incorporating reconfigurable antennas makes it possible to scale up the aperture in size while reducing the number of components and the overall power consumption. Beyond enabling large apertures, reconfigurable antennas can dynamically alter many of their primary characteristics, including polarization, gain pattern, and frequency to retune based on spectral and spatial considerations. For example, it would allow for optimizing the gain for the target bandwidth, changing the operating frequency without switching to a different array, and matching the polarization and direction of the signal to better paths in the channel. We are currently developing precoding and combining algorithms for the tri-hybrid architecture and studying how the architecture adapts to different types of antennas.
Reconfigurable antennas, dynamic metasurface antennas, and parasitic arrays
Reconfigurable antennas are an important new ingredient of MIMO communication systems. The WSIL is doing fundamental research to understand reconfigurable antennas in terms of communication theoretic performance measurements. Our current work involves devising new models for reconfigurable antennas, simulating and prototyping the antennas, and developing algorithms to integrate the antennas into MIMO systems as part of the tri-hybrid MIMO architecture.
[2] J. Carlson, M. R. Castellanos, and R. W. Heath, “Hierarchical codebook design with dynamic metasurface antennas for energy-efficient arrays,” IEEE Transactions on Wireless Communications, vol. 23, no. 10, pp. 14790–14804, Oct. 2024, doi: 10.1109/TWC.2024.3419107.
Communication at Upper Mid-band frequencies
The upper mid-band is the next spectrum gold rush. It comprises frequencies from 6 GHz to 24 GHz and is known as FR3 in the cellular world. There are several new challenges for MIMO systems in the upper mid-band. (1) The arrays are larger due to shrinking antenna sizes and there is a desire to retain the same coverage at lower frequencies. (2) The spectrum is occupied by many incumbents such as satellites, space-based sensing, and radars. The WSIL is exploring methods to address these new challenges. One line of research is to understand when and how near-field communication can be useful for multiuser MIMO in the upper mid-band frequencies. Another line of research is developing MIMO techniques to reduce the interference inflicted on incumbent spectrum users by exploiting information about their characteristics.
AI/ML for configuring MIMO communication links
Next-generation cellular and WiFi networks continue to increase the MIMO communication dimensions. Configuring these extreme MIMO links with low overhead remains a challenge. The WSIL has been an early pioneer of AI/ML algorithms for MIMO communication systems since 2009. Our early work was focused on smart rate adaptation in MIMO communication links, where the problem of modulation, coding, and number of MIMO streams was formulated as a classification problem. Our recent work focuses on codebook design for beam-based MIMO including the SSB and the CSI-RS codebooks. We have found that AI/ML can design new beam codebooks with several decibels of performance improvements while staying completely within the constraints of current 5G systems.
[2] R. M. Dreifuerst and R. W. Heath, “Hierarchical ML codebook design for extreme MIMO beam management,” IEEE Transactions on Machine Learning in Communications and Networking, vol. 2, pp. 688–702, 2024, doi: 10.1109/TMLCN.2024.3402178.
[3] A. Rico-Alvarino and R. W. Heath, “Learning-Based Adaptive Transmission for Limited Feedback Multiuser MIMO-OFDM,” IEEE Transactions on Wireless Communications, vol. 13, no. 7, pp. 3806–3820, Jul. 2014, doi: 10.1109/TWC.2014.2314104.
[4] R. C. Daniels, C. M. Caramanis, and R. W. Heath, “Adaptation in Convolutionally Coded MIMO-OFDM Wireless Systems Through Supervised Learning and SNR Ordering,” IEEE Transactions on Vehicular Technology, vol. 59, no. 1, pp. 114–126, Jan. 2010, doi: 10.1109/TVT.2009.2029693.
Position and sensor-aided MIMO communication leveraging AI/ML
Wireless communication systems are increasingly integrating sensor data to support applications such as automated vehicles and robotics. This data may be generated by sensors like radar, camera, and lidar located on the infrastructure and devices, or by integrated sensing and communications (ISAC) where the communication waveform is also used for sensing. This sensor data is collected in the same physical environment where radio propagation is occurring. As such, there are potential congruences between sensing and communication that can be exploited. The WSIL is developing strategies for reducing the overheads in configuring MIMO communication links by exploiting location and sensing information. We continue to develop AI/ML strategies for uncovering the relationships between sensor data and communication strategies.
[2] I. Kilinc, R. M. Dreifuerst, J. Kim, and R. W. Heath, “Beam training in mmWave vehicular systems: Machine learning for decoupling beam selection,” in 2024 IEEE International Black Sea Conference on Communications and Networking (BlackSeaCom), Jun. 2024, pp. 54–59. doi: 10.1109/BlackSeaCom61746.2024.10646235.
[3] Y. Wang, A. Klautau, M. Ribero, A. C. K. Soong, and R. W. Heath, “MmWave Vehicular Beam Selection With Situational Awareness Using Machine Learning,” IEEE Access, vol. 7, pp. 87479–87493, 2019, doi: 10.1109/ACCESS.2019.2922064.
[4] V. Va, J. Choi, T. Shimizu, G. Bansal, and R. W. Heath, “Inverse Multipath Fingerprinting for Millimeter Wave V2I Beam Alignment,” IEEE Transactions on Vehicular Technology, vol. 67, no. 5, pp. 4042–4058, May 2018, doi: 10.1109/TVT.2017.2787627.
Reinforcement learning fundamentals and applications to wireless communications
Reinforcement learning (RL) is a powerful approach for distributed decision-making with broad applications to wireless communication systems. It is well suited for problems in distributed networks including device-to-device communication, relays, and mobile ad hoc networks. The WSIL is investigating how reinforcement learning can solve challenging problems in MIMO link configuration. We are exploring the applications of more advanced regret formulations and are interested in devising new RL methods based on quantized feedback.
Reinforcement learning (RL) is a type of machine learning where an agent learns to make decisions by interacting with an environment.
[2] C. V. Nahum et al., “Intent-aware radio resource scheduling in a RAN slicing scenario using reinforcement learning,” IEEE Transactions on Wireless Communications, vol. 23, no. 3, pp. 2253–2267, Mar. 2024, doi: 10.1109/TWC.2023.3297014.
[3] Y. Zhang and R. W. Heath, “Reinforcement learning-based joint user scheduling and link configuration in millimeter-wave networks,” IEEE Transactions on Wireless Communications, vol. 22, no. 5, pp. 3038–3054, May 2023, doi: 10.1109/TWC.2022.3215922.
[4] V. H. L. Lopes et al., “Deep reinforcement learning-based scheduling for multiband massive MIMO,” IEEE Access, vol. 10, pp. 125509–125525, 2022, doi: 10.1109/ACCESS.2022.3224808.
Radio-over-fiber for distributed MIMO
Distributed antennas add value to MIMO cellular systems. Coordinating multiple transmission points effectively converts a cellular system into one large multiuser MIMO cell. This is the key idea behind cell-free massive MIMO and C-RAN. In the WSIL, we envision an architecture where the large arrays at different cell sites are jointly processed in a central processing system. We are investigating radio-over-fiber techniques as an alternative to eCPRI where signal processing functions and complexity are pushed back to the base station. We are investigating methods for pre- and post-equalizing distortions introduced by the optical link. We are also exploring optical computing to dramatically reduce the power consumption of MIMO beamforming and combining.