Source-linked AI summary
Massive MIMO is a Reality -- What is Next? Five Promising Research Directions for Antenna Arrays
Emil Björnson, Luca Sanguinetti, Henk Wymeersch, Jakob Hoydis, Thomas L. Marzetta
TL;DR
Massive MIMO has become commercially deployed in sub-6 GHz cellular networks, motivating research beyond its initial communication-focused development. The paper outlines five research directions for exploiting increasingly widespread massive antenna arrays, while identifying associated technical challenges and opportunities for synergy.
Problem
Massive MIMO has moved from a future cellular concept to deployed technology, creating a need to identify new research directions and applications for widespread massive antenna arrays.
Method
The paper surveys five directions involving Massive MIMO, digital beamforming, and antenna arrays, including extremely large apertures, holographic arrays, positioning, radar, and intelligent systems.
Results
95% of Huawei’s current commercial shipments had either 32 or 64 antennas, demonstrating that Massive MIMO had become a reality in conventional sub-6 GHz cellular networks.
Takeaways & Limitations
The five directions may be pursued together, because the paper identifies synergies and opportunities to combine them as Massive MIMO technology becomes mainstream and widespread.
Abstract
from arXiv · showhide
Massive MIMO (multiple-input multiple-output) is no longer a "wild" or "promising" concept for future cellular networks - in 2018 it became a reality. Base stations (BSs) with 64 fully digital transceiver chains were commercially deployed in several countries, the key ingredients of Massive MIMO have made it into the 5G standard, the signal processing methods required to achieve unprecedented spectral efficiency have been developed, and the limitation due to pilot contamination has been resolved. Even the development of fully digital Massive MIMO arrays for mmWave frequencies - once viewed prohibitively complicated and costly - is well underway. In a few years, Massive MIMO with fully digital transceivers will be a mainstream feature at both sub-6 GHz and mmWave frequencies. In this paper, we explain how the first chapter of the Massive MIMO research saga has come to an end, while the story has just begun. The coming wide-scale deployment of BSs with massive antenna arrays opens the door to a brand new world where spatial processing capabilities are omnipresent. In addition to mobile broadband services, the antennas can be used for other communication applications, such as low-power machine-type or ultra-reliable communications, as well as non-communication applications such as radar, sensing and positioning. We outline five new Massive MIMO related research directions: Extremely large aperture arrays, Holographic Massive MIMO, Six-dimensional positioning, Large-scale MIMO radar, and Intelligent Massive MIMO.
I. INTRODUCTION
Antenna arrays enable spatially selective transmission, reception, and sensing, extending beyond conventional directional beamforming to multi-user MIMO. Multi-user MIMO spatially multiplexes users by focusing signals toward intended receivers while reducing interference elsewhere.
- Beamforming: Antenna arrays can dynamically change their radiation patterns over time and frequency for transmission and reception.They support both angular beamforming and spatial filtering at arbitrary points.
- Beamforming: Spatial focusing at an arbitrary point can produce a superposition of angular beams without dominant directivity.Both angular focusing and point focusing are commonly called beamforming.
- Sensing: Antenna arrays can also sense propagation environments to detect anomalies or moving objects.
- Multi-user MIMO: Multi-user MIMO spatially multiplexes several users on the same time-frequency resource using antenna arrays at base stations.Receive combining separates uplink signals, while transmit precoding supports downlink multiplexing.
- Multi-user MIMO: With M antennas, an ideally M times stronger received signal can be achieved without increasing radiated power, reducing interference to other locations.If K users are multiplexed and M ≥K, the received signal can be M/K > 1 times stronger than in a classical system.
- Historical development: Multi-user MIMO did not become commercially successful in the 1990s or 2000s because it was complicated and expensive when voice-service deployment favored users per km2 over spectral efficiency.Early information-theoretic work also assumed perfect CSI, while practical systems have imperfect CSI.
A. Massive MIMO is a Reality
Massive MIMO has moved from a research concept to commercial deployment through large digital arrays, mature signal processing, and practical implementation pathways. Commercial sub-6 GHz products already use 64 fully digital chains, while digital mmWave arrays are progressing toward deployment.
- Technical foundation: Massive MIMO addresses prior multi-user MIMO shortcomings using very large antenna counts, TDD operation, and uplink-downlink channel reciprocity.This communication protocol supports arbitrary antenna numbers and channel conditions.
- Technical foundation: M ≥64 antennas provide high spatial resolution, robustness against small-scale fading, and interference suppression with imperfect CSI when M ≫K.
- Technical foundation: Massive MIMO research has matured across spectral-efficiency analysis, energy-efficient design, pilot contamination and decontamination, and power optimization.Recent textbooks cover both fundamentals and advanced topics.
- Commercial reality: In 2018, commercial sub-6 GHz deployments included products with 64 antennas connected to 64 fully digital transceiver chains.The AIR 6468 and competing Huawei and Nokia product lines demonstrate commercial availability.
- Commercial reality: 95% of Huawei’s current commercial shipments reportedly had either 32 or 64 antennas.
- Deployment path: Commercial networks can begin with simple MR beamforming and least-squares channel estimation, then upgrade signal processing through software as user demand grows.Operators initially prioritize cell-edge performance, including beamforming toward arbitrary spatial points.
- Deployment path: A 24-antenna fully digital mmWave array was experimentally verified in the 28 GHz band, supporting continued development toward commercial digital solutions.Early 5G mmWave products may use analog or hybrid implementations before digital solutions become prevalent.
B. What is Next?
The paper argues that widespread Massive MIMO deployment creates opportunities for new antenna-array applications and deployment concepts. It focuses on distributed, extremely large apertures whose near-field operation and nonstationary channels require new models and methods.
- Future deployments may use arrays with 64 or more antennas across sub-6 GHz and mmWave sites, enabling widespread spatial processing beyond mobile broadband.The paper asks how this spatial resolution can support new communication and non-communication applications.
- Extremely Large Aperture Arrays: Extremely Large Aperture Arrays distribute antennas across large areas or construction elements, rather than concentrating them in one visible, heavy box.One example integrates antennas next to windows in a tall building.
- Extremely Large Aperture Arrays: A 1512-antenna example spans 24 m × 60 m with meter-scale antenna separation, far exceeding the wavelength at considered 5G frequencies.
- Extremely Large Aperture Arrays: An ELAA consists of hundreds of distributed BS antennas that jointly and coherently serve many distributed users.
- Extremely Large Aperture Arrays: ELAA users typically lie in the radiative near-field, allowing the array to resolve propagation distance as well as angle but making channel modeling substantially harder.Some wave components may be visible only to subsets of the array, and conventional channel hardening cannot be expected in the same way.
- Table I presents the first two proposed research directions as umbrellas that collect multiple previously separate research topics as special cases.
A. Vision
The paper envisions ELAAs delivering much higher area throughput than compact Massive MIMO arrays. Distributed antennas, inexpensive hardware, and coordinated operation are central to this vision, while interconnecting thousands of components remains a challenge.
- Vision: ELAAs aim to provide orders-of-magnitude higher wireless area throughput than Massive MIMO with compact arrays.The vision relies on more antennas and distributed deployment to reduce average propagation loss and increase spatial resolution.
- Vision: Cooperation among distributed antennas can overcome the inter-cell-interference barrier associated with ultra-dense networks, at least in theory.The ultimate goal is mutually orthogonal user channels with per-user throughput similar to an additive white Gaussian noise channel without propagation loss.
- Vision: Smartphone-grade hardware and increasingly capable integrated circuits could make ELAA deployment cost-efficient, but interconnecting thousands of cooperating antennas remains challenging.
- Vision: With total output power fixed at 1 W, distributing it across M antennas can ideally produce an M times stronger received signal through signal focusing.The same distributed architecture can improve uplink SNR by collecting more signal energy.
B. Open Problems
The paper presents Holographic Massive MIMO as continuous-aperture communication, while identifying modeling, implementation, and deployment challenges for realizing its potential.
- Distributed extremely large arrays face interference-rejection and front-haul challenges when thousands of antennas must coordinate through centralized processing.
- Continuous-aperture systems pose open problems in channel modeling because near-field wavefronts and arbitrary antenna deployments make conventional array-response models difficult.
- The research direction aims to use antenna surfaces for communication, sensing, and other spatial-processing applications while requiring new physical and communication models.
- Optical holography provides an analogy: recording mixes a received field with a reference wave, and reconstruction illuminates the medium with a reference replica to recover the field.
- Two implementation approaches use either tightly coupled active antennas or nearly passive electronically steerable reflecting elements, including software-controlled metasurfaces.
- Holographic Massive MIMO uses an approximately continuous antenna aperture to actively generate beamformed RF signals or control reflections from signals generated elsewhere.
A. Vision
Holographic Massive MIMO envisions electronically controlled continuous surfaces integrated into everyday structures. Its extreme spatial resolution could support new near-field applications, but asymptotic analysis must respect physical energy constraints.
- A continuous aperture can reduce spatial aliasing associated with sub-critically spaced discrete antennas and may be easier to analyze using integrals.
- Continuous apertures could be integrated into walls, windows, and fabrics to emit and receive electromagnetic waves while providing high spatial resolution and spatial multiplexing.
- Dense apertures enable near-field applications involving evanescent waves, including potential extensions of implantable neural communication systems.
- Existing asymptotic Massive MIMO models can predict physically impossible energy capture, so they cannot establish the true asymptotic performance limit.
- Resonant evanescent-wave coupling has transferred 60 W over two meters with 40 % efficiency at a 10 MHz carrier frequency.
- Super-directivity may provide array gains far beyond conventional arrays and maximum-ratio processing when antennas are placed closely together.
B. Open Problems
The paper identifies multidisciplinary theory, electromagnetic modeling, mutual coupling, radiated-power calculation, and application validation as central open problems for Holographic Massive MIMO.
- Progress requires researchers who understand both communication theory and electromagnetic theory, especially when extending analyses beyond line-of-sight propagation.
- Continuous-aperture transmission has been approximated with discrete patches, introducing mutual coupling that must be addressed in practical implementations.
- Spatially stationary small-scale fading can be modeled through a plane-wave spectral representation derived from the homogeneous wave equation.
- Radiated power is a quadratic form involving current and a positive-definite impedance kernel, rather than a simple integral of squared current magnitude.
- Metasurface range extension has been demonstrated, but its gains over conventional relaying are not convincing, leaving a need for a compelling application.
- The proposed surfaces could form sharp beams without sidelobes and potentially approach Massive MIMO limits for capacity and energy efficiency.
IV. DIRECTION 3: SIX-DIMENSIONAL POSITIONING
Wireless positioning remains constrained by measurement quality and historically coarse accuracy. Massive MIMO adds antennas and bandwidth that can open new dimensions for higher-precision localization.
- Cellular positioning evolved from 2G cell-ID and coarse timing to more accurate time-difference-of-arrival measurements enabled by larger 3G bandwidths.
- Positioning generally collects measurements first and then computes a position estimate, so accuracy is fundamentally limited by the underlying measurements.
- Massive MIMO can provide new positioning opportunities because small arrays offer little benefit, whereas more antennas combined with more bandwidth can improve positioning precision.
A. Vision
Six-dimensional positioning aims to estimate a user's three-dimensional location and orientation using large antenna arrays, especially at high carrier frequencies. Massive MIMO arrays can also exploit angular, delay, and multipath information for sensing, mapping, and synchronization.
- A. Vision: Six-dimensional positioning estimates three-dimensional location plus roll, pitch, and yaw, using many-antenna arrays particularly at high carrier frequencies.
- A. Vision: AoA and AoD estimation provide new physical measurements, with AoA variance decreasing cubically and AoD variance quadratically as antenna count increases.The relevant antenna count is measured along the direction of interest, such as horizontal azimuth or vertical elevation.
- A. Vision: Larger apertures and fully digital arrays can improve positioning accuracy through finer spatial resolution and advanced signal processing.Larger apertures can be implemented at lower frequencies, while higher carrier frequencies can fit more antennas into the same physical size.
- A. Vision: Arrays on both the base station and user device enable simultaneous AoA and AoD estimation, allowing user orientation to be estimated alongside absolute position.
- A. Vision: At mmWave frequencies, sparse geometry-related channels allow communication signals to sense nearby objects through their angles and distances, supporting environmental maps.Multipath can thereby contribute to positioning rather than merely degrade it.
- A. Vision: Propagation paths can provide enough geometric constraints to jointly synchronize a user's clock or oscillator and estimate position, even when the line-of-sight path is blocked.
B. Open Problems
Large-scale MIMO radar extends array-based sensing but still faces unresolved modeling, algorithmic, engineering, and adoption challenges. Its promise comes from combining multiple probing waveforms, many receive antennas, spatial diversity, and high spatial degrees of freedom.
- B. Open Problems: Positioning requires tractable, sufficiently rich channel models and common evaluation use cases with highly accurate ground truth for the base station, user, and environment.
- B. Open Problems: Positioning and communication require jointly designed pilots, beams, waveforms, and computational procedures for single- and multi-base-station settings.
- B. Open Problems: Engineering challenges include synchronization among base stations and users plus calibration of antenna positions and orientations to the application's required accuracy.
- B. Open Problems: A 10 ns timing error can produce positioning errors of 3 m or more, making positioning requirements stricter than those of many communication applications.
- B. Open Problems: Precise Massive MIMO positioning raises privacy, security, and dual-use concerns, particularly for safety-critical applications such as UAVs.
- B. Open Problems: Large-scale MIMO radar uses multiple transmit waveforms and jointly processes signals from multiple receive antennas to detect and estimate target parameters.
- B. Open Problems: Spatially multiplexed probing signals can identify up to M targets, improve spatial resolution and interference rejection, and generate better-performing beampatterns.Here, M denotes the number of transmit antennas.
- B. Open Problems: Robust radar detection and estimation algorithms remain needed, with complexity scalable in antenna count and implementations suitable for centralized or distributed arrays.MIMO radar adoption is also slowed by continuing skepticism and the limited understanding of its advantages and limitations.
A. Vision
Large-scale MIMO radar applies Massive MIMO's large-array paradigm to target detection, parameter estimation, interference rejection, and coexistence with communications. The direction remains exploratory because its benefits, limitations, and robust scalable algorithms are not yet fully established.
- A. Vision: Large arrays can improve radar spatial diversity, spatial resolution, target parameter estimation, detection, and interference rejection through waveform design.
- A. Vision: The direction remains open because its advantages and limitations are not yet clear, and scalable robust detection and estimation algorithms are still needed.
- A. Vision: Preliminary work indicates that large-scale MIMO radar can exploit high spatial degrees of freedom to detect targets without prior knowledge of the received-data statistical model.This addresses radar clutter, whose statistical characterization depends on many environmental factors.
- A. Vision: Large-scale MIMO radar is envisioned as a way to address increasingly complex military and commercial scenarios, including applications involving small UAVs.
- A. Vision: Large antenna arrays in radar and communications may support seamless spectral coexistence in mmWave bands through their high spatial degrees of freedom.
B. Open Problems
The paper identifies open problems spanning large-scale MIMO radar, machine learning, CSI exploitation, and adaptable antenna-array geometry. These directions require scalable algorithms, real-world data and testbeds, suitable hardware, and solutions to practical deployment constraints.
- Large-scale MIMO radar: Large-scale MIMO radar needs robust detection and estimation algorithms whose complexity scales with the number of antennas.The paper also notes that the advantages and limitations of such radars remain insufficiently understood.
- Large-scale MIMO radar: Measurement campaigns and testbeds are needed to develop target and clutter models and validate radar performance under real-world conditions.
- Large-scale MIMO radar: Radar systems with hundreds of antennas require manufacturing economies of scale, while consumer-grade components may introduce hardware distortions.
- Machine learning: Machine learning can improve existing communication algorithms, reduce implementation complexity, or enable use cases that model-based approaches cannot feasibly support.
- Machine learning: Low-cost hardware with impairments, including one-bit ADCs, is a promising ML setting because nonlinear systems are difficult to model analytically.One-bit ADCs could enable ultra-low-cost Massive MIMO systems with dramatically reduced energy consumption.
- CSI exploitation: Storing and analyzing CSI together with side information could improve performance and support applications such as downlink prediction, transmitter identification, activity detection, and positioning.The paper presents ML as the primary tool for exploiting these large CSI datasets.
- Adaptable arrays: Flexible array geometry could dynamically adjust aperture and spatial resolution to user distributions while requiring less hardware than fixed alternatives.The paper also points to software-controlled metasurfaces as a way to adapt the radio environment itself.
A. Vision
The paper envisions Intelligent Massive MIMO in which machine learning becomes broadly integrated into communication systems and exploits stored CSI and RF data. Realizing this vision depends on overcoming data, latency, processing, deployment, and model-management challenges.
- A. Vision: Machine learning is envisioned as a first-class component of future communication systems, used where models are lacking, intractable, or too complex to implement.
- A. Vision: Connected arrays, learned channel estimation, and reinforcement learning can be combined to optimize array configurations and communication processing over time.Figure 9 presents this combination as an example of multiple ML-based use cases.
- A. Vision: The proposed Intelligent Massive MIMO concept covers ML applications across almost every part of future communication systems.
- Open challenges: Successful communication-system ML requires application-specific solutions because each application has distinct requirements and constraints.
- Open challenges: Training data must be cleaned and debiased, but operational-network data are difficult to acquire and open datasets are scarce.Data acquisition is costly, particularly for the physical layer, and ML benefits must outweigh those costs.
- Open challenges: Communication ML must meet micro- to nanosecond inference constraints, forcing processing near baseband hardware with limited computational resources.
- Open challenges: Reinforcement learning is difficult to deploy in real systems because learning is slow and may not be feasible during operation.
- Open challenges: Multiple interconnected ML models create technical debt and data dependencies that complicate maintenance when models or inputs change.
VII. CONCLUSIONS
The paper outlines five promising research directions involving Massive MIMO, digital beamforming, and antenna arrays. It presents them as complementary opportunities for renewed research while acknowledging that the list is not exhaustive.
- VII. CONCLUSIONS: The five directions can be studied independently or combined through synergies and opportunities across Massive MIMO, digital beamforming, and antenna arrays.
- VII. CONCLUSIONS: The directions are expected to experience a renaissance as Massive MIMO becomes mainstream and omnipresent.
- VII. CONCLUSIONS: The paper does not provide an exhaustive list of prospective antenna-array research directions.
- VII. CONCLUSIONS: Possible further paths include combining Massive MIMO with quantum communications, frequencies above 300 GHz, or molecular communications.