The AI Revolution of Walls: How Can Intelligent Metasurfaces Enable Buildings to “Understand” 6G Signals?
In the current situation where the power consumption of 5G base stations is extremely high, indoor signal attenuation has become a bottleneck in communication. Explore how 6G can leverage intelligent metasurfaces and the wireless friendliness of buildings to break down the barriers between the construction and communication industries, and revolutionize the optimization of 96% of indoor traffic demand.
The extremely high power consumption of 5G base stations has long been criticized in the industry. The main reason is that base stations can only be deployed outdoors, while 96% of the traffic demand comes from indoor users.
The signal strength drops by 90% - 99.9% when passing through walls.
How to serve indoor users, who dominate the traffic demand, is a problem that 6G must consider.
Looking back from the development of the first - generation mobile communication technology to today's 5G, the communication industry has been focusing on overcoming the constraints of buildings and indoor environments. From optimizing coding techniques to coordinating power and spectrum resources, the real - world performance gains brought by the rapidly expanding baseband indicators are becoming fewer and fewer. The higher frequency bands and ultra - large - scale antenna arrays driven by near - field communication technology, as well as the new paradigm of endogenous intelligence represented by technologies such as semantic communication and intelligent metasurfaces, will have to meet the diverse network requirements of various industries and application scenarios.
What lies ahead for the communication industry is the gap between complex technologies and fragmented demands.
The "Butterfly Effect" of Architectural Design
Minor changes in architectural design can bring about huge changes to signal islands.
Intelligent metasurfaces are considered a key innovation to support 6G mobile communication, marking a milestone in actively intervening in electromagnetic wave propagation to optimize communication networks. Before the invention of intelligent metasurfaces, the industry could only passively adapt to the wireless propagation obstacles inside buildings.
Outdoor metasurfaces are generally covered on the outer surface of buildings. The beams emitted from the base station to the metasurface are reshaped and focused and reflected to the shadow area of the building beyond the line - of - sight, thereby expanding the coverage area. However, nearly 96% of mobile traffic occurs indoors, an area that even the most intelligent outdoor metasurfaces cannot reach.
The performance limit of indoor wireless networks is restricted by building materials and structures, and architectural design has never been a friend to wireless performance. The preliminary research of the team led by Zhang Jiliang from Northeastern University shows ([1]) that slight deviations in the relative permittivity and thickness of wall building materials can lead to a loss of more than 14.4% in communication quality.
If the inertial thinking of passively adapting to the building's wireless environment is not changed, the benefits that the next - generation mobile communication technology can bring to users will be very limited.
As shown in Figure 1, if designers can fine - tune the wall material and thickness based on safety standards and energy - saving requirements, the transmission power can be reduced by 10 times while still maintaining the network speed at a 25dB signal - to - noise ratio ([2]).
The relative permittivity and thickness of building materials together define the "Building Wireless Friendliness" indicator ([1]).
The wireless performance of a building is an inherent property of the building itself. Every stroke that an architectural designer makes on the blueprint outlines the data network landscape that accounts for 96% of the total traffic.
The barriers from 5G to 6G lie not only in the information industry but also in the construction industry.
Figure 1: Wireless environment friendliness of buildings
Building Wireless Friendliness
The theoretical basis for breaking down the barriers between the construction and communication industries, "Building Wireless Friendliness", was formally proposed and put into practice in 2022.
The emergence of prefabricated building technology and additive manufacturing technology for building materials provides flexible and low - cost processing methods for wireless performance to break through the constraints of buildings. Prefabricated buildings can shorten the construction period by 80%, and concrete 3D printing can already achieve millimeter - level processing.
For this reason, the team led by Zhang Jiliang is exploring the potential of improving the wireless friendliness of prefabricated industrial buildings by regulating the electromagnetic properties of composite building materials through the integration of composite building structures and metamaterial - embedded building structures.
The research direction proposed by Zhang Jiliang's team has begun to attract the attention of the industry. In March 2025, the "Health and Care Infrastructure Construction Guide" issued by the UK National Health Service clearly stated that "wireless connectivity should be pre - planned before the building construction process."
Figure 2: Metasurfaces embedded in various building structures and materials
Intelligent Walls: The Path to Wireless - Friendly Buildings
Wireless - friendly buildings obtained by optimizing structures and materials are static, which is not enough to fully address the mobility problem of indoor users.
The latest research ([4]) of the team led by Professor Zhang Jiliang proposes embedding low - cost passive metasurface tiles into building structures (in the form shown in Figure 2), making metasurfaces ubiquitous and forming intelligent walls, thereby fundamentally improving the wireless performance of the network inside the building.
However, the general mobility of users introduces complex dynamic changes to the channel of metasurfaces embedded in buildings. To make the electromagnetic environment of wireless - friendly buildings dynamic and focus the reflected beams on moving users, it is crucial to have a deep understanding of indoor human behavior patterns.
The research for the first time attempts to present the wireless performance potential of metasurfaces embedded in buildings under the constraints of indoor human behavior.
This research has been accepted by the top journal in the communication field, IEEE Wireless Communications. The first author of the article is Wu Ziyang from Northeastern University, and the corresponding author is Professor Zhang Jiliang from Northeastern University. It was jointly completed with Professor Muhammad Ismail from Tennessee Technological University in the United States and Professor Zhang Jie, the co - founder of Ranplan Wireless.
- Preprint link: https://arxiv.org/abs/2507.14876
- Paper link: https://ieeexplore.ieee.org/document/11202168
First, we need to understand what problems human mobility brings to the building's wireless environment.
The mobility of user equipment is multi - scale.
Macroscopically, it is jointly constrained by the return tendency and the bounded Lévy walk. Users always tend to go to specific places at specific times, and when choosing destinations, there are obvious power - law distribution characteristics in the travel distance. This characteristic is scale - free, which means it holds at any scale. This research has confirmed through a large number of actual measurements and statistics that the macroscopic mobility of users in the indoor environment is still constrained by the above - mentioned laws.
On a smaller scale, users will choose specific ways to reach their destinations, including avoiding obstacles and the behavior patterns of other users, which further leads to frequent blockages of the line - of - sight links.
On the microscopic scale, the devices held by users are in constant slight vibrations for a long time. Many studies have shown that there are obvious differences in the statistical characteristics of device directivity under different behavior states such as walking, standing, and sitting.
The above - mentioned scale factors together dominate the tidal evolution characteristics of the channel of metasurfaces embedded in buildings.
Figure 3: Spatiotemporal evolution of the channel of metasurfaces embedded in buildings. (a) and (b) are snapshots of the spatiotemporal distribution of the channel statistics on the surface of the intelligent wall respectively. The distribution in each snapshot has been normalized, and the brightest area corresponds to the highest channel gain or link survival rate. The partial autocorrelation function of the intelligent wall channel state shows an increase in the Markov order. (c) shows the probability density function of the time - varying intelligent wall channel gain and the metasurface link survival rate.
Challenges Brought by Group Mobility Behavior
Since the high - frequency bands of 6G (such as millimeter - waves and visible light) are highly sensitive to occlusion and displacement, user mobility has become the core driving force for the evolution of channel modes, resulting in the absence of a universal channel model for metasurfaces embedded in buildings.
The uncertainty of human behavior forces the use of data - driven reflected beam tracking, but the channel evolution shows tidal characteristics and is predictable.
By utilizing this characteristic, only 10% of the metasurface area can be dynamically activated, and regional creeping activation can be achieved through low - complexity deep reinforcement learning, significantly reducing the control and power supply burden.
However, human behavior destroys the Markov characteristics of the channel and increases the dimensionality of the channel state space. The team led by Zhang Jiliang proposed a lightweight MuZero - like method, which is trained in the time - embedded latent space to generate control strategies resistant to environmental uncertainties.
Figure 4: Concept drift of the channel of the intelligent wall. The upper figure points out the sources of three types of channel concept drift. The lower figure shows the evolution process of the probability density function of the intelligent wall channel gain at different mobility stages.
There are significant differences in the intrinsic generalization ability at different wavelengths. For example, compared with visible - light communication, millimeter - waves have better diffraction ability, resulting in stronger randomness in their channel characteristics.
This randomness not only blurs the boundaries of shadow areas but also masks the details of the evolution of human behavior patterns.
Therefore, the algorithms trained for millimeter - wave channels naturally show better generalization ability. In contrast, the visible - light channel clearly depicts the projection of human behavior in the building environment, inevitably leading to the evolution of multi - modal channel characteristics and more serious generalization challenges.
Let's return to the original mission of buildings.
The "embodied intelligence" of the building's wireless environment cannot be achieved at the cost of compromising safety and basic functionality. These can be quantified and are easily used to constrain wireless - friendly buildings. However, concepts such as human behavior and architectural aesthetics are difficult to quantify but cannot be avoided.
A series of reinforcement learning and generative methods initially explored by the research team naturally have advantages in dealing with non - quantifiable concepts. It is believed that the "end - to - end" one - click generation of architectural design under the constraints of behavior science, aesthetics, etc. is on the horizon.
References
[1] Jiliang Zhang, Andrés Alayón Glazunov, Wenfei Yang, and Jie Zhang, "Fundamental Wireless Performance of a Building," in IEEE Wireless Communications, vol. 29, no. 1, pp. 186 - 193, February 2022, doi: 10.1109/MWC.121.2100244 (https://ieeexplore.ieee.org/document/9599590)
[2] Yixin Zhang, Jiliang Zhang, Xiaoli Chu, and Jie Zhang, "Wireless Friendliness Evaluation and Optimization for Sandwich Building Materials as Reflectors," in IEEE Transactions on Antennas and Propagation, vol. 72, no. 3, pp. 2697 - 2711, March 2024, doi: 10.1109/TAP.2024.3356052 (https://ieeexplore.ieee.org/document/10414382)
[3] Planning for the provision of wireless connectivity, NHS: https://digital.nhs.uk/services/networks - and - connectivity - transformation - frontline - capabilities/connectivity - hub/advice - and - guidance/wireless - infrastructure - building - guidance - for - health - and - care - organisation/infrastructure - guidance/
[4] Zi - Yang Wu, Muhammad Ismail, Jiliang Zhang, and Jie Zhang, "Tidal - Like Concept Drift in RIS - Covered Buildings: When Programmable Wireless Environments Meet Human Behaviors," in IEEE Wireless Communications, accepted, 2025, doi: 10.1109/MWC.2025.3600792 (https://arxiv.org/abs/2507.14876)
This article is from the WeChat public account "New Intelligence Yuan", author: KingHZ. It is published by 36Kr with authorization.