Digital twins for
radio spectrum

The recent concept of Digital Twins (DTs) is starting to be applied to data-driven models of the spectrum environment. Although it is still at an early stage, the concept has huge potential implications for the future of spectrum management. This article explores the possible use cases, technology enablers, and challenges. 

What is a digital twin?

A DT is a virtual representation of a real-world physical system, process, or object continuously collecting and using real-time data to update itself. It thereby keeps track of the state and behaviour of the original target, enabling better monitoring, analysis, simulation, and optimisation without interfering with the actual physical asset or system.  

Through sensors and data integration (and increasingly AI techniques), DTs can predict maintenance needs, test the outcomes of different scenarios or possible actions, and improve asset/system performance, while reducing risks and costs.  

Typical use cases for DTs include representing individual assets like factories, vehicles, or machines, as well as larger systems such as smart cities or environmental zones like watersheds and forests. 

Digital Twin 1
Digital Twin 2

How does the DT concept apply to spectrum?

The same broad principles can be applied to “hard” assets such as machines, or less tangible but still real-world phenomena such as the atmosphere, weather systems, electricity flows, or radio spectrum. Creating DTs for these assets still requires sensors, data integration, modelling, and appropriate computational platforms.

Where an industrial digital twin might focus on preventative maintenance and asset performance optimisation, a spectrum digital twin might examine how to mitigate the risks of interference or optimise the effective utilisation of a given frequency band. It could operate at a local level (for instance, inside a building or enterprise campus), across a mid-sized area such as an airport or city, or potentially across an entire nation and beyond.  

A significant difference to consider is that a spectrum DT will likely have to deal with multiple separate frequency bands occupying the same physical space, with the potential for interference between them.

While wireless companies regularly use design, planning, and monitoring tools, they are often static or updated irregularly. A mobile network operator (MNO) might use periodic drive tests to check their network coverage and performance, as well as data from their network elements and end-user device feedback. However, this does not mean they have a proper “twin” or realistic model of the spectrum environment. Regulators and other industry participants are similarly limited in their views of the “spectrum-scape”.

A spectrum-specific DT would typically combine data from radio transmitters with propagation maps generated from measurements and models. This information is then overlaid onto geographical information systems (GIS) for large-scale environments or building information models (BIM) for smaller-scale or indoor settings. It integrates a real-time snapshot of sensed and reported radio activity with a historical log of past activities, providing a comprehensive view of spectrum usage. 

The DT would be able to:

management-icon

Improve existing operations of spectrum management or a specific network 

depiction-icon

Maintain and improve its own realistic depiction of the “spectrum-scape” 

capabilities-icon

Enable potential new capabilities and services, based on the DT’s knowledge and data 

The key components & abilities of a
spectrum DT would include:

Real-time sensing

Real-time sensing and data collection mechanisms, with a high degree of distribution 

Multi-dimensional capabilities

Multi-dimensional capabilities, incorporating not just time and space but also frequency, directionality, user/device type, mobility, and power levels 

AI/ML

Modelling and prediction capabilities, likely based on AI and machine learning 

Scenario development

Simulation capabilities, allowing for scenario development and testing 

Visual interfaces

Visual interfaces for human users and APIs / defined interfaces for software integration with other systems (including adjacent DTs, such as those for network infrastructure) 

Regulatory compliance

Parameters reflecting regulatory rules, licenses, and constraints 

Use cases and stakeholders for Spectrum Digital Twins

Various specialised spectrum DTs (or overlays) will likely evolve to meet the needs of different stakeholders such as regulators, MNOs, enterprises, and government / military users. 

There are many potential use cases for a spectrum DT, depending on its accessibility and granularity. These could include:  

Data-driven planning
Improved data-driven regulatory spectrum planning and policy development, especially in the context of innovations such as new sharing schemes, wireless power transmission, or 6G development
Spectrum-sharing schemes
Acting as a critical data source for spectrum-sharing schemes, whether dynamic (such as CBRS or AFC) or administrative such as UK or German localised access licenses.
Improved efficiency
Improved operational efficiency for networks of various types, allowing for simulations to optimise trade-offs of coverage vs. capacity vs. interference vs. cost in different settings
Better valuation
Enabling better valuation of spectrum assignments, based on historical / current utilisation trends, or predictions of future use
Dedicated resources
Enabling mobile operators or other spectrum holders to identify and dedicate resources for localised private networks, either from their existing spectrum portfolio or from additional local licenses where available
Simulating conditions
Simulating the effects of extreme weather or climate change on RF propagation
Preventative maintenance
“Preventative maintenance” of spectrum assignments and operations by tracking and forecasting interference in particular places or contexts
Compliance monitoring
Identification of “spectrum anomalies” and assessment of risks or dangers from them. Regulatory bodies can monitor compliance with regulations and identify potential violations.
Environment modelling
Modelling the spectrum environment for multi-orbit satellite systems, direct-to-device hybrid satellite / terrestrial networks, or other 3D use-cases such as HAPS (High Altitude Platforms).
Better performance
Help military and public safety agencies train for complex spectrum scenarios, gain better situational awareness, and derive better performance in contested environments
Frequency of visibility
Improve the liquidity of the secondary marketplace for spectrum sales / leases / sub-licencing, by improving the visibility of which frequencies are available and where
Sharing & pooling
Enabling new spectrum usage models to be tested and deployed, such as new sharing and pooling models or varying uplink / downlink ratios to suit local application needs.
Optimize & manage technologies
Enabling broadcasters to optimise their coverage areas or manage the planning and execution of transitions between bands or different technologies
Optimize spectrum use
Allow critical infrastructure sites such as airports, ports, and utility grids to optimise spectrum use for different applications, or mitigate and identify sources of interference nearby. There are also potential applications in border regions.
Improved channel management
Improve coexistence and coordination in unlicensed bands, for instance, between technologies such as Wi-Fi, fixed wireless links, and Bluetooth, or for channel management in areas with multiple Wi-Fi networks
Added layer of protection
Assist in protecting “receive-only” spectrum users such as scientific monitoring, radioastronomy, and environmental sensing.

The roles of spectrum sensing and AI 

 

Multiple technology enablers are required for creating spectrum digital twins, but two especially critical ones are sensing abilities and various AI techniques for analysis, modelling, simulation, and prediction.   

There are numerous sources of information about the spectrum environment. Still, ideally, a DT will collect data from “independent” sensors beyond the self-reporting and RF measurement from expected / registered radio systems such as cellular base stations and users’ handsets. Regulatory authorities and other spectrum management agencies will be particularly interested in operating their own sensing systems rather than relying on licensees’ reported data—which, in any case, may not be realistic for legacy or secure systems. 

In essence, there needs to be a more real-time and distributed equivalent of drive tests and other periodic testing, especially if the DT is designed to be multi-technology and multi-band. Attention also needs to be paid to the patterns of interference and propagation and how those change with time or other situational conditions. That allows the twin to be updated directly from sensed data and with modelling where needed, making detecting anomalies easier. 

There is also likely to be a strong link with the evolution of software-defined radios (SDRs) and other flexible, programmable platforms such as Open RAN and Open Wi-Fi that can repurpose normal wireless transmission and reception as a sensing function.  

Depending on the desired use cases for the DT, spectrum sensors may be concentrated in particular geographic zones, such as maritime areas, airports, near borders, around urban centres, or military bases or deployment regions. They may be collocated with existing wireless assets such as broadcasting infrastructure or cell towers or linked to deployments of other environmental sensing and monitoring systems. 

busy-airport

AI and Machine Learning will also be critical for DTs, alongside sensing. Advanced algorithms can process sensing data to identify patterns, predict potential issues, and endogenously improve the accuracy of spectrum usage models. Other AI functions include: 

  • Predictive analytics, combining historical and real-time data to predict spectrum usage patterns, potential conflicts, and optimal sharing arrangements. These predictions can enable proactive management strategies to mitigate interference risk and optimise utility and value. 
  • Process automation, for instance, in terms of issuing licenses or enforcing existing regulations. Generative AI tools could also simplify the processes for obtaining spectrum and help enable “spectrum as a service” by combining a DT’s models with user-friendly access mechanisms. 
  • Anomaly detection and alert filtering could quickly identify unusual spectrum usage patterns or potential security threats, enabling rapid response to unauthorized transmissions or interference sources, while avoiding the difficulties of “false positives.” 

Challenges for spectrum DTs

It should be noted that creating and operating spectrum DTs has numerous challenges, which will need to be addressed with iterative research, testing, and deployment cycles. This emergent domain will likely proceed in multiple steps and phases. Issues to address include: 

  • Data collection, storage and integration – there will be vast amounts of heterogeneous data from various spectrum users and sensors. This may include real-time data on spectrum usage, signal propagation, and interference patterns across different frequency bands and geographical zones. 
  • Modelling complexity – this will involve the intricate interplay of mapping and topographical data, together with fast-evolving communication protocols, multiple configurations of transmitters and receivers (including those in motion), many physical environments (including indoors), and even the impact of new types of device or application. 
  • Simulation accuracy and speed – this is a common problem across many sectors using DTs. There is inherent latency in the sensing / modelling / prediction cycle, especially at high levels of granularity and collection of large volumes of measurements from widely-distributed data sources. The nature of many radio technologies means that the physical use of spectrum can vary on a sub-millisecond basis, for instance, with MIMO and phased array antennas. The processing and timing requirements for spectrum sensing and analytics may also prove to be one of the important use cases for edge computing 
  • Interoperability – this will likely be a significant challenge, given the different systems, models, semantics, and interfaces between different systems using or monitoring spectrum. While interoperability within one technology is usually quite feasible, dealing with shared spectrum bands with multiple types of users (e.g. radar, cellular, Wi-Fi, Bluetooth, and microwave links) will pose huge challenges. This is a key reason why independent sensing functions may be essential. 
  • Maintaining synchronisation of the spectrum digital twin with the real world will be challenging – it will require not just real-time sensing and modelling of spectrum users but also adaptations to the physical environment (e.g., building construction, foliage growth, and even vehicle movements) and also evolutions in new way systems transmit and receive radio, such as updated antenna designs. 
  • Visualisation and interfaces – There will be a strong need for spectrum DTs to encompass ways to output data effectively to human and machine applications. Given the multi-dimensional nature of spectrum data, this will pose some significant design challenges, although GenAI may prove useful here. 

 

As well as direct technical challenges, there are additional considerations around standardisation, regulation, and commercialisation. In particular, there are potential risks and threat surfaces to any form of spectrum monitoring or modelling, especially where some of the user groups are related to critical infrastructure or emergency / defence user groups. 

Conclusions and timelines for spectrum DTs

While spectrum-specific DTs are certainly an interesting future trend, there is currently limited real-world deployment (at least that which is publicly acknowledged). While there are numerous research projects in this space, it seems likely that full deployment will only occur over the next 5–10 years in tandem with the evolution of adjacent technologies such as 6G, radio-centric AI tools, and better systems of distributed sensing and data collection. 

There are three vectors to watch here: 

Evolution and deployment of key technical enablers, such as sensing grids and spectrum-optimised AI models

Growth of key DT customer groups and use cases, such as new dynamic spectrum-sharing models, military spectrum sophistication, and new AI-native 6G standards capable of using DT-derived data and simulations 

Regulatory support for DTs, both in terms of internal need for advanced spectrum management and in addressing privacy and security implications for other users. 

Also, it is currently difficult to gauge whether spectrum DTs will operate primarily on a standalone basis or if they will be integrated into other “system” DTs.  

For instance, MNOs will likely run digital twins of their entire networks, including both the active radio systems and spectrum, as well as the hard infrastructure components such as cell-towers, backhaul, and indoor wireless systems. They will want to optimise their investments in spectrum alongside physical assets such as rented rooftops, power supplies, and fibre connectivity. However, they will have limited data about other spectrum users’ performance and interference either caused or experienced. 

Conversely, a dedicated spectrum management agency within government will probably want a more multi-band, multi-technology “pure” DT, but may lack some of the input data from the network infrastructure itself. Over time, we can probably expect spectrum DTs to develop a variety of interfaces and integration options. 

In summary, we should expect a deepening of interest, investment and utility from spectrum DTs. Potentially, they could form a central part of future spectrum management, as enablers for regulators to better manage demand for spectrum and optimise radio frequency utilisation and value, while minimising interference risk in real-time. As part of that, we should also strive for cross-sector collaboration, in order to integrate data and models from multiple technologies and spectrum stakeholder groups. 

CRFS RF sensor overview

CRFS RF sensor overview

Portable and rugged high-performance RF sensor for real-time 24/7 spectrum monitoring and geolocation of transmitters

Dean Bubley-1
Dean Bubley

Dean Bubley, founder of Disruptive Analysis, writes guest posts for CRFS. He is an independent analyst and advisor to the wireless and telecoms industry and has covered the evolution of private cellular networks since 2001.