What applications and technology trends prompt the need for artificial intelligence/machine learning algorithms in RF test and measurement?
Emerging use cases such as 6G, 5G Advanced, IoT, and AI-driven wireless systems are creating a complex wireless landscape where the scarcity of spectrum and higher data rates are critical issues. In this scenario, complex interference, multipath propagation, and signal variations are challenging traditional methods of signal characterisation. To address these challenges, AI can manage large, complex datasets to identify patterns and anomalies much more swiftly than conventional methods, particularly in finding and characterising interference sources and RF signals in dynamic wireless environments. AI can also be utilised to optimise test flows, automate procedures, reduce downtime through predictive maintenance, and accelerate the writing of test scripts.
How are different approaches to regulating RF spectrum use creating a need for flexible solutions?
While the USA and China have a single regulator overseeing large markets, the EU has regulators in each member country, resulting in needs for cross-border coordination and thus a highly fragmented market with many distinct operators.
The European Conference of Postal and Telecommunications Administrations (CEPT) plays a crucial role in harmonising radio spectrum regulations across Europe, but ultimate spectrum allocation decisions remain with individual national governments. CEPT’s mandate from the European Commission includes tasks like identifying frequency bands and spectrum use conditions to support EU policies, as well as developing technical conditions for new technologies like 5G and 6G.
Nonetheless, the International Telecommunication Union (ITU) and the World Radio Conference (WRC) facilitates international coordination on the radio frequencies utilised to prevent interference between radio systems. It also advocates for shared global use of the radio spectrum and aids in the development and coordination of worldwide technical standards. Consequently, global spectrum utilisation is complex, involving both national and supranational organisations. From a testing perspective, AI can assist in adjusting test equipment to comply with all these standards by swiftly reconfiguring the testing systems. As 6G progresses, cognitive radio, spectrum-sharing techniques, and dynamic RANs (Radio Access Networks) will create an ever-evolving RF landscape governed by AI. Adaptable AI-driven test systems will be key here.
How does AI build on the capabilities of existing spectrum monitoring solutions?
AI-based algorithms, particularly those based on machine learning, can analyse, characterise, and classify complex radio signals more quickly and accurately than manual traditional spectrum analyser methods. AI offers the advanced monitoring and signal characterisation required by emerging spectrum-sharing techniques, such as dynamic spectrum sharing and new emerging use cases for 6G. Moreover, AI can locate, identify, and characterise sources of interference significantly faster and more accurately. It can also be used to automate rules-based spectrum sharing techniques based on information from RF-sensing.
Why is adding AI to test processes as important as designing more sophisticated hardware?
Currently, test hardware generates vast amounts of data, and more advanced hardware will produce even more. As this data must be analysed, the analysis and characterisation of complex RF signals are of primary importance. AI finds anomalies and characterises data faster and more accurately. In RF sensing, AI excels at detecting signals and characterising them through machine learning. Integrating AI with hardware allows test systems to advance with evolving wireless networks without traditional limitations. AI will effectively optimise the use and efficiency of the underlying hardware to manage a dynamic RF environment, particularly for analysis and signal characterisation.
How are AI-based algorithms incorporated into test processes?
AI software is typically run on test equipment or on a separate computing platform. For example, Anritsu has integrated proven AI machine learning technology developed by DeepSig into its MS2090A/MS27201A Field Master Pro spectrum analyser, allowing it to surpass the capabilities of current spectrum monitoring systems. DeepSig software enables the test system to detect and classify signals while concurrently analysing the spectrum environment to provide contextual analysis and facilitate decision-making.
Why is it important to start with reliable, robust datasets, and how are they acquired?
AI is only as effective as the data model on which it relies. It uses various algorithms, each with its strengths and weaknesses, making some models more suitable for specific tasks than others. Moreover, one must consider the type of data and its complexity when selecting or developing an AI model. Training AI models is essential, involving the process of feeding algorithms with data to enable them to recognise patterns and make predictions. This process depends on thorough data preparation to ensure that the model can efficiently carry out specific tasks with high accuracy. In RF test cases, robust data sets can be obtained from previous test results and knowledge of specific waveforms, including both measured and synthetic data. AI for a specific task, such as signal characterisation, must use relevant data related to the task and not be compromised with irrelevant data to ensure optimal performance.
How much does AI speed up learning of signals from new sources?
AI can significantly speed up learning signals from new sources by using machine learning training models. For instance, Anritsu adopts a deep learning, data-driven approach based on DeepSig machine learning training tools to integrate new radio signal models into their testing capabilities rapidly. RF signals of interest from various new sources, such as cognitive radios, drones, and IoT devices or any radio emission, can be learned quickly and accurately within days rather than months to satisfy rapidly changing customer requirements.
How will machine learning help to form the foundation for AI-native RF sensing for 6G?
As the radio spectrum is a scarce and expensive asset, it needs to be managed, shared, and optimally utilised in wireless networks, especially with the emergence of new wireless technologies such as 6G, drones, and the IoT. Machine learning will enable systems to adapt to new wireless signals and automate networks to respond dynamically to changing loads and requirements. Native RF sensing is necessary to identify and characterise signals. Networks need this information, for example, to implement interference mitigation, load balancing, and adapt to outages. Spectrum-sharing techniques, dynamic RAN, and cognitive radio are technologies that depend on continuous RF sensing for effective operation.
Envisioned for 6G, originally to address spectrum scarcity from sub-6GHz to THz applications, Joint Communications and Sensing (JCAS) aims to integrate sensing functions into communications networks with lower energy consumption. To provide communication capacity and sensing accuracy at the same time, MIMO is expected to play an important role due to its spatial beamforming and waveform shaping abilities. The challenge is to bring down the cost and energy consumption associated with MIMO. Emerging technologies, such as Cloud radio access networks (C-RANs), unmanned aerial vehicles (UAVs), and reconfigurable intelligent surfaces (RISs), provide potential solutions. JCAS also aims to empower connected and automated mobility, remote health monitoring, Industry 4.0, and autonomous vehicles that not only exchange data but also rely on precise information about their surrounding environment. JCAS is anticipated to utilise classical signal-processing optimisation models alongside advanced machine-learning techniques.
How are established test and measurement businesses like Anritsu partnering with AI specialists to create new solutions?
Anritsu is bringing advanced AI capabilities to its test systems to address complex challenges within wireless communication systems by utilising DeepSig’s AI machine learning technology. Traditional RF sensing techniques face challenges in rapidly evolving wireless environments, which demand more sophisticated monitoring and signal characterisation. Anritsu addresses this by adding DeepSig’s machine-learning capabilities to its test systems. Anritsu’s use of AI will enable customers to enhance network performance, optimise spectrum use, and achieve real-time adaptation to changing RF conditions. Further, such test systems will effectively be future-proof as they are adaptable to new environments with new signal sources through machine learning, as is expected with 6G.
What part will humans play in future test scenarios that incorporate AI?
AI will make humans more productive and enable tests that are not currently possible due to time constraints and the volume of data that needs analysis. Humans will still need to oversee AI and ensure that it works as specified. As networks become much more complex and automated, the growing need for RF testing will continue to grow. AI will serve as a tool for engineers and technicians to manage complex
and dynamic networks, which would be unfeasible without advanced automated AI tools.
This article originally appeared in the September’25 magazine issue of Electronic Specifier Design – see ES’s Magazine Archives for more featured publications.
