Modelling & Simulation
Mapping radio waves in seconds with the help of machine learning
Finding just the right location is crucial for cellular network operators when it comes to maximising the coverage of a new base-station.
Radio frequency (RF) propagation modelling and simulation comes to the rescue, mapping where transmitter signals will go according to the surroundings – they will bounce off buildings, for example, causing shadow regions and be affected by the terrain.
But traditional physics-based modelling and simulation tools can take weeks to derive an optimum location – and that is no use when time is of the essence – e.g. for aid agencies operating in a disaster zone after an earthquake.
That’s where rapid RF propagation modelling and simulation comes in – using machine learning (ML) models to crunch the numbers. Once an ML algorithm is trained on data from traditional modelling tools, it can give an answer to a new problem in a fraction of the time and with far less computational resource.
Real-world challenge
In time-critical scenarios, the urgent need to establish reliable communication networks necessitates a rapid and efficient alternative to traditional physics-based models, but which also retain the key physical effects within an environment which govern how RF signals propagate.
That’s why the Plextek team developed a proof-of-concept ML model, based on modern generative AI techniques, and trained it on numerous coverage maps generated using conventional high-fidelity radio propagation tools – a process that required several days of compute time.
Once trained, the ML model could accurately predict coverage for previously unseen terrain in a fraction of a second, using only a standard PC or laptop – giving unprecedented speed and scalability in establishing effective and robust communication networks.
Crucially, the technique captures key physical effects such as diffraction, shadowing and channelling of the RF propagation to ensure accurate predictions.
The Plextek advantage
- ML algorithm development
Creating algorithms capable of learning from data to predict RF propagation patterns efficiently, representing a core technology that powers rapid modelling capabilities.
- Physics-based RF modelling proficiency
Deep understanding of traditional RF propagation physics to ensure accurate baseline models, crucial for training effective ML models.
- Data-driven network optimisation
Data analytics to enhance network performance and coverage, ensuring optimal use of resources in any terrain.
- Cross-terrain signal prediction
Ability to accurately forecast signal behaviour across diverse environments, from urban landscapes to rural settings, enhancing connectivity and network reliability.
- Computational performance enhancement
Enhancing the speed and efficiency of simulations on standard computing hardware, markedly decreasing the time required for propagation modelling.
- Emergency communications deployment
Setting up communication networks swiftly in disaster scenarios, a critical component of rapid response efforts.
- Advanced RF propagation analysis
Conducting assessments of RF propagation to inform the planning and optimisation of communication networks, incorporating complex environmental factors.
Modern Generative AI techniques are revolutionising our ability to perform complex modelling tasks which have, until now, not been considered feasible or practical with available computer hardware. By embracing the latest developments in AI/ML, we are now opening up a whole new set of potential capabilities – such as RF propagation modelling of complex environments at scale and pace.
Prof Aled Catherall
Chief Technical Officer
What sets us apart when it comes to rapid RF propagation modelling?
Combining Plextek’s deep knowledge of physics-based principles with our expertise in ML, we can deliver accurate and dependable coverage maps, ensuring rapid deployment of comms networks.
- ML
- RF
- ML-driven RF modelling
- ML in RF Planning
- Emergency response networks
- Real-time coverage mapping
- Scalable network planning
- Propagation modelling
- Efficient communication networks
