Comment: Using AI to build your next-gen wireless system

The wireless industry is experiencing an unprecedented surge in demand, with more than 7.1 billion human mobile users and a growing number of wireless machine-to-machine (M2M) connections. The central challenge engineers face when designing wireless systems and networks is their complexity. Traditional predefined designs are inadequate or inflexible when handling system complexity and unadaptable when requirements and environments change. Founded on the principle of learning and adaptability, emerging AI-native technologies promise to address the complexity challenge. 

What are AI-native wireless systems, and why are they superior to traditional designs? 

An AI-native wireless system integrates AI algorithms into its framework, offering better coverage, higher capacity and reliable robustness. AI-native systems adapt to their environment, overcoming scalability limitations and reducing signal processing costs. When designing these systems, engineers need real-world data sets, often sourced from physical prototypes or real-world signals. Most engineers use digital twins to augment data to train AI-native systems as they ensure there is sufficient data to handle adverse situations and efficiently manage system elements.

Designing and integrating an AI-native wireless system

Developing an AI-native wireless system is a complex process that involves creating a design workflow that includes gathering data, training and testing the model, and implementing and integrating the model into the wireless system. 

1.    Gathering and generating data 

The first step in creating an AI-native wireless system involves data collection, by either acquiring over-the-air (OTA) signals or synthesising data from a digital twin. Synthetic data enabled scalability testing, fault tolerance and anomaly detection, while aiding in environment modeling and system configuration. Engineers then use this data for training and validation of AI models, testing and simulation and performance tuning. The next step involves simulation and data modeling.

2.    Training and testing model 

Training a model for wireless systems requires determining key parameters like bandwidth, latency, and signal strength. Engineers then optimise machine learning algorithms for essential functions, considering real-time performance factors. After training, the model undergoes testing and adaptation to ensure reliable real-world performance. The final step involves pruning the AI network by converting it to a fixed point and removing unnecessary neural layers. This process prepares the model for implementation in the wireless system balancing performance with efficiency.

3.    Implementing the AI model 

Implementing an AI model in a real-world system involves three key steps. First, engineers must access scaling and resource assessment, including evaluating processing power, memory requirements and data throughput. The second step utilises automatic code generation to deploy pretrained AI models on desktop or embedded targets, streamlining implementation and reducing error. Finally, engineers can validate the implemented system’s performance against the original model to ensure there are no discrepancies or performance issues.

4.    Integrating the model 

The final step involves the integration of the implemented AI models within the overall wireless system. This phase ensures that the newly implemented AI solution works harmoniously with the rest of the legacy system. Before full-scale integration, engineers must ensure interoperability with existing system components by analysing the end-to-end system performance rather than individual algorithms and subsystems.  

Challenges of using AI to design wireless systems 

Integrating AI into wireless systems presents a variety of hurdles, including balancing conflicting performance metrics and ensuring superior performance relative to legacy systems. The goal is to achieve a balance that supports operational objectives by delivering high-quality overall performance. 

Balancing performance metrics

Optimising one metric in design often compromises another, requiring a balancing act to meet a system’s goals. For instance, increasing network throughput may raise power consumption and latency. Engineers can use modeling and simulation to explore scenarios, enabling informed decisions on optimal configurations without disrupting the actual systems.

Ensuring superior performance 

Transitioning to AI-enhanced wireless systems is challenging but crucial for superior performance. AI models that continuously learn are key, requiring diverse data sets for training. Simulating the integrated systems before deployment ensures proper interoperability with legacy systems. Engineers then use analysis and design tools to facilitate compatibility testing and identify performance bottle necks.

Conclusion 

The wireless industry is at a critical juncture. With the upcoming rollout of 5G Advanced and 6G standards, the next generation of wireless systems will deploy more AI-native technologies. Engineers tasked with designing modern wireless systems have realized that integrating AI is no longer optional; it is essential. By incorporating AI-native design principles, wireless engineers can develop systems and networks that meet today's needs and are equipped to evolve with tomorrow's wireless requirements and advancements. 

Houman Zarrinkoub, principal product manager, MathWorks