TL;DR:
- Wireless AI plays a crucial role in future wireless systems and 5G Advanced.
- On-device AI enhances E2E optimization by distributing processing tasks and personalized data.
- AI-native processors like Qualcomm’s Snapdragon X75 boost AI performance.
- Cross-node AI is essential for true E2E performance optimization.
- 5G Advanced Release 18 use cases focus on ML-driven channel feedback, beam management, and precise positioning.
- AI-based channel estimation overcomes the limitations of model-based CSI feedback.
- Sequential training enables ML-based CSI in multi-vendor systems.
- ML-driven beam prediction improves network throughput and power efficiency.
- ML enhances positioning accuracy for indoor and outdoor environments.
- 6G air interface must be AI-native to support future network demands and innovation.
Main AI News:
In the ever-evolving landscape of wireless technology, the convergence of 5G and advanced artificial intelligence (AI) marks a pivotal moment in the industry’s trajectory. As we embrace the transition to 5G, traditional rules-based design methods are no longer sufficient to manage the increasing complexity of wireless systems. This realization has ignited a surge in interest in generative AI, underscoring its vital role in shaping the future of wireless networks.
The Impending Transformation: 5G Advanced
With the advent of 5G Advanced, wireless AI is set to redefine the way networks are designed and operated over the next three to five years. This leap forward will introduce innovative AI applications that enhance network performance and device optimization, heralding a new era of seamless wireless connectivity. Crucially, wireless AI will emerge as a key pillar of 5G Advanced, spearheading end-to-end (E2E) design and optimization across protocols and network layers.
The On-device AI Paradigm
As smartphones and other devices have already embraced AI for several years, the industry now turns its focus to incorporating AI into the network itself. The current approach of implementing AI independently either on devices or within the network presents limitations in achieving E2E systems optimization.
Enter on-device AI – a game-changer for the E2E optimization of 5G networks. By distributing processing tasks across millions of devices, on-device AI harnesses the collective computational power to achieve significant improvements. Moreover, it enables AI models to be tailored to a user’s personalized data, leading to enhanced reliability and data sovereignty. This approach extends far beyond smartphones and extends its influence to consumer devices, sensors, and diverse industrial equipment.
The Rise of AI-native Processors
The development of AI-native processors, such as Qualcomm’s Snapdragon X75 5G modem-RF chip, is a critical step towards realizing on-device AI’s full potential. By integrating dedicated hardware tensor accelerators, this second-generation AI processor exhibits more than 2.5 times the AI performance of its predecessor, revolutionizing the processing power available on devices.
Toward True E2E Performance Optimization
While on-device AI serves as a pivotal advancement, achieving genuine E2E performance optimization requires collaborative efforts across the entire network. To make this a reality, wireless system designers must integrate AI training and inference at a systems-wide level, embracing the concept of cross-node AI. This approach emphasizes the need for both AI expertise and deep domain knowledge in wireless technology.
Wireless AI in Action: 5G Advanced Release 18 Use Cases
The potential of wireless AI is being realized in 3GPP’s Release 18, setting the stage for extensive use in the future 6G landscape. Three use cases take center stage in this release:
1. Cross-node Machine Learning (ML) dynamically adapts Channel State Information (CSI) feedback between base stations and devices, optimizing network and device performance in coordination.
2. ML-driven intelligent beam management enhances network capacity and device battery life by selectively measuring and predicting future beams, reducing power consumption.
3. ML improves positioning accuracy for devices in indoor and outdoor environments, elevating both direct and ML-assisted positioning capabilities.
Redefining CSI with AI-based Channel Estimation
Accurate CSI is integral to reliable communications. The traditional model-based CSI approach, while useful, still faces challenges in compressing and handling signaling overheads. In contrast, AI-based channel estimation employs a data-driven air interface that dynamically learns from its environment, achieving higher accuracy and improved link performance, particularly at the cell edges.
Overcoming Vendor Hurdles with Sequential Training
The implementation of ML-based CSI feedback can be complex in multi-vendor systems. To overcome this obstacle, Qualcomm introduces sequential training, allowing individual devices to be trained using their proprietary data before sharing it with the network. This approach eliminates the need to share neural network models across vendors, streamlining the integration of AI.
Pioneering ML for Beam Prediction
AI takes center stage once more in intelligently predicting beams on millimeter-wave radios, significantly enhancing network throughput and reducing power consumption. Rather than continuously measuring all beams, ML algorithms intelligently select and interpolate between beams, demonstrating equivalent performance to conventional measurement setups.
Enabling Precise Positioning with ML
Qualcomm’s groundbreaking demonstrations reveal the potential of ML for enabling precise positioning in both outdoor and indoor industrial networks. Utilizing multi-cell roundtrip and angle-of-arrival-based positioning, coupled with RF fingerprinting, ML-powered positioning overcomes non-line-of-sight channel conditions, fostering unprecedented positioning accuracy.
The AI-native Air Interface: A 6G Revolution
As the industry looks beyond 5G, 6G looms on the horizon, demanding a substantial leap in performance and spectrum efficiency. To achieve this, the 6G air interface must be inherently AI-native, transitioning from a model-driven design to a data-driven paradigm. By integrating ML across all protocols and layers, distributed learning and inference across devices and networks will unleash a new era of wireless connectivity.
The Road to Endless Possibilities
Embracing 5G Advanced wireless AI/ML as the foundation of 6G innovation promises a plethora of new network capabilities. From refining existing communication protocols to learning new ones, the AI-native air interface adapts dynamically to various deployment scenarios, radio environments, and use cases. This unprecedented customization will empower operators to serve diverse markets with tailored solutions, solidifying the wireless industry’s boundless future.
Conclusion:
The integration of wireless AI and 5G Advanced represents a transformative shift in the cellular network market. The industry’s focus on AI-driven optimization, enhanced performance, and precise positioning will redefine the wireless landscape, enabling operators to cater to diverse markets with tailored solutions. As the foundation for 6G innovation, AI-native air interfaces will propel the market into an era of endless possibilities, revolutionizing the way communication systems are designed and operated. Embracing this technological advancement will be vital for businesses to stay competitive and meet the ever-increasing demands of the wireless industry.
