NVIDIA AI Aerial: AI-Native Wireless Communications

2025-10-01

AI Summary

Traditional cellular networks suffer from fragmented signal processing (DSP) and machine learning (ML) modules, leading to inefficient deployment and poor runtime integration. Method: This paper proposes a tightly integrated DSP-ML software stack for AI-native 6G wireless systems, featuring a Python-to-GPU compilation pipeline that unifies CNN-based ML models and DSP algorithms into high-performance GPU-executable modules, and a digital twin–driven closed-loop development paradigm enabling end-to-end deployment—from simulation-based training to real-time wireless testbeds. Contribution/Results: It achieves, for the first time at the software stack level, seamless DSP-ML co-execution, validated via a CNN-based channel estimator in a PUSCH receiver. Experiments demonstrate significant improvements in inference latency and channel estimation accuracy, alongside high scalability and cross-environment consistency—providing a practical, deployable pathway for intelligent lifecycle management in 6G systems.

Technology Category

• Machine Learning: Hardware-aware ML
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Security and Privacy: Security and privacy of machine learning and AI applications

Abstract

6G brings a paradigm shift towards AI-native wireless systems, necessitating the seamless integration of digital signal processing (DSP) and machine learning (ML) within the software stacks of cellular networks. This transformation brings the life cycle of modern networks closer to AI systems, where models and algorithms are iteratively trained, simulated, and deployed across adjacent environments. In this work, we propose a robust framework that compiles Python-based algorithms into GPU-runnable blobs. The result is a unified approach that ensures efficiency, flexibility, and the highest possible performance on NVIDIA GPUs. As an example of the capabilities of the framework, we demonstrate the efficacy of performing the channel estimation function in the PUSCH receiver through a convolutional neural network (CNN) trained in Python. This is done in a digital twin first, and subsequently in a real-time testbed. Our proposed methodology, realized in the NVIDIA AI Aerial platform, lays the foundation for scalable integration of AI/ML models into next-generation cellular systems, and is essential for realizing the vision of natively intelligent 6G networks.

Problem

• Integrating DSP and ML in 6G wireless systems
• Compiling Python algorithms for GPU execution
• Enabling AI-native cellular networks via unified framework

Innovation

• Compiles Python algorithms into GPU-runnable blobs
• Integrates DSP and ML in cellular network stacks
• Uses CNN for channel estimation in PUSCH receiver