Deepwave Digital developed the first deep learning spectrum sensor for 5G networks.

2020-01-14

Deepwave Digital is using artificial intelligence radio transceivers (AIR-T) to create the first deep learning sensor for 5G networks. This network, called Citizens Broadband Radio Service (CBRS), will be the telecommunications industry's first autonomous spectrum sharing service that utilizes real-time RF sensing. The ability to determine whether priority users are active on specific channels is crucial for the operation of CBRS. When there are no priority users, spectrum can be reallocated for commercial networks to provide new enterprise services or additional bandwidth for existing services. AIR-T is a unique platform that combines radio frequency (RF) hardware with embedded NVIDIA Jetson TX2, enabling high-throughput digital signal processing (DSP). For CBRS, the Deepwave team has implemented deep neural networks (DNN) on AIR-T, which can detect, classify, and report the presence of priority users with high accuracy. By leveraging AIR-T and its AirStack development environment, the Deepwave team has demonstrated the ability to create, test, and deploy enterprise-grade signal processing solutions on the Jetson product line and AIR-T. With the explosive growth of IoT devices and LTE/5G-enabled smartphones, spectrum usage has significantly increased. As a result, spectrum has become increasingly congested, which can degrade network performance and reliability. Historically, spectrum has been managed by mandating each communication system to operate within a fixed frequency range, making spectrum management simple but requiring expertise in user demand. With technological advancements, user demands also change, leading to underutilization of certain frequency bands and congestion in others. The primary objective of CBRS networks is to provide a more dynamic approach to spectrum management. Deepwave Digital has developed deep learning algorithms as part of the Environmental Sensing Capability (ESC), which has undergone certification testing. This sensor will be a key component of the services provided by Deepwave's strategic partner, Key Bridge Wireless. At GTC DC 2019, Deepwave outlined various methods for executing DSP on NVIDIA GPUs and AIR-T. The key component of AIR-T is the onboard NVIDIA Jetson TX2, which provides ARM CPU and Pascal GPU as computing resources. For the current version of the ESC Sensor, Deepwave utilizes several GPU-accelerated libraries such as TensorRT, cuFFT, and CUDA itself. NVIDIA has recently released an open-source DSP toolkit called cuSignal as part of the RAPIDS accelerated data science project. By accelerating the popular SciPy Signal library with GPUs, cuSignal demonstrates the ability for Python programmers to easily write GPU-accelerated signal processing applications, making it easier for DSP engineers to leverage GPUs. Deepwave is currently evaluating whether cuSignal will be included in future versions and comparing it with our traditional workflow of directly using CUDA, cuFFT, and other software libraries. The deployment of the Key Bridge Wireless ESC network supported by Deepwave Digital AIR-T and DNN will launch in 2020 and begin serving enterprise customers. The network will initially be deployed along the coastlines of the continental United States, Alaska, Puerto Rico, Guam, and Hawaii.

RECENT AI NEWS

RECENT AI TOOLS