RF Intelligence Operating System
AirStack Core software is the centerpiece of Deepwave’s platform. It allows the AIR-T to deliver AI-generated RF intelligence.
Core’s unique combination of capabilities allows RF signals to be analyzed at the edge so only AI-generated results are shared with connected IT systems. Sending only AI model results means the data output from the AIR-T is greatly reduced and – more importantly – low-latency intelligence is delivered rather than large volumes of raw RF (IQ) data.
For example, Deepwave’s platform is used to drive spectrum sharing for 5G telecommunication networks. The AirStack Core software optimizes high-performance compute on the AIR-T to support 5G, handles AI inference models on the AIR-T, and then orchestrates AI model outputs and data to the 5G network.
Here is a closer look at Core’s capabilities and key features:
DataOps, MLOps and Edge Management
AirStack Core software enables communication and control of AIR-T edge devices through application programming interfaces (APIs). AirStack Core APIs support RF data retrieval for model training, MLOps workflow tasks such as model deployments, and software defined radio (SDR) features to control transmit and receive.
Accelerated Computing
Core is the accelerated computing engine for the AIR-T platform as it orchestrates low-latency data transfer across an integrated software defined radio, and an FPGA, CPU, and GPU chipset – which optimizes memory management to minimize latency. This is the backbone of the AIR-T GPU-accelerated SDR system.
By utilizing GPU/CPU shared memory, the AIR-T system efficiently executes AI-driven algorithms and parallel processing tasks while minimizing latency. With the addition of a signal processor, wideband signal data passes through an FPGA for preprocessing before being written to shared memory, enabling enhanced data reduction, modulation, and other signal-processing functions.
Low Latency Data Transfer Technology1
AirStack Core is purpose-built to drive compute-intensive Software Defined Radio applications. Designed for high-performance throughput applications, Core’s architecture enables direct data transfer from RF receivers to a GPU-accelerated SDR system, optimizing memory management and minimizing latency. By utilizing GPU/CPU shared memory, the system efficiently executes AI-driven algorithms and parallel processing tasks. With the addition of a digital signal processor (DSP), wideband signal data passes through an FPGA for preprocessing before being written to shared memory, enabling enhanced data reduction, modulation, and other signal-processing functions.
RF Intelligence Engine
Deepwave’s Core software makes the intelligent edge possible. It combines GPU computing, AI science tools, and containerized software architecture to power hardware acceleration platforms such as the AIR-T.
Key Features
-
- Embedded AI & Signal Processing – Supports real-time neural network inference and cognitive algorithms on edge hardware.
- Containerized Application Deployment – Enables the use of built-in and third-party applications through Docker-based containerization.
- Seamless FPGA & GPU Integration – Optimized for Deepwave hardware with direct access to FPGA and GPU acceleration.
- Sensor Health & Status Monitoring – AirStack is preconfigured to report IIoT metrics, such as temperature, location, and computing utilization, right out of the box.
- Low-Latency & High-Performance – Designed for real-time processing in aerospace, defense, and industrial applications.
1 Patented technology. See article here for more information