NVIDIA Holoscan Sensor Bridge Empowers Developers with Real-Time Data Processing

In the rapidly evolving robotics and edge AI landscape, the ability to efficiently process and transfer sensor data is crucial. Many edge applications are moving away from single-sensor fixed-function solutions and in favor of diverse sensor arrays. These include vision, audio, temperature, force/torque, and communication sensors, IMUs, idaradar systems, ultrasounds, motors, and actuators.

NVIDIA Holoscan Sensor Bridge (HSB) is an advanced sensor over Ethernet streaming technology, designed to enable real-time data streaming and simplify high-speed sensor and actuator integration. NVIDIA Holoscan incorporates HSB as the interface connecting physical sensors to the processing pipeline. It plays a critical role in enabling low-latency, end-to-end workflows where tight integration between hardware and AI processing are essential.  

This blog post dives into the challenges of high-speed sensor processing. It also highlights how HSB addresses these issues with ultra-low latency, ease of use, scalability, and robust safety features, for industries like robotics, medical imaging, and industrial automation

Challenges in high-speed sensor processing 

Physical AI and robotics developers face extreme challenges in sensor integration due to several key factors.

1. Development time and diversity
Developing sensor drivers and integrating them into systems is time-consuming due to the different types of sensors and their unique interface requirements. This complexity can slow down the overall development process.

2. Real-time processing and low latency
Achieving real-time processing and low latency is crucial for applications like robotics and medical instruments. Fast data acquisition, processing, and transmission must avoid delays that can lead to suboptimal performance or system failures.

3. Complexities in sensor streaming
Continuous sensor data streaming applications face challenges such as managing high data rates, ensuring data security, functional safety, and synchronizing data from multiple sensors. 

What is the Holoscan Sensor Bridge?

Holoscan Sensor Bridge is designed for low-latency data streaming and control. It transfers sensor data using user data protocol (UDP) over the Ethernet directly to GPU memory on systems like NVIDIA Jetson and NVIDIA IGX, reducing latency and CPU usage. Optimized for use with NVIDIA ConnectX SmartNICs and camera-over-Ethernet technologies, it enables real-time processing for video, edge AI, and robotics. HSB streams raw sensor data into the Holoscan SDK, supporting a unified pipeline from acquisition to inference and visualization.

Key features and benefits of HSB

Holoscan Sensor Bridge delivers tremendous value for developers to accelerate their development time without sacrificing performance. 

1. Ultra-low latency: Providing ultra-low latency ensures that data from sensors is processed and transferred with minimal delay. This is critical for applications that require real-time data processing, such as humanoids, autonomous vehicles, and medical imaging.
2. Ease of use: The software-defined architecture enables easy enumeration and API programming without FPGA programming expertise. This accelerates sensor software development for multiple sensors like cameras, IMUs, lidar, DACs and ADCs.
3. Scalability: With multi-modal sensor integration support, users can handle large volumes of data, which is essential for smart cities and industrial automation. HSB IP can support from 100 Mbps to 100 Gbps of bandwidth with customizable configuration parameters. 
4. Safety and security: It includes features such as redundancy, failure detection, watermarking, and Mac SEC support, contributing to achieving SIL 2 level safety. 

Introduction to Holoscan Sensor Bridge software 

With the NVIDIA Holoscan SDK developers can build high-performance streaming applications by composing modular operators into customizable pipelines. With the Holoscan Sensor Bridge host software, you can build custom pipelines and process data from network-connected sensors using ready-to-use operators for tasks such as image conversion, signal processing, inference, and visualization. Sensor objects provide device-specific APIs for configuration and monitoring, making it straightforward to adapt pipelines for different sensors and applications.

Holoscan applications are structured by dividing the main application class and defining the data pipeline in a configure method, connecting operators as needed. The platform is flexible and open: you can extend or modify operators (provided as source code) to meet unique requirements. 

Boosting performance with HSB vision sensors 

Modern embedded edge systems require cameras with diverse interfaces, high resolution, fast frame rates, low latency, and precise synchronization. HSB meets these needs by delivering up to 5x lower latency than USB cameras and 1.5x lower latency than MIPI cameras, with glass-to-glass latency as low as 17 ms. Using RDMA and camera over Ethernet, it enables direct data transfer to GPU memory with virtually zero CPU utilization, enabling quicker response times and real-time operation. With the HSB v2.0 release, developers can use the latest tools to accurately measure latency within their specific use cases.

With traditional mobile industry processor interface (MIPI) cameras, an Argus camera pipeline typically requires multiple kernel-space drivers, sending data to CPU memory before passing it to GPU memory. With HSB, traditional kernel-space drivers are replaced by user-space APIs, eliminating the need for developers to create separate drivers for camera and control functionalities. 

This streamlined approach enables developers to focus on application logic rather than low-level driver development. It also offers flexibility, enabling integration with a range of Image Signal Processor (ISP) options—including NVIDIA CUDA-based ISPs, soft-ISP implementations on HSB hardware, or the internal ISPs found on NVIDIA Jetson AGX and NVIDIA IGX. The following figures illustrate how HSB pipelines adapt to various configuration choices, highlighting the versatility and modularity of the platform. 

Reference examples for using different ISP options can be found in the user guide: 

• GPU-based CUDA ISP example that showcases how to connect a CUDA based ISP to an HSB Pipeline
• Jetson hardware ISP example that showcases how to connect the NVIDIA Jetson Hardware ISP to an HSB Pipeline 

Multi-sensor time synchronization 

Another key feature is HSB’s support for Precision Time Protocol (PTP), built on the IEEE 1588-2019 standard. This feature enables the Sensor Bridge to synchronize its internal clock with the host system, ensuring every piece of sensor data is precisely timestamped at the moment of acquisition. It achieves synchronization accuracy within 1 microsecond, and often surpasses 100-nanosecond precision—making it exceptionally well-suited for high-performance applications in medical imaging, robotics, and autonomous systems. With accurate timestamps, developers can confidently track exactly when each sensor event occurs, enabling data alignment across multiple sources and reliable coordination in distributed systems. 

HSB ecosystem

As a sensor-to-compute technology platform, HSB supports a vibrant ecosystem of sensors, hardware, and service partners, helping customers with faster time to market. Sensor partners offer a broad range of production-ready cameras and sensor solutions powered by HSB, delivering real-time performance and high-bandwidth processing. Hardware partners provide FPGA-based evaluation boards that enable developers to quickly design custom connectivity with low-latency, flexible sensor configurations and interfaces. Service partners specialize in leveraging the NVIDIA Holoscan software APIs to implement AI solutions and sensor integration, enabling optimized end-to-end performance.

Fast, accurate sensor data analysis is transforming medical imaging and diagnostics. Holoscan Sensor Bridge powers real-time analysis in endoscopes, surgical robotics, and radiology. Companies like Virtual Incision use HSB to accelerate camera development. Shane Farritor, CTO from Virtual Incision, highlights “With Holoscan Sensor Bridge, we’ve been able to revolutionize how we develop and deploy cameras. The ability to replace Verilog with Python and C code has drastically accelerated our development timeline from months to weeks. Programming for CUDA has made a world of difference in efficiency and performance.”

In software-defined radios, 6G, and test and measurement applications, analog sensors like data converters are essential. Holoscan Sensor Bridge connects these sensors to GPUs, leveraging NVIDIA IGX GPUDirect RDMA for rapid, efficient signal processing.

Getting started with the Holoscan Sensor Bridge

NVIDIA provides comprehensive documentation, developer kits, and reference examples to help users get started with the Holoscan Sensor Bridge. To explore the Holoscan Sensor Bridge:  

1. Get an NVIDIA Jetson AGX Orin Developer Kit or NVIDIA IGX Orin Developer Kit for your edge AI or robotics applications. 
2. Start with Holoscan Sensor Bridge evaluation boards from our hardware partners, Lattice (HSB  FPGA IP) and Microchip (HSB FPGA IP).
3. Download the HSB software and HSB user guide.
4. Watch how to use fully Integrated HSB sensors from our sensor ecosystem:
   1. Leopard Imaging showcases their 5.1 MPixel depth stereo HSB camera connected to Jetson AGX Orin for real-time vision applications and simplified integration. 
   2. Econ showcases their small-form-factor HSB camera connected to Jetson AGX Orin.
   3. D3 showcases an IP69 Rugged camera connected to Jetson AGX Orin through HSB, running a vision language model.
   4. YUAN leverages HSB for medical device integration of multiple modalities, such as HDMI, SDI, and analog signals. 
5. Learn more with HSB tutorial Ridge Run: Improving Latency with Holoscan Sensor Bridge.

Discover how AI, robotics, digital twins, and advanced simulation technologies are powering smarter, safer, and more efficient industrial automation at GTC Paris.

Watch the NVIDIA GTC Paris keynote from NVIDIA founder and CEO Jensen Huang at VivaTech 2025, and explore GTC Paris sessions.

Stay up to date by subscribing to our newsletter and following NVIDIA Robotics on LinkedIn, Instagram, X, and Facebook. Explore NVIDIA documentation and YouTube channels, and join the NVIDIA Developer Robotics forum. To start your robotics journey, enroll in our free NVIDIA Robotics Fundamentals courses today.

Get started with NVIDIA Isaac libraries and AI models for developing physical AI systems.