AT GTC 2024, NVIDIA President and CEO Jensen Huang’s highly anticipated keynote speech to a packed stadium, highlighted a groundbreaking set of new products, outlined a vision for the future of AI and GPU technology and highlighted the critical role of partners in driving innovation forward. He emphasized the priority for NVIDIA in automation and robotics.
Computer Vision Needs for Autonomous Robots:
Orbbec is well positioned to address this market opportunity for enhanced computer vision systems required for autonomous robots. Leveraging our portfolio of depth+RGB cameras renowned for high resolution and precision, Vision AI developers can craft sophisticated algorithms tailored for specific industries. With NVIDIA’s GPU-accelerated computing, developers can unlock new potentials in sectors such as manufacturing to logistics and healthcare.
During the conference, we had the opportunity to showcase our advancements in computer vision through exciting demonstrations.
Edge Computing:
We recognize the growing demand for real-time processing capabilities. To meet this need, Orbbec offers cameras integrated with NVIDIA’s feature-rich edge computing System on Modules (SOMs). This integration eliminates the necessity for dedicated attached compute with each camera, facilitating real-time processing of advanced depth algorithms. Seamlessly working with NVIDIA edge computing systems, Orbbec’s cameras enable low-latency and energy-efficient operations in a myriad of applications, including robotics, manufacturing, IoT, and healthtech.
Robotics:
Orbbec’s 3D vision technology plays a pivotal role in enabling AI-driven autonomous robotic systems to effectively perceive and interact with their surroundings. By providing rich and accurate depth data, our cameras empower robots in tasks such as object recognition, navigation, and manipulation. Leveraging NVIDIA’s Isaac ROS, users can deploy autonomous mobile robots (AMRs) and robot arm applications seamlessly, showcasing our commitment to enhancing robotic capabilities.
We had the opportunity to showcase our smart bin picking solution at NVIDIA GTC. This innovative solution, based on the Femto Bolt depth camera and UR5 Robotics Arm AccuPick 3D bin-picking solution from Solomon, uses Femto Mega and Orin AGX for guiding Universal Robots arm for detecting, picking, and placing items with unmatched precision. It effortlessly manages complex tasks involving overlapping, irregular, and even transparent or reflective objects.
This smart bin picking solution demonstrates Orbbec’s commitment to pushing the boundaries of what’s possible in robotics. Through technological advancements and AI, we continue to drive innovation in an increasingly automated world.
Simulation/Modeling/Design:
Orbbec cameras serve as invaluable tools for manufacturers, capturing essential data to inform design and research and development teams. The technology seamlessly bridges the gap between the physical and virtual realms, transforming real-world observations into virtual insights.
Virtual models of Orbbec cameras allow developers to utilize synthetic data for simulations in platforms like Omniverse and Isaac Sim, fostering innovation and iteration in a virtual environment.
One notable prototype featured was the Simultaneous Localization and Mapping (SLAM) Solution, which utilized Bird’s-Eye View RGB+Depth Vision alongside Orin AGX technology. This Autonomous Mobile Robot (AMR) prototype integrated a setup comprising bird’s-eye view and front-positioned Orbbec RGB+Depth sensors, along with ultrasonic long-distance sensors. This integration enabled real-time 3D reconstruction of the vehicle’s surroundings in voxel representation.
What sets this system apart is its seamless fusion of hardware and software components. By harnessing the power of CUDA computation, Orbbec stereo processing, PyTorch, TensorRT, and Isaac Sim, we achieved sophisticated object detection and obstacle avoidance. These capabilities are instrumental in enhancing autonomous navigation, ensuring optimal safety in dynamic environments.
High-Precision Scanning System:
We had an example of the High-Precision 50µm 3D Scanning System Based on NVIDIA Orin, developed with Purdue University, and powered by NVIDIA Orin Nano. This self-contained, portable unit simplifies high-precision scanning tasks, offering dual depth field of view (FOV) and built-in AE and HDR for detailed 3D captures. It seamlessly exports data directly to 3D printable formats, facilitating rapid prototyping and design iteration.
The versatility of this system extends to various industrial applications requiring in-situ high-precision 3D scanning and its capabilities empower industries to achieve unprecedented levels of accuracy and efficiency.
Generative AI:
Innovations in generative AI open new horizons for our robotics customers, who are now experimenting with unstructured language commands for robots. This integration holds promise for intuitive human-robot interaction, propelling the field of robotics forward.
Ecosystem Collaboration:
Strategic partnerships are key to our strategy for expanding our reach and impact. Through partnerships across various industries such as logistics, manufacturing, retail, healthcare and more, we are seamlessly integrating our 3D vision technology with NVIDIA edge systems into various applications and platforms. This collaborative effort not only enhances interoperability but also ensures compatibility with emerging AI, robotics, and automation solutions, ultimately expanding Orbbec’s influence in the market.
Orbbec remains committed to delivering cutting-edge solutions that pave the way for a smarter, more connected future. We aim to push the boundaries of AI, specifically in robotics, autonomous systems, and edge computing while driving efficiency and collaboration. Our dedication to pioneering autonomous systems ensures that our solutions not only meet but exceed the evolving demands of today’s dynamic landscape.
Frequently Asked Questions
1. What were the main themes of NVIDIA GTC 2024?
In his keynote to a packed stadium, NVIDIA President and CEO Jensen Huang introduced a groundbreaking set of new products, outlined a vision for the future of AI and GPU technology, and highlighted the critical role of partners in driving innovation forward, emphasizing automation and robotics as a priority for NVIDIA. Orbbec is positioned to address that opportunity through its portfolio of depth and RGB cameras known for high resolution and precision, which Vision AI developers can use to craft sophisticated algorithms tailored to specific industries. Combined with NVIDIA’s GPU-accelerated computing, this opens up potential across sectors from manufacturing to logistics and healthcare.
2. How do Orbbec cameras work with NVIDIA edge computing?
To meet growing demand for real-time processing, Orbbec offers cameras integrated with NVIDIA’s feature-rich edge computing System on Modules. This integration removes the need for dedicated attached compute with each camera and allows advanced depth algorithms to be processed in real time. Working seamlessly with NVIDIA edge computing systems, the cameras support low-latency and energy-efficient operation across applications including robotics, manufacturing, IoT, and healthtech.
3. How does Orbbec 3D vision support robotics on NVIDIA platforms?
Orbbec 3D vision technology helps AI-driven autonomous robotic systems perceive and interact with their surroundings. By providing rich and accurate depth data, the cameras support robots in object recognition, navigation, and manipulation. Using NVIDIA Isaac ROS, developers can deploy autonomous mobile robots and robot arm applications seamlessly.
4. What was the smart bin picking demonstration at GTC 2024?
Orbbec showcased a smart bin picking solution combining an Orbbec depth camera with the AccuPick 3D bin-picking solution from Solomon and a Universal Robots UR5 arm, with NVIDIA Orin AGX providing the compute for guiding the arm to detect, pick, and place items with high precision. The system handles complex tasks involving overlapping and irregular objects, as well as transparent or reflective ones, which are traditionally difficult for depth sensing.
5. How is Orbbec technology used in simulation and modeling?
Orbbec cameras capture data that informs design and research and development teams, bridging the physical and virtual realms by turning real-world observations into virtual insights. Virtual models of the cameras let developers use synthetic data for simulations in platforms such as Omniverse and Isaac Sim. One prototype shown was a SLAM solution using bird’s-eye view RGB and depth vision alongside Orin AGX. This autonomous mobile robot prototype combined bird’s-eye view and front-positioned Orbbec RGB and depth sensors with ultrasonic long-distance sensors, enabling real-time 3D reconstruction of the vehicle’s surroundings in voxel representation. Fusing CUDA computation, Orbbec stereo processing, PyTorch, TensorRT, and Isaac Sim delivered object detection and obstacle avoidance for safer autonomous navigation in dynamic environments.
6. What high-precision scanning system did Orbbec show at GTC 2024?
Orbbec presented a high-precision 50 micrometer 3D scanning system based on NVIDIA Orin, developed with Purdue University and powered by NVIDIA Orin Nano. The self-contained, portable unit simplifies high-precision scanning tasks, offering dual depth field of view along with built-in automatic exposure and HDR for detailed 3D captures. It exports data directly to 3D printable formats, which supports rapid prototyping and design iteration, and it suits industrial applications that require in-situ high-precision 3D scanning.
