Trossen Robotics puts NVIDIA Jetson AGX Orin 64GB in Rivet and Workbench

Trossen Robotics said its new Rivet mobile manipulation platform and Workbench benchtop system will ship with NVIDIA Jetson AGX Orin 64GB as factory-integrated compute. The move gives developers a common AI stack across tabletop and mobile robotics development, with preorder available now.

Why it matters: - Trossen Robotics is standardizing the compute layer across two physical AI platforms, which can cut the rework needed when teams move from benchtop testing to mobile robot development. - The shared architecture is meant to keep perception, policy development and validation on the robot instead of pushing data back to remote systems.

What happened: - Trossen Robotics announced Rivet and Workbench, two new physical AI development platforms built around NVIDIA Jetson AGX Orin 64GB. - Standard configurations of both products will ship with the Jetson module factory-integrated by Trossen Robotics. - Rivet is a mobile manipulation platform for teleoperation, robot learning, multimodal data collection and supervised application pilots. - Workbench is a compact stationary system for repeatable data collection, policy development and application evaluation at a table or workcell.

The details: - Jetson AGX Orin 64GB delivers up to 275 TOPS of AI performance and supports power settings from 15W to 60W. - The module includes an NVIDIA Ampere architecture GPU, 64GB of LPDDR5 memory, deep learning and vision accelerators, high-speed interfaces and multimodal sensor support. - Trossen integrates the computer, thermal management, power delivery, networking and sensor connectivity at the factory. - Rivet combines configurable manipulation arms, an omnidirectional mobile base, vertical reach adjustment and integrated sensing in a compact system. - Workbench uses the same core compute, sensing and networking architecture in a stationary format. - The shared Jetson architecture gives teams a more consistent edge-computing target across mobile and stationary workflows. - Developers can use NVIDIA JetPack SDK, NVIDIA Isaac ROS, CUDA-accelerated libraries, models, tools and ROS 2 packages alongside Trossen Robotics’ open SDK, drivers, robot descriptions and simulation resources. - Trossen Robotics said the combined environment is intended to speed robot bring-up, data collection, skill development, evaluation and real-world validation. - The company said the onboard stack reduces systems engineering before teams begin collecting robot data and building applications.

Between the lines: - The announcement shows Trossen Robotics is betting that standardizing hardware and software around NVIDIA’s Jetson ecosystem will make physical AI development more portable across use cases. - The pitch is not just performance. It is workflow continuity, from first data capture to supervised field testing. - Processing at the edge can also reduce dependence on moving high-bandwidth sensor data to external infrastructure.

What's next: - Rivet and Workbench are available for preorder now. - Trossen Robotics said product configurations and additional information are available at the company’s website. - The company noted that product features, specifications, configurations and availability may change without notice. - Trossen Robotics also said actual AI performance may vary by workload, software, power mode, thermal conditions and system configuration.

The bottom line: - Trossen Robotics is making NVIDIA Jetson AGX Orin 64GB the default brain for both its mobile and benchtop physical AI platforms, aiming to simplify development and keep more robotics work on the robot.