The Robotics Operating System (ROS): Key to New Robotics Development and Growth

In May 2024, Open Robotics—the non-profit and owner of the Robotics Operating System (ROS)—released Jazzy Jalisco, the 10th iteration of ROS 2. This release comes with a guaranteed 5-year support and maintenance commitment. In June, NVIDIA released the third version of its Isaac ROS, a dedicated ROS-based development platform for Isaac, now with tie-ins for Perceptor, Manipulator, and robot motion planning workloads. The continued, and growing, support for ROS bodes well for emerging developers across the automation landscape.

In fact, eager to gain a foothold in the inchoate and much anticipated “robotics boom,” chipset vendors are incorporating the ROS framework en masse: Qualcomm’s Robotics RB Development Kit platforms, NVIDIA’s Orin, powered by the Jetson series, and AMD’s Kria products are all dedicated mobile Robotics Development Kits (RDKs) with native ROS support. The objective is for product developers on the ground to utilize ROS, and the vast computational resources of the platforms provided, to rapidly develop and scale robotics offerings, generating demand for high-resource chipsets. Currently, a stop-off on the road to the “killer app” for these vendors is the Autonomous Mobile Robot (AMR) market, which is forecast to grow to nearly 3 million shipments annually by 2030. AMRs, generally dependent on Simultaneous Localization and Mapping (SLAM) for navigation, require a high level of compute for processing (warranting NVIDIA, Qualcomm, or AMD chipsets), while existing in a heterogeneous component market that generally necessitates the convenient communication paradigms of ROS for orchestration. For Intel’s part, the company previously maintained a dedicated GitHub repository that provided a RDK—a suite of open source algorithms—that could be deployed in conjunction with the RealSense camera; contributions to this repository ended in 2022. RealSense support for ROS continues, but a dedicated development platform linked to Intel Central Processing Units (CPUs) for mobile robots is conspicuously missing.

ROS was created at the famous Silicon Valley incubator, Willow Garage, to standardize mobile robot development—and it has been very successful. Close to 50% of AMRs currently deployed depend on ROS for some mission-critical functionality (examples include Clearpath Robotics, Fetch Robotics, Otto Motors, Vecna Robotics, and MiR). The reasons for ROS’s popularity are accessibility, speed, and permissible licensing. When a new mobile robot is developed, setting up communication between the various components, synchronizing sensor data, and writing hardware drivers are all labor-intensive tasks, consuming engineers’ time that would be better utilized building products. ROS removes this hardship by introducing an intuitive framework, cross-language compatibility, and data standardization. Anecdotal evidence indicates that using ROS can reduce development time from months to weeks or, in some instances, from years to months. Due to the framework’s popularity component manufacturers have found it advantageous to create (and make freely available) ROS drivers for their hardware to enable software engineers to easily incorporate components into their products. Examples of companies with ROS support for mobile robots include SICK, Basler, and Ouster for supporting their industrial camera and LiDAR portfolios, and, extensively, Intel with its RealSense Camera.

ROS’s popularity wanes in other verticals. Industrial robotics manufacturers have historically shirked open-source technologies, including ROS. Specialist Original Equipment Manufacturers (OEMs), including KUKA, FANUC, and ABB have well-established clients, Systems Integrators (SIs), and engineering teams with decades of experience in, and understanding of, the proprietary software ecosystems for automated industrial manufacturing processes. Although each of these companies have worked to create ROS-compatible drivers for their industrial robots, a development that can, in part, be attributed to the efforts of the ROS-Industrial Consortium, (the drivers can be found on GitHub), their real-world use is limited to academia, Research and Development (R&D), and emerging deployment cases such as palletization in the growing third-party logistics vertical. Further, industrial robots traditionally depend on rugged microcontrollers, rather than high-end Graphics Processing Units (GPUs) and computer hardware. If and when this will change is unclear; research interviews suggest that OEMs will develop any software innovation in-house, rather than leaning on open-source libraries.

The robotics landscape remains in flux. Mobile robotics could prove to be a volatile horse to hitch one’s wagon to. Although the market is forecast to be large, consolidation is inevitable. Equally, Collaborative Robots (cobots), humanoids, and drones may not “take off” in the way that proponents hope.

NVIDIA has hedged its bets in the robotics space: Isaac Perceptor will improve mobile and cobots; Isaac Manipulator will benefit collaborative robots and industrial arms by increasing their dexterity; Project GR00T targets humanoids by leveraging multi-modal generative AI; and Isaac Sim will enable simulated robot training and prototyping. For each of these tools, ROS provides the “software glue” that will accelerate developer uptake and the Isaac platform’s overall value.

• For Product Developers: Due to the proven time and cost savings, it would be counterproductive to attempt to develop a robot, or robotics-adjacent, product without ROS.
• For OEMs: Hardware manufacturers of all stripes would be missing out on easy gains by not developing ROS 2 drivers for their products, enabling developers on the ground easy adoption and integration.

NVIDIA understands that the value in robotics is not something conceived in board rooms or in the dreams of product owners. The value of robotics is determined organically in industry, where an experienced eye can identify utility and begin productization. Enabling organic development and growth for robotics, building where demand exists, is the key to uptake and value for all stakeholders. The ROS is the facilitator of that growth. NVIDIA via its universal adoption of the framework, value-adds, and pivotal role in governing bodies such as the Open Source Robotics Alliance (OSRA) may become its de facto manager and champion. Such an eventuality may help assuage concerns surrounding ROS’s open-source hangups, while growing the framework’s capabilities.