NVIDIA Sets Its Sights on New Corners of the Commercial Robotics Ecosystem |
NEWS
|
NVIDIA’s keynote announced the expansion of synthetic data generation in Isaac GR00T for training humanoid robots via teleoperation; Cosmos, a physics-grounded world foundation model; and Mega Omniverse Blueprint, a new fleet manager tied to the Omniverse simulator and Isaac Sim, indicating a serious further commitment to demonstrating the commercial viability of NVIDIA’s robotics ecosystem.
Mega Omniverse Blueprint is a significant development. The software, an extension of Omniverse, which is often used for industrial simulations, is being trialed in partnership with the KION Group. Mega is an Autonomous Mobile Robot (AMR) fleet manager coupled with NVIDIA Isaac Sim’s robot training arsenal and the new Cosmos foundation model. The fleet management space presents an opportunity for NVIDIA to demonstrate the value of Cosmos and reinforcement learning to optimize established processes within mature markets.
NVIDIA Cosmos, an open-source world foundation model, will allow AI models to better predict the behavior of objects and scenes via concepts such as object permanence, accurate geospatial information, and associating data with physical laws. Cosmos, coupled with Mega, could offer dynamic route optimization beyond the current simplistic path planning that defines the fleet management status quo—offering vast efficiency improvements. Google DeepMind also announced its planned development of world models in January. The extent to which world models will extend robot autonomy is unclear, but adding further complexity to robot training will undoubtedly increase computational overhead.
Could NVIDIA Unite the AMR Market? |
IMPACT
|
NVIDIA’s software-based robotics products, primarily centered around the Isaac platform, have, until this point, focused on the robot as an isolated system. For example, Isaac GR00T is a multimodal foundation model designed for training individual humanoid robots in various environments. Although this methodology is scalable—the training for one humanoid can quickly be copied to another—the focus is on an individual machine, rather than a process or workflow. A sizable portion of value can be captured by the optimization of a large workflow, rather than one step of a process in isolation.
NVIDIA wading into the commercial fleet management and orchestration space is a significant development. Fleet management has been an innovative and growing market for some time with an increase in third-party Software-as-a-Service (SaaS) vendors and large robotics groups pushing to incorporate interoperability across robot brands. Third-party fleet management companies, often referred to as Cloud Robotics or RoboOps (a term coined by InOrbit), include Formant, Cogniteam, InOrbit, and Brain Corp. The Go-to-Market (GTM) strategy for these firms has been to partner with Original Equipment Manufacturers (OEMs) and System Integrators (SIs) to offer products that streamline the deployment of AMRs and reduce the burden of management and maintenance on end users. Product features often include interoperability (frequently via VDA 5050); teleoperation; machine vision value adds (notably Brain Corp); predictive maintenance; interchangeable Simultaneous Location and Mapping (SLAM) algorithms; and dynamic path planning. Several robotics groups—including MassRobotics, Open Robotics, and OpenRMF—advocate for general AMR interoperability via their own, or common, communication protocols. Despite the best efforts of these groups and companies, methodologies remain fragmented with no dominant technology or framework for AMR orchestration. Most OEMs develop their own fleet managers, which SIs have become accustomed to.
The value that NVIDIA can offer for fleet orchestration via Cosmos and the Isaac portfolio could be vast. If Return on Investment (ROI) can be demonstrated, fleet management providers—all of which are already partnered with NVIDIA—will buy into Mega, sharing positional data for paths to be optimized in real time (analogous to the popularity of the Universal Scene Description (USD) format). This would allow NVIDIA to create de facto interoperability and provide advanced optimization across all AMR brands.
NVIDIA Must Leverage Its Presence to Capture the Orchestration Space |
RECOMMENDATIONS
|
Other tech behemoths are trying to conquer the AMR orchestration space, notably the hyperscaler Amazon Web Services (AWS) with its RoboRunner product, a key feature of which is interoperability. However, RoboRunner appears to have achieved minimal market penetration, to date. NVIDIA possesses several advantages that AWS lacks. First, many AMRs already depend on NVIDIA Graphics Processing Units (GPUs) to perform demanding real-time workloads, including SLAM. The Robot Operating System (ROS)-native architecture behind Isaac and Mega Omniverse Blueprint will be a major draw for development teams that already use ROS within their robot product architecture. Further, NVIDIA has active partnerships with all the fleet management companies listed above, providing an immediate channel to market for orchestration innovation. Finally, by using the USD data format, NVIDIA has achieved Omniverse buy-ins from most major industrial simulation products, including Siemens’ Process Simulate, Dassault Systèmes’ DELMIA, and Emulate3D from Rockwell Automation—opening manufacturing verticals for Cosmos and Mega to provide optimization.
Dominating both edge hardware and cloud-based orchestration services, NVIDIA is well positioned to offer performance gains to AMR service providers within all market verticals—from warehousing and manufacturing to the emerging service, healthcare, retail, and agriculture markets. Importantly, NVIDIA is already embedded within the startups and Small and Medium Enterprises (SMEs), hoping to capitalize on automation opportunities in these spaces. Gaining a foothold within the warehousing sector may be harder. NVIDIA ought to develop products that appeal to both the SIs that service AMRs within logistics and warehousing and the Warehouse Management System (WMS) providers that dominate this ecosystem. Ultimately, Cosmos and Mega must demonstrate efficiency improvements beyond what is currently available. Impressive digital twins—which is the current value offered by Omniverse and the USD file format—will not be enough to disrupt and drive demand within the fleet management market.
