Autonomy Solution Providers Are Leading the Way for a New Robotics Industry Based on Hardware-Agnostic OS for Mixed Fleets

For the past thirty years, the robotics industry has been defined by a large number of different providers building similar products with proprietary operating systems, controllers, and software to perform largely similar tasks. This applies particularly to mobile robotics, with different providers offering separate form factors, sensor solutions, use cases, and highly differentiated hardware offerings.

This might be called the Artisanal Age for robotics and was understandable given the significant challenges faced in developing robots for different use cases. But as technology has matured and deployments have become more common, hardware costs have gone down, and the corresponding drop in sensor prices has made it less difficult to build sophisticated mobile robots at scale. Now, the value is shifting to the middleware and software that can power these robots and develop the advanced value-added services that enable and extend the possibilities of robotic adoption. Ultimately, these companies will extend their offerings beyond different hardware and will thus have a twofold effect on the robotics ind

1. Companies will harmonize and help establish interoperability and standardization as the norm for robotics. This will greatly expand the capacity of end users to adopt robots. Given the increased flexibility and cloud capacity of these providers, Robotics as a Service (RaaS) will become the norm among all new verticals, with standard system integration, distribution, and direct-sale chains enduring in mature markets.
2. Companies will shift value from the ability to build robots to the ability to deploy solutions at scale, with value concentrated in management, analytics, operations, and monitoring. Their success or failure will depend more and more on integrating multiple robots and use cases (drones, unmanned ground systems, underwater vehicles etc.) with more general Internet of Things (IoT) offerings like smart wearables, connected infrastructure, and the smart workforce.

In essence, these companies are not developing vertical innovation that greatly expands the capabilities of one robot. At the moment, mobile manipulation, improved mechatronics, flexibility, and so on will still be spearheaded by individual manufacturers, Original Equipment Manufacturers (OEMs), universities, and government departments. The value for these companies is the economies of scale they will reap as they spread, and in turn the wider distribution of cloud computing and value-added services will be a broad-based force multiplier for robotics, accelerating development to paces previously not yet seen.

A number of companies are aiding this need and might be broken down as such:

• Cloud Service Providers: These general-purpose platforms are offering general purposes services, Software as a Service (SaaS), and the ability to scale operations effectively. Microsoft Azure and Amazon Web Services (AWS) Robomaker are the leaders in this space.
• Robotics Software Platform: A Robotics platform that focuses on management, analytics, monitoring, and so on. They are often cloud-connected and can be plugged into more comprehensive OS solutions.
• Autonomy Solution Provider (ASP): An ASP differs from simple robotic platforms in that they offer fairly comprehensive technology stacks that provide all the traits of autonomy to robots, standard vehicles and equipment.
• Robotics Manufacturers: Many of the companies are developing in-house software for their systems. They are limited by their own hardware. An example might be IAM Robotics.

ASPs represent an increasingly important part of the commercial robotics market, as the dividends of improved mobility-related technologies proliferates. They are in a particularly strong position because they sit at the intersection of hardware development, software, and implementation. They are thus better attuned to the specific expertise of robotics design and development and have experience in integrating robots into workspaces. In many cases they will work with system integrators and distributors to expand their operations and are thus closer to the traditional robotics industry than cloud service providers and software platforms.

Among the most well-known of these companies, Brain Corp, has rolled out 2,000 autonomous vehicles and has extended its operations from America to the Netherlands with the recent opening of a distribution center. A Swiss provider, BlueBotics, has deployed over 1,500 robots, while legacy industrial giant Kollmorgen provides varying degrees of navigation for different Automated Guided Vehicles (AGVs) and standard vehicles. They have a partnership with Italian robotics behemoth Comau and have deployed some 15,000 systems worldwide. As a legacy company, Kollmorgen does not provide an enormous analytics or software product line.

A new generation of companies is taking advantage of this space and trying to be even more flexible than the current leaders. Stocked Robotics, for example, is offering to retrofit vehicles in a short space of time and for a limited price and are contemplating making the hardware free and relying on revenue from their analytics and value-added services instead. Currently, Stocked is defining its value as superior to other ASPs in that they can offer a simple retrofit to a wide range of electric forklifts without significant design or engineering work. Arguing that the average robotic forklift costs upward of US$150,000–US$175,000 per unit, Stocked offers to automate any vehicle for US$50,000 with its Sierra autonomy platform.

Scenarios for the future of robotics have increasingly been described through the metaphor of smartphones. The analogy is tempting, given the differing philosophies of the two primary Operating System (OS) providers for the mobile space—Apple iOS and Google Android.

Apple represents the far end of the closed proprietary system. They do not have open-source offerings and specifically tailor their OS to their own hardware. As a result, their potential for short-term improvements and continual crowd-sourced updates is limited, but this is offset by improvements in performance through optimization.

On the other side is Google Android, which relies heavily on Linux Kernel. With certain licensing caveats, it is open to a wide range of hardware vendors and therefore has the benefits associated with a more permissive and license-based system. While the core OS for Android is free for all phone manufacturers to deploy, they have to enter a licensing agreement with Google to have access to Google services, such as Google Maps and other applications.

Budding ASPs are definitely trying to follow the route showcased by Android, where a Linux-based system may have some open-source features but where the most valuable offerings, like path-planning, navigation and apps are commercially. The shipments for Android-enabled phones from Samsung, LG, Google, and Huawei exceed Apple’s shipments by an enormous order of magnitude (see Figure 1).

(AOSP refers to Android Open Source Project)

Given that robots are ultimately commercial systems and not general purpose tools like smartphones, it is not possible to run exactly the same business model. Robots are filled with high-end sensors and will always represent extremely complex systems in comparison to an average smartphone. In fact, they are increasingly expected to perform complex tasks in groups, making scaling them significantly more challenging than smartphones. Simply developing a strong technology stack and licensing it across different robotics developers is only part of the challenge. Deployment, maintenance, and monitoring are huge strains on the ASP, and given the desire of many new end users to shift to subscriptions, the capital cost for all this responsibility falls on them.

So it is not quite as simple as iOS versus Android. What is more likely is that Robot Operating System (ROS) futures will trail those of smart wearables. Wearables are similar to smartphones but are more varied and often more complex, especially when relating to medical devices (see Figure 2). Closed proprietary OSs will do very well for specific robotic applications, maybe for piece picking or for mobile manipulation, but will not represent the majority of the market. Android has a strong market share for wearable OS but nowhere near the same dominance seen in phones, and a large percentage of the market is built on individual proprietary systems (categorized under “other”). An increasingly popular category is Linux-based OS that is commercially leased (similar but distinct from Android) to offer a flexible system that is also ruggedized. This represents something similar to Brain Corp’s Brain OS and will form the basis of the future mobile ROS landscape.

Division of labor via the development of ROS-based ASPs offering common navigation solutions will be supplemented by specialization through the development of best-in-class software and DevOps tools that focus on specific use cases like monitoring, diagnostics, or fleet management.

What industry thinkers, and especially ASPs, now need to consider is how and where the technology stack gets monetized. Just providing navigation, localization, or mapping is not particularly profitable even at scale—it only provides a strong market position. Indeed, even developing the ability for the robot to locate is one of the least valuable and most difficult parts of the engineering process for any robotics company. The fact this had to be repeated ad nauseum in the development of individual, atomized robotic systems could be viewed as a bad opportunity cost; but with this largely solved, attention will move up the technology stack, as shown below in Figure 3.

This opens up the field for cloud-service providers and software providers, and so ASPs, who also have to provide the middleware and autonomy enablement, are going to have to partner with, acquire, and offer Application Programming Interfaces (APIs) for bespoke solutions that the customer will want.

This is the beginning of a sea change in robotics, amounting to a fundamental division of labor that will open up the market to diversification. An example is the cloud robotics vendor InOrbit, who provides development operations for robotics deployments and specializes in visualization, fleet management, data collection, and aggregation. They work with ASPs like Brain Corp and offer some the highest value services by providing a common platform for heterogenous fleets. Given the need for scalability, many robots of differing types will develop across the workspace and require the deployment of superior software services to manage.

At the same time, the prevalence of ASPs and the development of specialized software companies like InOrbit will allow for more diversification in hardware and use cases as integration become easier. As noted by Figure 3, there is a large but narrowing area of potential long-term value in services, apps, and data. While there will be an explosion in software providers to offer this, consolidation will be the likely outcome. ASPs are well placed to start this because they control the fleets and have a head start in developing their own software offerings.

Ultimately the new age of robotics will be defined by

1. The success of a few ROS-influenced OSs that can automate lots of different pieces of hardware and provide general solutions to common problems, like navigation, odometry, and so on. Robotics companies will try to develop ASP-like capability by licensing their software.
2. The proliferation of software vendors that build best-in-class solutions to plug into OSs. These players will be both competitive and complimentary to ASPs and will be based on partnerships and acquisitions.
3. Specializationin both software and hardware, particularly for hard use cases (quadrupeds, underwater drones, mobile manipulation).
4. Reactive consolidation being pushed by cloud service providers. There will also be attempts by autonomous car developers like Cruise and Nuro to develop cross-industry solutions for robots.

As ASPs move forward and preempt these developments, they stand in a very strong position to develop as the central family of platforms for the mobile robotics ecosystem.