Success Stories and Key Lessons from the Robotics Industry in Light of the COVID-19 Crisis

There was the usual mixture of wonderment and indignation when Boston Dynamics’ SpotMini was deployed by the Singaporean Government to monitor social distancing in public parks. The concerns of onlookers notwithstanding, the deployment is largely for show and not practical. The quadruped is teleoperated for a start, requiring a remote controller to move the machine about. Though somewhat antithetical to the notion of autonomy, teleoperation of mobile unmanned systems makes sense in some cases. For industrial inspection, remotely controlling SpotMini, or a more specialized inspection device from companies like Gecko Robotics, is worthwhile; humans can direct and interact with an environment without risking their own personal safety. In mining sites as well, teleoperation of haulage trucks from a central command station offers added redundancy over a full autonomy solution. While not fully autonomous, a remotely operated vehicle can still achieve major efficiencies over fully manned systems. One teleoperator can oversee multiple systems and has access to more information infrastructure. For this reason, it is much more reasonable to suggest that, in the medium term, unmanned trucks and taxis will be remotely operated by a new class of virtual drivers than to expect autonomous driving.

As for a surveillance robot in a public park, there needs to be a higher degree of autonomy, otherwise cost efficiency goes down significantly. A basic map route, as deployed in a warehouse or retail store, may not make sense in a security context. The SpotMini is deployed in Singapore for monitoring social distancing, but presumably such robots might be expected to monitor crime, loitering, or potential trespassing on private property. If so, there has to be a degree of unpredictability and dynamic movement to help capture information and respond effectively to alerts. This will require walk-through mapping of a specific environment and the uploading of routes, as well as the necessary sensors, either onboard or external, to allow the quadruped to navigate as pedestrians, congregate around it. Global Navigation Satellite System (GNSS) is sufficient for much of the navigation, but for added redundancy, Simultaneous Location and Mapping (SLAM) algorithms and radar, laser and cameras will be used.  

Attention also needs to be paid to the charging system needed for such a system to be practical. A battery swap is unlikely to be optimal as it means manually changing the power source for a quadruped. Doing this for a number of systems every eight hours would be time consuming and defeat the purpose of the robot in the first place. Opportunistic charging is therefore the better way to go. Lawncare robots like Husqvarna already have developed docking stations for opportunistic charging, but quadrupeds as a form factor may need a more sophisticated solution. Wireless charging for mobile robots is developing apace, with vendors like Wibotic spearheading the technology’s development.

The notion of a security quadruped is not fantastical, but as described, there is a lot needed to get it up and running. At least for the short term, their value will be most keenly felt in visual inspection for industrial sites, where current vendors like Zoa Robotics and ANYbotics are ploughing their resources.

The cobot market had already been negatively trending before COVID-19 hit in earnest in 2020. Universal Robots, the main supplier of collaborative robot arms with over 50% of the total market, saw its 1Q 2020 revenue slide 14% from the same period in 2019, from US$56 million to US$48 million. After years of solid growth, the cobot market can expect a very challenging 2020.

It is doubly worrying because Universal Robots is still king of the cobot space. This means other players are therefore much more likely to suffer. UR’s continual dominance was exemplified when system integrator Voith switched the partnership from Franka Emika to Universal Robots in late April. Franka’s arm has been plagued with issues relating to price increases and has not found the same popularity that UR’s has. ABI Research profiled both companies in our Industrial Collaborative Robot Competitive Assessment (CA-1268) in 2019 and, in line with the report’s projections, the industry’s gains are largely being enjoyed by UR, and a handful of other competitive players. The two most notable are Taiwan-based Techman Robot and South Korea-based Doosan Robotics. Both companies enjoy a strong geographic advantage—being in Asia, they have much more opportunity to scale and are located in and around economies where a much larger number of robots are sold, i.e., China, Japan, South Korea, and Singapore.

Another big concern for the cobot space is that it relies heavily and disproportionately on Small and Medium-sized Enterprizes (SMEs) for their revenue. While some larger companies like Jabil will deploy hundreds of cobots, they do not lend themselves to bulk orders or extensive solutions from larger vendors. Much of their business is from smaller vendors trying out one or two systems and adding more incrementally. Cobots are essentially designed to automate 80% of a certain task at a workstation (bin picking, machine tending) at 20% of the cost of a human employee. In a downturn, smaller vendors tend to halt funding and are more susceptible to fail, and while large industrial robot vendors can rely on big one-time orders, such as BMW’s recent purchase of 5,000 Kuka Robots, cobot vendors are not so lucky. Coronavirus hit just when the cobot space was consolidating and evolving from its embryonic stage, and the coming crisis will accelerate the main trends, including:

• More Direct-Sale and Services from the Vendor: Though relying on distributors for smaller customers, cobot vendors will aim to do more direct-sale purchases with larger end users, as well as develop service models to lower initial investment costs.
• Comprehensive Solutions: As represented by the recent announcement of the ActiNav picking system, companies like UR are packaging disparate components into comprehensive solutions for specific tasks like bin-picking and machine tending, essentially offering a turnkey solution to customers and moving up the value chain.
• Larger Payloads and Improved Hardware: Cobots need to be become more capable of handling larger payloads and higher performance in terms of speed. This is leading to the development of more seven-axis robots that have longer reach and can lift more, an example being UR’s 16 kg payload robot.

These trends will accelerate with the virus and are exactly what the fledgling technology needs to become more than an auxiliary subgroup of the established industrial robot offering.

Self-driving passenger cars as a viable reality were already out of immediate sight prior to the coronavirus. Now they are largely out of mind. Some companies, like Starsky Robotics, have already gone under. Meanwhile, some of the biggest funders of self-driving cars, like Softbank, are running into financial challenges themselves. A technology market that has no route to commercialization is about to face its first downturn, and there will be some casualties. That being said, a lot of the technology developed by these companies can be applied to other applications. There is a significant pipeline of talent when it comes to self-driving car developers moving into last-mile delivery robots (vis-à-vis Nuro) and with mobile robotics in general. As an example, Posmates teamed up with self-driving software developer Phantom Auto to develop the teleoperation for its last-mile delivery bots. It could well be that, far from going to waste, the money ploughed into self-driving cars will manifest in mobile robotics for almost all other use cases.

Autonomous robots for the last mile have been very widely advertised during the pandemic. Starship Technologies, which has close to 500 bots in operation, is approved in eight American states and is also helping supply food delivery to enclosed environments and particular city districts like Milton Keynes in London. The company offers to deliver goods from grocers and restaurants, and initially offers a free month before charging 20% of the delivery cost. Another company, Refraction AI, has deployed 20 systems and charges 15% of the order for its service, competitive with manual delivery services.

So how do these vendors operate? The robots are semi-autonomous, relying on Global Navigation Satellite System (GNSS) for positioning and a 4G network to move around locations. They are overseen by teleoperators who can intervene if and when there is an autonomy exception. Currently the teleoperation solutions in place are limited to one controller and one robot, but the aim will be to develop systems where one teleoperator can manage an entire fleet of robots.

While some robots, like those of Starship and Refraction AI, deploy on pavements and sidewalks, Nuro is notable for deploying autonomous road vehicles that carry a considerably larger payload. Nuro has gained significant attention for its enormous US$1 billion investment from Softbank, as well as its ability to get approval for testing and deployment of fully autonomous systems in states like Texas and California. The company has done some high-profile delivery operations for entities like the Mayo Clinic in light of the virus and has thus demonstrated the utility of self-driving delivery, but challenges remain. The company is operating some systems at field hospitals in California, and is limited to maximum outdoor speeds of 5 mph, while the vehicle’s technical maximum is 25 mph. What is more, for additional testing the autonomous vehicle is usually tracked by manned vehicles, a situation complicated by COVID-19. While the pipeline of activity shifting from self-driving passenger vehicles to last mile delivery exists, the latter still needs a lot of testing and regulatory overhaul before it can start scaling effectively.

In short, COVID-19 will see the expansion of various last-mile robotic fleets, because they have shown their value in light of social distancing and the pandemic response, but, for the most part, they will be confined to campuses and highly specific locations.