Finding cost-effective insurance coverage can be an unpredictable process for some robotics developers and start-ups.
- The insurance industry is only starting to engage with the next generation of robotics technology
- Not enough robots are sold to provide the critical mass of data that insurers need
- Lack of regulatory and legislative guidelines complicates matters for both robot makers and insurers
Scott Eckert gets hit a lot by robots
Eckert isn’t a stunt double, however. He’s CEO of Rethink Robotics, the Boston-based robotics company that created Baxter, a low-cost industrial robot specially designed for use in close proximity to human workers.
And neither Eckert, nor Baxter, has a serious malfunction: It’s all part of a routine designed to demonstrate Baxter’s safety features, which include 360-degree sonar and front-camera for human presence detection and collision detection and avoidance mechanisms.
“I will routinely have the robot simply hit me in the middle of performing a task,” explained Eckert. “The robot’s arm will hit me while it’s working, just to demonstrate that it’s really not a big deal when that happens.”
“And when the robot hits me, you see both that it isn’t a big deal, and the robot will respond smartly, if that were ever inadvertently to happen,” he added. “Baxter is self-evidently safe.”
Baxter is one of a new breed of industrial robot: cheap, relatively small and lightweight (compared to the majority of industrial robots), and easy to program.
Most importantly, Baxter is designed to work alongside human workers. The safety cages and yellow warning tape that typically separate human and robot workers from each other are no longer required.
Robot-style dog and pony show
Eckert and Baxter went through their routine when insurance company representatives came to see the robot in real-world action.
“Once you see the way Baxter works, the robot’s inherent safety becomes quite clear,” said Eckert. “The insurance agents could immediately see that there are multiple aspects of the design –the way the robot moves, it’s inherent compliance, the collision detection, the human presence detection– that make it very intuitive to deploy Baxter in a factory without safety cages or safety guarding.”
As a result, at least in part, of that demonstration, he noted, Rethink Robotics was able to secure standard product liability insurance protection “avoiding the potential extra expense and complication associated with customized, specialty insurance packages.”
“We needed nothing beyond the normal sets of insurance you would expect a company of our size to have,” Eckert said. “Nor have any customers ever asked or raised the issue of any insurance that they might need to carry. So, insurance has not emerged as an issue for Rethink Robotics or for our customers.”
For those, like Baxter, that qualify, insurance companies offer a range of off-the-shelf products aimed at commercial robotics manufacturers. These typically include workers compensation, products liability, errors and omissions liability, contingent business interruption, R&D property and umbrella excess insurance.
Insurance denied, project abandoned
Not all robot developers fare so well. One Europe-based robotics researcher who preferred to remain anonymous told Robotics Business Review that a pilot project designed to test an autonomous humanoid robot in an urban environment had to be abandoned when affordable insurance coverage could not be found.
With so many unknowns at play –What range and type of risk might a malfunctioning autonomous robot pose to property and persons in an urban setting? — insurance companies could only offer protection via costly specialty insurance.
What is already an unpredictable landscape for robotics companies is set to become even more difficult to navigate as more autonomous and semi-autonomous robots enter our workplaces and homes.
Despite the best efforts of robot designers and safety experts, the greater freedom and flexibility enjoyed by these machines will present numerous challenges for insurers.
Unfortunately, the insurance industry in general doesn’t yet understand robotics well enough to provide insurance for next-gen robotics systems at a competitive price, but that’s starting to change, said Robert Wenzel, Director of London Global Laboratories (LGL), a consultancy that specializes in commercializing robotics R&D.
Traditionally, robots have been deployed in controlled, industrial environments, said Wenzel, which made it relatively easy to reduce risk by restricting human access to the robot. The latest developments in robotics show a clear trend towards robotic co-workers and robots for personal care, however. “These systems have close human-robot interactions and the robotic limbs could severely harm users,” he said.
ASIMO draws a blank
“A malfunction could happen relatively fast,” observed Wenzel, citing the recent incident at a press event where ASIMO’s sensors could not distinguish smartphones from hands as an example of how a robot’s AI can malfunction in unpredictable environments.
“Now is the time for the insurance industry to seriously discuss these issues and in the past year or two, driven to some extent by autonomous car research, that is beginning to happen,” he said.
As if to enforce Wenzel’s point, Swiss Re, the global reinsurance giant, recently hosted a conference on human enhancement technologies at its Centre for Global Dialogue in Switzerland.
However, the company also declined our request for an interview, telling us that Swiss Re is at an “early stage of evaluating the risk landscape” around robotics.
Similarly, The Society of Actuaries, with 22,000 actuarial members in the U.S., Canada and worldwide, was also unable to provide a source for an interview about robot risk and insurance issues.
There is no question that insuring certain robot applications will be more challenging than others, especially as robots become a ubiquitous technology and customers and end-users become more dependent on robots, said George Weimer, Assistant VP and Senior Underwriting Specialist for the technology insurance specialty for the Chubb Group of insurance companies
“As dependence on robots in every day applications increases, there will be greater expectations in terms of safety, performance and reliability,” said Weimer. “Specialty insurers that have been providing products to dynamic, cutting edge industries for many years will be in the best position to support the robotics industry with innovative and tailored insurance solutions.”
Robotics is a very diverse industry segment with a growing range of new commercial, industrial, and consumer applications that pose unique challenges for insurance companies, Weimer acknowledged.
“Even if a robot performs as intended, lawsuits may result from allegations that the robot caused injury or damage, which may, in fact, have been caused by the end-user’s improper use of the robot,” he said.
The level of insurance coverage a robotics company will require and the need for specialized protection depends, at least in part, on the specific end-use application.
“Standard and specialty insurance products are currently available from several insurers that address many of the risks faced by developers of robotic products,” explained Weimer. “However, the risk appetite and willingness to offer certain types of insurance for a robotics firm will vary greatly among insurers depending on the specific end-use of the robot product.”
For example, a company developing a robotic product for surgical applications may require a specialty insurance product to cover the medical trial phase, while a relatively standard insurance product may be sufficient for the commercially released version.
Part of the reason the insurance landscape is so unpredictable, is because insurers, reinsurers, and robotics developers don’t have robot-specific legislation and regulations to work within, says LGL’s Wenzel.
“Government is always much, much slower than technology. But, as an entrepreneur, you have to take the risk, develop your system, and bring it to market,” said Wenzel. “The Internet started this way. E-commerce started in this way. And I am convinced that robotics will start this way too.”
If robotics industry standards were legally enforced and the insurance industry set standards which robotics companies had to meet to qualify for specific insurance products, then robot developers might be better placed to avoid the cost of specialty insurance.
The kind of certainty that insurance companies look for comes from legislation and case law, with legislation being “much preferable,” said Donald Light, director, Americas Property/Casualty Practice at research and consulting firm Celent.
“Case law has to accumulate and doesn’t necessarily have any impact from jurisdiction to jurisdiction, so it’s a very slow and awkward process,” he said. “If there’s economic pressure to create enough certainty it will express itself through legislation, I think, not through case law.”
Insurance companies also seek security through data. In a relatively niche market like robotics, that data simply isn’t available.
In probability theory, the Law of Large Numbers (LLN) states that every time an action with an random outcome is performed (such as flipping a coin), the overall average moves closer to the expected value (a 50-50 split between heads and tails).
The LLN is an invaluable tool for insurance companies as it enables them to understand the nature of risk over time. Unfortunately, for next-gen robotics companies, the law only applies where a large number of observations is possible.
“Insurance companies like the LLN,” explained Light. “The nature of the insurance mechanism is that if you have a lot of similar risks then you can understand the nature of that risk and the probabilities of losses occurring. Then you can make intelligent pricing decisions so that you can pay your claims and still make enough money to operate and make a little bit of a profit. That’s the insurance financial model.”
The problem is that there aren’t a lot of robots being sold and used, and so the LLN does not apply. As a result, robotics remains “a difficult risk for insurance companies to understand.”
In terms of measuring risk via the LLN, a robotics manufacturer that sells 300 robots in a year is effectively the same as a manufacturer that sells one robot per year, says Light. And the result is likely to be the same: a costly, specialty insurance policy that’s negotiated on a one-off basis.
“But if there’s a dozen robot manufacturers selling 3,000 robots a year, then we’re getting into the LLN territory where some insurance companies will develop standard policies and even start competing against each other on price,” said Light.
For that stage to be reached however, the robotics market needs to grow and mature so that the data available to insurers reaches LLN proportions.
Decision-making autonomous systems
The real challenges will come with fully autonomous systems, in which the decision-making role traditionally performed by humans is taken over by machines, says LGL’s Wenzel.
“In today’s car insurance, humans are the common decision-making factor,” according to Wenzel. “Today’s car insurance products place a value on the quality of human decision making when driving a car. In the future, they will have to place a value on the decision making of the car itself.”
This will also require insurance companies to take into account the robot’s sensors, actuators, and safety systems. And since these components vary so much from product to product, says Wenzel, insurance companies have to assess risk on a per product basis, which almost invariably leads to costly specialty insurance coverage.
And in the absence of robot-specific legislation and regulation, and the LLN in force, this situation unlikely to change anytime soon.
So, what can robotics developers, startups, and others, do to improve their chances of successfully navigating this unpredictable landscape? Find out in Part II, where expert sources in robotics and insurance share their tips and advice.