From the legal perspective, Andrea Bertolini shared the many ways legislators might regulate autonomous tractors. Christophe Gossard from John Deere GmbH & Co. KG focused his presentation primarily on risk assessment and safety and security concerns.
In the roundtable that followed, Antoine Chatelain (nursery employee and robot operator at Chatelain Nursery), Aymeric Barthes (CEO of Naïo Technologies), Gordon Clements (General Manager Solutions for VARTA Ag) and Greg Meyers (CIO/CDO for Syngenta) discussed the various limitations that needed to be overcome for driverless tractors to be developed, deployed and, eventually, adopted worldwide. This section looks at each of these viewpoints in greater detail.
Civil Liability, Time Machines and the Case for Insurance
With the implementation of any new technology, there are unprecedented risks. In this way, driverless tractors are no different from weeding robots, robotic vacuums, or long ago, the first automobiles. Whenever a machine causes harm to an operator, there must be a legal means for compensating the victim.
“Civil liability is basically defined as the set of rules that determine who is bound to pay damages to compensate the victim when the victim suffers harm,” says Andrea Bertolini, a lawyer specializing in robotics. “At European level, typically liability, and civil liability in particular, requires fault. Basically, that the party that is being held responsible can be considered negligent or violating a rule of law.”
With respect to robots and advancing technologies, fault can be considered in a few different ways. The first is based on whether the machines themselves can be held responsible for causing harm. According to Bertolini, there is some debate about whether advancing technologies that can “learn” like robotics and artificial intelligence applications can be deemed agents, and therefore, held accountable for causing harm. He cites a provocative article from years past that posited even a smart thermostat that automatically adjusts the temperature in a room could be considered an agent and held liable for harm.
Bertolini is not convinced by such proposals. Because today’s machines are not sophisticated enough to be agents. They have neither the skills nor the abilities to cause harm. Under the current legal framework, he says, the programmer or producer of the applications would more likely be held responsible.
The issue became slightly more complicated when in 2017, the European Parliament considered a proposal about whether to attribute “electronic personhood” to advanced machines.
“Some people understood this to be the acknowledgment that machines are agents and therefore self-responsible, but that was actually not the case,” Bertolini says. “What our European Parliament was proposing, among other alternatives, was to attribute personhood like it does to corporations also to some kinds of advanced machines. That would be a purely instrumental concept.”
Since that proposal, there has not be much follow-up, and advancing technologies do not have electronic personhood according to the existing legal framework. Bertolini proposes that driverless tractors and similar machines be considered in a different context: the words of Professor Paolo Grossi. The Italian law professor and Italian Constitutional Court judge famously said, “the law is not a brick that falls on our heads; the law is all around us.”
“This is to say that if overnight, somebody invented the time machine to undergo time travel, that would immediately and already be regulated,” Bertolini says. “Meaning, that if somebody suffered harm while using the time machine, even if there is no time-machine-specific regulation, those that got harmed using the time machine could sue the producer of the time machine in the court, and the judge would have to find a solution to that case with existing norms that are already available.
“This also allows me to say that it is not useful to debate whether or not we want technology to be regulated because technology is already regulated the moment it comes into existence,” he continues. “Rather, we are only debating what kind of legislation we want for it.”
Today, all machines fall under two different liability norms: product safety regulation and product liability directive. Product safety regulation, Bertolini explains, is “that body of requirements that are set through primary legislation at the European level to determine what parameters a given application needs to meet in order to be considered a safe and also to be then sold legally on the European market.”
Advancing technology suppliers must comply with these standards to sell their products on the market, but these parameters do not limit the company’s liability. That’s where product liability directives comes into play. These directives aim at two things: ensuring product safety and ensuring victims who suffer harm are compensated. When the user encounters harm, the producer of the good is strictly liable. The victim does not need to show negligence, only that the product was defective. This is tougher than it sounds.
“There are grounds to doubt that, at European level, this body of legislation is functioning properly as it as it should,” Bertolini says. “There are a number of studies showing that there is a very limited litigation that allows the application of those norms. It's very hard for claimants to demonstrate that the product was actually defective and therefore that the producer is obliged to compensate, and this causes a lot of concerns when it comes to advanced technologies.”
When victims have little recourse against the products that harm them, there are fewer early adopters of new technologies.
“It seems to be a useful thing in a producer's perspective to say, ‘if it's harder for victims to sue us in a court, we will be favored somewhat,’ but that is not actually the case,” Bertolini says. “I did studies to show that this will slow the uptake of technology in many cases because if users do not feel safe around those kinds of applications, they might not want to switch to the more autonomous versions. They would rather keep the more familiar technologies.”
Instead of relying on product liability directives, Bertolini suggests a more practical risk management approach. Take the example of driverless cars, as they have similar characteristics to autonomous tractors and other advancing technologies. If the producer of the technology was held “absolutely liable under all circumstances,” Bertolini says, “he would acquire insurance.”
The cost of that insurance would then be transferred onto the product’s users who would pay for this insurance as part of the cost of the product. This would result in compensation without litigation for any potential victims. The costs get distributed along the value chain through contractual agreements.
In the case of the driverless car then, this evens the playing field between the advanced technology and a traditional car. Now, users are more likely to choose the driverless option than if there weren’t a framework for addressing liability and damages.
While this is a good option for addressing harm within the current legal framework, Bertolini believes insurance has its limitations. There is more work to be done.
For one thing, there has to be a market for this specific type of insurance, as driver’s or homeowner’s insurance is unlikely to address the unique problems created by advancing technologies.
“I do believe that all robotic applications are different from one another in that you cannot really find a one size fits all solution like the European Parliament is attempting to do,” Bertolini says.
For example, agriculture robots generally exist within segregated environments, away from the general public. The driverless tractors will not need to account for the erratic behavior of other drivers, as a driverless car would.
“This is an element that should be taken into account when considering a regulation for agricultural or robots,” Bertolini says. “It’s always a good to take into account the specific technological features of each machine. I always have this bottom-up approach, and I do really think that it is necessary for conceiving new legislation. So, even when it comes to insurance, it is a tool to address some of the liability concerns, but it is not always the best tool.”
Safety, Security and the Journey to Market
From the manufacturer’s perspective, John Deere’s Christophe Gossard presented the two types of challenges that needed to be addressed before driverless tractors become a real-world possibility. The first is managing safety and security; the second is ensuring the legal requirements to get autonomous machines onto the market are not cumbersome.
When it comes to product safety in Europe, Gossard says, risk assessment is done through the manufacturer. When there is no standard available (especially for newer technologies), there is likely an international or industry standard that can be applied. This is very different from places like the U.S. where risk assessment is conducted in court through a team of lawyers.
For new technology risk assessment, the process is supported by the original equipment manufacturer (OEM). The OEM’s job, Gossard explains, is to validate the conformity of equipment that can be used in the fields and may potentially circulate on the public network.
In his presentation slides, Gossard notes that manufacturers can use European and international standards to allow proper risk assessments under the future regulation for machineries replacing the current Machinery Directive (2006/42/EC). From there, manufacturers can integrate new technologies (IoT devices, AI/ML, cyber-security, autonomous features, etc.) into the Essential Health and Safety Requirements, while maintaining high safety and security requirements, and protecting the OEM against potential litigations.
When it comes to connectivity, Gossard contends that there needs to be a robust architecture based on a Permissioned Distributed Ledger (PDL)/private blockchain. A PDL combined with a certification process will provide trust to the end user, he says, while ensuring openness to interconnect other equipment completing the eco-system required to make run an autonomous vehicle.
The new legislative framework and declaration of conformity are very important for an industry that wants to deploy autonomous vehicles, Gossard says. There are benefits, Gossard explains, including self-certification, removing market barriers for the industry and SMEs, cutting unnecessary costs and avoiding a disproportionate administrative burden for the manufacturer. He believes that it doesn’t make sense for small operations to have a full legal team to deal with any issues that arise from advancing technologies.
As the industry gets closer to bringing driverless tractors onto the market, new challenges will arise. One of the major issues around advancing technologies centers on data. The sovereignty of the data is really at the core of the discussion right now for the European Commission, Gossard says, and there is a lot of discussion from Germany and France about the architecture of standards for developing new products and new technology, especially the data created at the sensor level.
No one knows what will come of these discussions. Gossard, however, prefers to focus on the future, dealing with the issues as they arise.
“My company used to say we need to have the feet on the ground and the eye on the horizon,” he says. “The tractor of today requires a lot of time, a lot of hours invested, and very little value is provided to the farmer. What we need to survive is for the farmers to have solutions for the future.”
It's Possible if Farmers Can Rely on the Machines
At a family-owned farm, nursery and garden center in Le Thillay, France, 18-year-old Antoine Chatelain has taken the lead on bringing robotics into the mix. Chatelain Nursery belongs to Antoine’s father, Laurent, but the younger Chatelain manages the technology. He sees driverless tractors as another means of improving farm work.
“There are tractors already using GPS out in the field, but what we are seeing is that not all tasks can be automated with current technologies,” Chatelain says. “I think that artificial intelligence and other systems can greatly improve what we can do, and I think, in the end, we will only be required to do only part of the job. The other part of the job will be reduced greatly by AI, autonomous tractors and other vehicles.”
Specifically, Chatelain envisions a future where advancing technologies help enhance decision making. The farm will never fully drive itself, he says. Instead, the value proposition of AI will continue to be performing intensive, repetitive tasks, creating space for new robotics-related jobs to develop.
If the current generation of agricultural robotics is any indication, the technology needs to be finessed before a driverless tractor becomes a viable option. From Chatelain’s experience, the biggest issue is reliability.
“Sometimes the robot works great, and we don’t see any issues with the work,” he says. “The robot does its job, and it is a great success. But at a moment’s notice, the machine can hurt your crops. You may have some issues that can attributed, in part, to wrong settings being entered or environmental variables that are not in kind with what the robot can do.”
Chatelain says the cost of ownership can also be preventative. Once purchased, however, the systems require little human supervision and are very inexpensive to run.
“For us, driverless tractors and robots are really about reliability and the trust we can put in those systems,” he says.
It’s Possible if Robot Manufacturers Can Deliver Proper Training
Naïo Technologies CEO Aymeric Barthes believes agricultural robots are already making a massive impact on the farming industry. There are already an abundance of robots that work on vegetables and vines. It’s only a matter of time, he says, before driverless tractors disrupt the status quo even more.
“I think the most important thing we have to keep in mind is safety,” he says, adding that this is where proper training comes into play.
“When using robots, you need to really train a lot with the robots to understand how they work. to understand how to integrate the robot in the fields, how to set up the robot in the right configuration,” Barthes says. “We have technical support integration during the time when you use the robot, so we have a lot of things to do too, and the training is really key.”
This means not only training the end user, but also the distributors. When customers understand the technologies, they are more likely to use them and spread the word. Distributors that know how to work the robots operate with a more powerful means of selling the product and alleviating potential fears.
On the issue of safety, driverless tractors have a seemingly quicker pathway to the market than, say, fully autonomous cars. Driverless cars operate in the public sphere. The biggest benefit farms provide on the road to implementation and adoption is privacy.
“Agricultural robots work in fields away from public areas,” Barthes says. “When it’s possible, the farmer can help create a safe environment by putting barriers and signage around the areas where autonomous machines will be operating. This is very different from driverless cars because you cannot privatize the public area.”
In addition to managing potential safety concerns, Naïo Technologies, the innovator behind the Oz, Ted and Dino weeding robots, focuses on profitability. Robots, Barthes says, must be profitable in order to succeed. One way that agricultural machines can increase their value is by addressing the power issues. Driverless tractors will need to address battery-related problems, too, and that’s precisely why companies like Naïo Technologies are already thinking ahead.
It's Possible if the Technology is Robust
“When you have batteries on the robot, you need to charge the batteries, and if you have to, you have to send someone to charge the batteries, you have a problem,” Barthes says. “You have an issue with the profitability of the model. We need to make that aspect of the agricultural robot autonomous like the vacuum robot in the house. So, we designed a trailer with solar panels to charge the robots autonomously.”
As the General Manager Solutions for VARTA Ag, Gordon Clements cites batteries and power as examples of the important role technology plays in developing effective autonomous machines. After all, the batteries that power autonomous tractors and other agricultural robots need to be durable, intelligent and properly optimized.
“From a power perspective, farming without a tractor driver is already possible,” Clements says. “The challenge is to consider the battery and also the charger and the availability of charge.
“The notion of charging a battery is probably the key to success because it's not only in the robot that we need artificial intelligence and machine learning,” he continues. “The big challenge is really combining the charger and the battery together and bringing artificial intelligence and machine learning directly to the battery, so that we can optimize the battery and the charger as we go forward.”
Currently, VARTA provides two types of solutions. The first is a custom battery that is designed especially for a client that is building a robot with specific dimensions and power needs. This battery will be unique, expensive, and sold exclusively to that client. For startups, VARTA offers a less expensive solution. Option two is an off-the-shelf model that is a simple, reliable, and safe power supply.
With batteries, safety is a key concern from the beginning. All of the materials that go into making a battery have safety standards they must pass. Then, Clements says, the components must be put together in a safe way. An automated production line is used to minimize deviations in the product quality and ensure the battery can operate within its safe window.
“The battery, especially in the application of agricultural robotics, is subjected to shock vibration, pretty harsh environmental conditions, in terms of the weather, rain, direct sunshine,” Clements says. “All of these things need to be considered in terms of the packaging and how the battery is actually housed, and provided you get all of these things correct and your vendors able to deliver these things to you, then the safety issues can be managed in such a way that it represents a very small risk to the end user, if any risk at all.”
When a battery needs to be replaced, the goal is to replace the modules instead of the batteries themselves. This, Clements says, will enable the farmers to safely change the power sources on the farm without expert assistance. An even better option is to use wireless changing solution. VARTA recently created one with power supply company IN2Power.
“It's much more relaxed,” Clements says. “You can manage the charging regime, and if we can learn what the power consumption looks like and what the profile of the power consumption looks like and what the charging regime looks like, we can basically collect that data, and we can optimize them for each individual application or even for each individual farmer.”
It’s also essential for the batteries and the robots to speak to one another. Not all agricultural robots nor driverless tractors willl talk to a standard battery module, Clements says, so a universal gateway is necessary. VARTA has already developed one.
“We can basically implement any protocol that the end customer needs,” he says, “and give him a completely integrated suite without having to deviate from the standard battery module and lose the advantages that he gets from having a standard module.”
It’s Possible if the Driverless Tractors Demonstrate Success
While it can be easy to get caught up in the futuristic fantasy of driverless tractors, Syngenta’s Chief Information and Data Officer Greg Meyers thinks the farming industry is already there.
“In many ways, agriculture is quite far ahead of consumers in this space,” he says. “I mean, tractors have
been steering themselves for 20 years, which is farther along than consumer vehicles are, and for the record, tractors with autosteer steer better than human beings do and with less effort.”
Fully autonomous tractors are on their way, but that doesn’t mean farmers will immediately embrace the technology. Many farmers, Meyers points out, use their parents’ and grandparents’ methods.
“Agriculture is a complicated endeavor from the moment you put the seed in the ground, and there's thousands of different things that could stop a seed’s ability to reach its full genetic potential,” Meyers says. “All those things require several hundreds, if not thousands of decisions that human beings have to make. In many cases, farming is still an art. It's something that relies a lot on intuition, judgment, and experience.”
Before the industry has access to a driverless tractor, Meyers continues, there will be many steps along the way. The progression is slow, but the autonomous features will continue to make the current equipment increasingly intelligent. These advancing technologies still have a lot to offer.
“Human beings are not good at managing a lot of variability, they’re good at averages,” Meyers says. “Farming today is still a lot of averages. You apply still pesticides, and you broadcast a similar thing across the whole field.
“But, in the long run, you'll be able to perceive relatively small differences in some plots of land and potentially be able to do something different plant by plant,” he continues. “That's not something a human being has the cognitive ability to do, but it is something that I think computers and data science and, ultimately, machinery, as they work together, can unlock. There will be a completely new opportunity to radically change the way agriculture is done.”
Regenerative farming, which is based on maximizing biodiversity in the soil and on a field through various practices, is one example. These practices are time-consuming. Someone who needs to spend a lot of time in a tractor each day won’t be able to commit to large-scale changes, even the kind of changes that often appeal to their customers.
“It's important to point out that farmers are very different than consumers because they only have 40 times in their lifetime to get it right and to make a living and to pass along the farm, so they're not going to take a lot of risk on things that are unproven,” Meyers says. “I think until there’s trust around how these algorithms work, and they can demonstrate success on that particular field, not just in demos and tests, but actually on their particular field, it will take a while for them to be willing to take the production risks associated with relying on computers and algorithms to make decisions for them.”
Regardless of their analytical capabilities, however, smart technologies remain imperfect. Mother Nature tends to throw a lot of variables at agricultural robots. Hardware that can effectively brave the elements and overall environment can sometimes be hard to find.
Meyers says these things take time, but some of the very repeatable tasks, like seeding and harvesting, benefit from mature mechanics and AI. Driverless tractors will follow suit. Even the safety factor is coming along.
“Just like in the 1920s, everybody was worried about cars and all the issues about having 800-pound machines moving through the streets and hurting people,” Meyers says, “but we navigated that, and we will navigate this as well.”