Agriculture and technology experts are trying to design a new regulatory framework for autonomous tractor adoption, but they worry that the needed data collection will be limited by California’s labor laws.
Industry needs to collect a large amount of data to support autonomous tractor adoption, says Michael Miiller, California Association of Winegrape Growers director of government affairs.
But a decades-old Cal/OSHA law banning driverless tomato harvesters could curtail trials from happening in California.
“We have the need to collect data under that law, and from that data we can update the law to reflect the new technology … that bridge is the part that I think is going to be the most interesting part,” Miiller said during the Occupational Safety and Health Standards Board’s inaugural Autonomous Agricultural Tractors Advisory Committee meeting.
Miichael Miiller (Agri-Pulse Photo)Cal/OSHA voted to establish a stakeholder committee in November, after years of rejecting petitions to change the autonomous agriculture ban. The 26-person committee includes representatives across labor, manufacturing, state agencies, universities and agriculture.
Committee members were tasked with scoping key issues ahead of their next meeting in May, after which they will draft preliminary recommendations for Cal/OSHA board consideration.
California Department of Pesticide Regulation senior environmental scientist Emily Bryson said state agencies would not have the bandwidth to filter the amount of data necessary to change the law.
She explained that since technology is outpacing government abilities, once regulators get their hands on available data, it may already be obsolete or close to it. And Yancy Yap, a senior safety engineer with Cal/OSHA, questioned how the safety of autonomous tractors could be practically reviewed if technology is sourced from multiple manufacturers.
Yap said that, from a regulator’s perspective, different sources of synthetic light detection and ranging, or LiDAR, and camera equipment will require them to consider how different systems are working together, even if they went through a variety of safety performance tests.
The “right-to-repair” issue is a contentious topic in agriculture. John Shutske, a University of Wisconsin professor and agricultural safety specialist, said right-to-repair would likely become a greater issue as the industry moves toward autonomy. But without clear examples, it is hard to say how the unintended consequences of mixed design systems would play out in the field.
It’s easy to be “in the know” about agriculture news from coast to coast! Sign up for a FREE month of Agri-Pulse news. Simply click here.
Shutske said the state’s complicated labor law will not allow for a simple solution.
He suggested data is just one layer of the puzzle, offering that the committee should be investing time in researching safety standards for design and safety. He said the International Organization for Standardization’s agriculture-specific standard — ISO 18497 — would be a good starting point.
The standard outlines agricultural machinery design concept principles and “deals with the significant hazards relevant to agricultural machinery and tractors” with different degrees of automation.
Warning that compliance standard literature can be complex, Shutske said safety experts down the line should do more to help small and midsized manufacturers understand and comply with standards, since most original equipment manufacturers have in-house teams dedicated to standards compliance.
He also pointed to the American Society of Agricultural and Biological Engineers, which has a separate catalog of standards.
“But it's an absolute necessity,” Shutske said, adding that inventors also need to be brought into the fold. “I have run into those situations where people are just astounded that there's this whole new body of data and information that can be used to assure safety in — or at least improve the level of safety in — the design process.” 
Bryan Little (Agri-Pulse Photo)
Bryan Little, senior director of policy advocacy at the California Farm Bureau, suggested that agriculture could leverage functional safety analysis lessons from other industries, which Shutske said has already been adapted in certain ISO standards for agriculture.
Shutske especially pressed that functional safety should be at the forefront for the entire stakeholder and design chain to fully think about the risk assessment process.
“We need to bring as many people to the table — designers, dealers, service people, you know, techs and all people, including the end users, including farm workers and farm employees,” said Shutske.
Shutske helped lead a series of Safety for Emerging Robotics and Autonomous Agriculture (SAFER AG) workshops in 2022 and 2024, an effort supported by the U.S. Department of Agriculture’s National Institute of Food and Agriculture.
During both years of SAFER AG, Shutske said participants posed the idea of federal agencies supporting research on autonomous data collection, on all factors: close calls, monitoring, detecting and storing data. He suggested this would allow California to be home to a federally funded effort and circumvent some of the committee’s concerns with data management.
The 2024 SAFER AG workshop homed in on risk management techniques, compliance standards and design guidelines. Shutske said there is a key difference between identifying the amount of time a piece of equipment went without incident versus understanding exactly how it avoided incident.
“It would be awesome if we could say, ‘What exactly was it that that machine saw during those operating hours?’” he said. “When I say ‘saw,’ a lot of these machines have cameras, they have LidDARs, they have other types of safety systems.”
He suggested manufacturers collect data on what happens in the field, what the technology encounters and how the technology responds to negative incidents versus successful encounters.
Miiller asked whether it would be an issue for California to use out-of-state trials to circumvent issues posed by the Cal/OSHA labor law, but Shutske said it is more a question about whether people trust data than if it was generated closer to home.
“There may be some data sets to kind of draw from, from around the country, from manufacturers and from partnerships with universities,” said Shutske.
He reminded the board that California is “not totally starting from scratch” and cited Carnegie Mellon’s research on automated tractor close calls and safety successes. The study was conducted on a Florida orange grove and confirmed the accuracy of autonomous tractor safety features like path tracking, detecting obstacles and self-monitoring.
Igino Cafiero, director of high value crops and autonomy at John Deere, noted that simulations could supplement the need for trialed data. A recent paper published by the company supports the use of LiDAR to train data across a variety of scenarios.
“Then we internally audit each of the other teams to make sure we've done it,” said Cafiero, comparing Deere’s process to data collected across many different real-world operations. “It's much more deterministic and much more planned out.”
And Tim Bucher, chief executive officer at Agtonomy, said the committee should look at what other states are doing to regulate autonomous vehicles. Agtonomy often operates out of state, Tim said, and suggested the committee aggregate a series of vehicle demos to show Cal/OSHA exactly what human-operated equipment looks like in the field.
At their next meeting, the advisory committee’s members plan to clarify data collection parameters and whether that could use proxy information. They also aim to identify potential issues with right-to-repair and equipment inspections.
For more news, go to Agri-pulse.com.
