Safety and precision are important for robots, particularly in industries such as agriculture and construction. From mobile robots to drones and marine systems, Trimble Inc. provides technologies for positioning, modeling, connectivity, and data analysis.
The Westminster, Colo.-based company offers autonomous products and cloud services. Trimble yesterday closed on a $2 billion joint venture with AGCO Corp. to form PTx Trimble to combine Trimble’s precision-agriculture business with AGCO’s JCA Technologies unit for factory-fit and retrofit applications in mixed fleets.
Giri Baleri, product manager and marketing leader in autonomy at Trimble, spoke with Mobile Robot Guide about the company’s emphasis on positioning accuracy and safety for autonomous systems.
How did you come to Trimble?
Baleri: My bachelor’s is in mechanical engineering, and my first job was in inertial navigation for all things robotics and aviation. It was the early days of auto-steering in agriculture and the “glass cockpit” in aviation, as well as the DARPA Grand Challenge.
From there, I worked in machine controls and GNSS [global navigation satellite system] technologies, taking data from sensors to automatically steer farming vehicles. Then I worked at early-stage startups for remote imagery and predictive analytics for growers.
I’ve been with Trimble for three years now, with a core focus on positioning. We add components to other elements of full-stack autonomy to offer value. This includes semi-autonomous and fully autonomous vehicles, including our driverless sprayer. I’ve spent most of my career in customer-facing roles.
How have precision navigation, localization, and mapping improved in the past year?
Baleri: They’ve improved vastly in past five years — primarily GNSS — more than in the prior 30 years. There are a lot more new satellite constellations — BeiDou, Galileo, and more — to build more robust, available solutions with modern L5 and L6 frequencies, providing added resilience.
There’s also a lot of correction networks. Cellphones have sub-meter accuracy, but for vehicles, you need centimeter-level accuracy. You need correction signals for any interference, from 20,000 km [12,400 mi.] in the sky down to the vehicle on the street.
Trimble established its CenterPoint RTX network years ago, and now cellphone companies are trying to provide correction services. In addition, jamming and spoofing are now coming into play, as bad actors try to disrupt the market. This has all driven the need for greater autonomy.
For the past year, Trimble has been working on IonoGuard because the solar activity cycle also impacts reliability. We’re trying to predict such disruptions and take care of availability and accuracy.
As autonomous systems become more prolific, functional safety becomes more relevant. Trimble ProPoint has gone through a rigorous process and received certification.
Also, Trimble announced path-planning technology, which enables an automated machine going in a specific trajectory and speed in a real-world environment to know where it is in a safe and predictable manner.
Safety to be ‘table stakes,’ says Trimble
While safety is well-understood for mobile robots and autonomous vehicles, what requirements should developers and users keep in mind as use cases expand?
Baleri: Functional safety is critical for any machine control system. Unlike automotive, where the goal is Level 4 or 5 autonomy, safety standards are not as well-established. What you really need is ISO 26262. There has to be education on the part of the customers.
Off-road context is still necessary to make completely autonomous farm equipment, but we’re taking incremental steps to get there. In cars, you need to get from Point A to Point B, but that’s only an enabler for ag or construction, where a specific task or workflow is more important. If you have someone in a cab, autonomy can help offload one set of tasks so operators can focus on other things.
How can progressively automate steps toward the goal of getting the operator out of the vehicle? People still need to trust the technology and that it can do the task they’ve done for generations.
Eventually, [safe autonomy] will become table stakes. Standards differ across vertical markets, but there are a lot of similarities. A common-denominator approach provides price optimization opportunities. First, what’s the problem you’re trying to solve, and then you can call back to automotive standards.
Can you give some examples of how partners have applied Trimble’s navigation technology?
Baleri: Trimble used GNSS to automate solar pile-driving operations. It’s critical to make sure that posts are located and oriented accurately, and it’s an example of how we work with sensors.
Also, we worked with Yanmar, a large tractor company in Japan that used our technology to create an autonomous tractor using GNSS and orientation sensors.
In addition, we worked with Mitsui on an autonomous ferry in Japan to move between islands where finding labor was a big challenge.
On the development side, Trimble collaborated closely with European manufacturer HORSCH on GNSS, path planning, and machine controls, as well as complete autonomous control of a spray feeder. It can sense when to turn sprayers on and off.
We conducted an experiment at a dam-construction site in British Columbia in which manned and unmanned vehicles collaborated. The Dynapac soil compactor had autonomous speed, steering, path planning, and obstacle detection and avoidance. It could be used in a leader-follower situation, such as a dozer moving dirt and then working with the autonomous compactor.
Trimble builds a connected ecosystem for farming, construction
Autonomous vehicles came under renewed safety scrutiny last year. Can Trimble’s technology help address such challenges?
Baleri: For vehicle-to-vehicle coordination, we have the Connected Construction cloud, where all data from different machines goes. Users can then do strategic-level controls, such as sending a work order, but this was already done well before with manual vehicles.
Trimble is looking at task execution in Connected Farm. If we enable customers to build their own solutions, they could come into our ecosystem for coordination and workflow management to build them.
Why would customers build their own?
Baleri: From a systems integrator and an OEM perspective, companies like John Deere want to build their own systems tailored to specific markets, geographies, or customers.
But once you expand to multiple machines, how do you coordinate them? Then you need a platform like Connected Farm or Connected Construction.
What needs are similar across the industries that Trimble serves, and which are different? Is it a matter of range, indoor/outdoor infrastructure, or speed?
Baleri: There are similar key drivers – labor shortages, efficiency, and sustainability. The core tech building blocks are precision, perception, controls, and path planning.
However, the use cases, workflows, and operating design domains are different. For example, in automotive, a mound of dirt is an obstacle, but to a dozer, it’s a task.
Decisions vary quite a bit, and speeds are different. You don’t need the same level of accuracy for every application in agriculture, so you may not need RTK [real-time kinematics] or sub-decimeter accuracy.
But if you’re spraying, you don’t want to trample existing crops, so then you need centimeter accuracy. A harvester is different from a sprayer, and the operating environments can be different — dust can cover sensors — so designers need to overcome those obstacles.
Understanding the domain is part of why we partnered with AGCO, which we first announced in September. This is not the primary driver, but to better serve the farmers and OEMs together.
Cloud versus edge for data
Speaking of infrastructure, how much do current or soon-to-come systems rely on 5G versus onboard processing? What advances would you like to see in sensing and compute?
Baleri: There’s no doubt that we’d like to see every acre connected with 5G, but most major highways in the U.S. are well covered in 4G for automotive.
In the ag context, most often, tractors are farther out and are operating with no cellular coverage. It’s the same with remote construction sites such as solar farms in the California desert. Then you have no choice but to do edge computing.
Previously, the cost for processing was expensive, but that has changed quite a bit, with costs coming down for teraflops [trillion floating-point operations per second]. We’re at an inflection point, with the technology evolving and costs dropping.
For instance, lidar sensors that used to cost $100,000 and are now below $10,000 as automotive use scales up. That’s a huge decrease in cost and capabilities.
We’ve automated the largest mining equipment, and the complexity and cost of meeting the needs of early adopters results in lots of advancements. More connectivity is better, but there also ways to work around barriers. When a vehicle is in a barn, we can offload the data and send it to the cloud, and then assess how a job was done and make adjustments.
Agriculture, construction continue leading the drive to autonomy
In 2024, which industry do you expect to benefit most from advanced navigation, localization, and mapping?
Baleri: Automotive has had some setbacks. Companies made a lot of promises, but underestimated what it will take to get to Level 4 or 5.
Agriculture and construction are more looking for incremental progress. They’d like to get less-experienced operators but get them to be as good with technology. From my perspective of looking at industry needs, they’re both taking the lead.
Multiple robotics companies are trying point solutions, such as for strawberry picking. Some are electrifying or retrofitting tractors, and multiple startups are getting funding for autonomous plowing, weeding, picking, or harvesting. At the same time, the big tractor OEMs have acquired the likes of Bear Flag and Blue River.
We’ve worked with tractor retrofitters such as Sabanto to accelerate adoption of autonomy.
In construction, we’re seeing the need for getting more productivity with less input. It’s not about removing the operator, but providing progressive automation to make operators more efficient.
What are some applications where Trimble’s precision technology could be used more?
Baleri: We have a couple of customers in yard automation. Trimble is at the docking station, helping to make sure that the implement or trailer is aligned with the drive unit. That’s pretty critical to moving pallets around quickly, efficiently, and accurately.
We’re seeing quite a bit of interest in solar farm construction. Previous process needed three operators. Now with automation, they can get down to one for significant savings.
Another area is autonomous lawnmowers. There’s a lot of interest in automating commercial mowing, and Trimble is working with systems integrators and partners.
In addition, there’s not as much automation in specialty crops — orchards, vineyards, fruits and vegetables, particularly in California — so there’s more to be done. Lots of food gets wasted, rotting in the field or not making it to grocery stores in a timely manner. This requires more collaboration between the point of picking and storage and transport to grocery shelves.
AI can help build trust in automation
Interest in artificial intelligence has reached new heights. How is that helping navigation, and how else might it be applied to existing data?
Baleri: AI is a big focus for Trimble, and it plays a crucial role in developing autonomous systems. Navigation systems use data from different sensors to detect obstacles, recognize them, and plan trajectories.
GNSS is not guaranteed in all environments, such as next to a building, under a tree canopy, or in urban canyons. To improve localization and accuracy, we need sensor fusion and perception from lidar, radar, and cameras. IMUs [inertial measurement units] play a critical role in providing best estimates of position.
Trimble is in the space that’s constantly collecting large data sets, such as to build models of ionospheric effects on navigation. We were in agriculture and construction before autonomy, so we have a vast knowhow of how AI can help human intelligence by providing data insights to make better decisions faster.
The equipment, AI, and vast data sets are enablers. If technology can put something in historical context, then decision makers can focus on the strategic level rather than on dull information.
Autonomy is a marathon, not a sprint. We’re learning along the way, and we have to establish and build trust in the technology. If people see that it’s reliable, then they can be more comfortable with getting out of the cab and eventually watching at a distance.
