Robots: How Close Are We?
Yesterday, the Wall Street Journal published an article on homegrown inventions that are advancing the technology in the field. Very much in line with my observations on tillage, it does seem like this type of tinkering and automation makes sense for the agriculture industry.
The article claims that “homespun hacking” (including tinkering, adding robotic parts, outfitting machinery with open source software, etc.) has become a trend in North America’s breadbasket. Furthering and innovating around technology makes sense. Technology has always been at the heart of agriculture, especially as many different farming operations have scaled up the number of acres they manage and cover. Adapting, updating, and finding ways to make labor more efficient all are highly relevant to succeeding in the agriculture business.
In a perfect world, we would imagine robots running around and conducting the mundane tasks efficiently, freeing up human labor for other purposes around the farm. Tractors would drive unmanned, up and down the field rows, and would coordinate with each other. The field would look more like an advanced manufacturing facility: a complex set of moving parts that ran like a well-oiled machine.
The venture capitalist in me dreams of the day where that perfect world becomes reality: where the latest and greatest computer vision, machine learning, robotics, and automation technology come together to transform the field into a manufacturing house. People often refer to agriculture as a “semi-structured” environment where there are consistencies (e.g., how far apart rows are planted) with some variability (e.g., weather conditions). As a result, it seems like a great place to push the boundaries of machine intelligence, forcing machines to adapt to new conditions and situations.
That being said, the last few days at Tom Farms have really shown me the challenges to making this dream a reality. I think I used to underestimate how non-trivial the “variability” in agriculture can be. It wasn’t until yesterday that I fully realized how often things do break and how unexpected these breaks can be.
It was a beautiful, sunny day in Indiana around 2:00 PM in the afternoon. I was on my way to CereServ to meet Mark, the semi-truck driver who would be delivering gallons of insecticide and herbicide to the sprayer. We loaded the tanks using a sophisticated pump system and some significant manual help. While the pumps functioned well, it still required a human operator to run the pumps, drop in the appropriate amount of anti-foam, drop in the appropriate amount of dry chemical, and log everything into the hand-written notebook. In any case, after about 20 minutes at CereServ, we were off to meet Matt at field 4050.
As we arrived to the field, Matt (who was operating the sprayer) was just filling up with the last tender truck. Everything seemed to be working fine as far as I could tell. However, I quickly learned that it was taking the team longer to re-fill the sprayer due a piece of faulty equipment on the sprayer itself. While not a huge hindrance to the operation, the issue did slow the process down a little bit. I already started to get a sense that not everything always works as planned.
I jumped into the cab with Matt, who introduced me to the complexity of the sprayer. With booms extending 120 feet wide and a slightly different setup than the cultivator tractor, the sprayer was perhaps even more impressive in terms of technology. In addition to the auto-steer function, the sprayer also monitors exactly what ground has been covered and is adjusted to avoid “double-spraying” automatically. It’s calibrated to deliver approximately 25 gallons of liquid / acre at a PSI averaging between 35–45. Moving at around 13 mph, the spraying process moves much more quickly than tillage. At this rate, we were planning on refilling the sprayer a few more times in that afternoon and finishing the field in no time.
After a number of passes, this seemed to be going smoothly and Matt graciously let me try my hand at driving the sprayer (probably a bad idea). Of course, it was only a simple back and forth across the field (e.g., no complex patterns or drawing field boundaries). We swapped seats. I looked down at my new steering wheel. It definitely was very foreign — nothing at all like driving a car.
I pushed the orange handle forward to move the tractor ahead, hit “1” and “2” to turn on the booms and auto-steer, and slowly but surely guided the vehicle down the row. There wasn’t much else to it other than monitoring the screens and preparing to turn at the end of the row.
As I approached the end of the row, Matt advised me to reduce my speed and turn around. Just as I was starting to re-align the vehicle with the auto-steer mark on the GPS, the screen disappeared and a “STOP” notice came on: “Hydraulic oil level low. Check oil level according to Operator’s Manual procedure.”
I thought to myself: “Great. One pass with me in the driver’s seat, and I already broke the equipment. I should never be allowed to drive.”
We stopped the vehicle and both hopped out. Matt took a look around back and realized where the pipe had broken on the right side of the boom. He told me that it wasn’t my fault and that these things were bound to break. (I didn’t believe him as he’s too nice to tell me the real truth about my terrible operating skills.) In any case, we called the semi-truck, hopped in, and went to go fetch the service vehicle.
At the end of yesterday afternoon, we had been expected to have completed 4050 and another field nearby. But clearly, that wasn’t going to happen. I think Matt spent the next few hours trouble-shooting before he got it up and running again. And as I just found out, the hydraulic pipe on the other side of the boom broke this morning. Things just don’t always seem to as planned.
This is the number one reason why farmers are skeptical about self-driving tractors. It’s not simply just denial or sadness at the prospect of not being able to ride and operate their own tractors. It’s that things break. And they break a lot. Everything from a major hydraulic leak to a piece of debris getting caught in a planter to a cultivator arm breaking while hitting a rock makes this notion of robots a bit more challenging. There sometimes just needs to be a person involved to troubleshoot and fix these issues, to remove debris from the field, or to know how to adjust the equipment when things aren’t quite working out as we had hoped.
Perhaps, for some of the more simple tasks, developing robots makes sense. But for some of the more complex tasks, we’re a long way from getting machine intelligence and machine durability to a point where it’s more effective and cheaper than the traditional use of labor. As a result, while I admire the tinkering and entrepreneurship from the farmers mentioned in the article, I also understand the skepticism coming from commercial scale farmers. There’s no doubt that the adoption of new technology is inevitably necessary — it just might need to be taken one step at a time.
Oh and I’ve learned my lesson. I think I’ll stick to “riding along” from now on…