A quick search of Google Scholar will reveal that the engineering literature is home to more papers on agricultural robots than ever. The majority are from China, which might seem strange given the historically low cost of labor there.

But times are changing: Foxconn, the electronics manufacturer famous for building Apple products as well as worker suicide, is moving to incorporate one million robots into its assembly lines in just three years. Wages are rising in China, as are the demands of workers. In an age in which increasingly complex tasks can be performed by semi-autonomous machines, robots have become the ultimate scabs.

Down on the farm, it’s no different. Industrial agriculture has already simplified the otherwise topographically complicated landscape a robot must navigate. Orchards, after all, have been planted in regular rows for centuries.

The same technology that allows for self-driving cars is driving this nascent revolution. These advances include optical range-finding technology like LIDAR, which can give robots the equivalent of human visual acuity and even better depth perception.

Another important ingredient, according to an abstract just submitted to the Automation and System Technology in Plant production conference of the Nordic Association of Agricultural Scientists, is the open-sourcing of robot control software and hardware.

Using FroboMind software, which runs on top of the open-source Robot Operating System from Willow Garage, researchers were able to drive the FroboBox, a straightforward Debin Linux-powered computer whose flexible outputs can be used to drive almost any kind of automate-able farm machinery.

In one particular application, the FroboBox drove an experimental ASuBot, which is kind of like a self-driving riding lawnmower. Here’s a video of it in action.

As we’ve seen in countless other areas of software development, the more the details of a system can be abstracted away through the use of robust, open-source components, the more time programmers and engineers have to actually build something that works. Just as Apache and HTML made the web possible, it appears that the Robot Operating System and platforms built with it are allowing researchers to move past theory and into the real world.

But at what point do robots become cheaper than migrant labor? That depends: in Denmark, where the FroboMind is being pioneered, labor costs are substantially higher than in the U.S. China has an incentive to move up the value chain in addition to dealing with its own rising wages, so, as in clean energy, its engineers have an interest in creating products for export. In any country, when capital is cheaper than labor – that’s when we’ll see devices like these roll out.