In agriculture, self-driving has been around and deployed for a decade. John Deere has 80 million acres under automated cultivation, growing at an incredible rate yearly. Want to talk about self-driving, the ag sector is way ahead of the game.
I tried to find a reference to this. Do you mean that John Deere sells/leases automated machinery to farmers who farm 80 million acres collectively, or that John Deere runs equipment on its own farms for testing cost offset by farming revenue?
The world and environment for a self driving tractor is so simple compared to any street in the world though. They can rely heavily on GPS for starters and the environment/fields are practically blank slates with no other vehicles to really contend with. It'd be surprising if they weren't ahead of the game considering all the advantages to navigating a wide open field with excellent GPS availability.
Not quite - a failure in automation for a tractor is measured in inches. If you end up driving over crops instead of between them for a pass, the farmer is likely out thousands of dollars.
On the highway, at least the obstacles are as big as cars and people. In a field, the obstacles can be a planted row that wasn't even sprouted above the ground yet.
Ultimately it's still driving a predetermined static path through a field with static obstacles with positioning provided by a modified GPS system that improved accuracy by providing additional information about the satellites positions. They've done really good work in their own space but very little of it can really be applied to self driving cars.
Further, they have other more complex constraints. "Swath control" means continuously adjusting the implement when crossing over already-treated ground (modulating spray nozzles etc to avoid double-treating).
Also there are many obstacles in a field -waterways, the occasional tree or post, fences etc. All has to be mapped and accounted for.
