Earlier quoted context omitted.
Humans are currently too inexpensive to make a harvesting robot worth the cost (sadly enough). Let's say you spend $10M on R&D (a very low number in my opinion), each robot costs $50k to build, and maintenance/repairs cost $5k/year. You decide to charge your first customer $100k because you have to cover some of your R&D, and you pass all the maintenance costs on to them. For that $105k per year, a farmer could hire…
More variables: picking only ripe fruit; handling the fruit more carefully; picking more rapidly so you can share the robot with many growers; refrigerating the fruit from the moment of picking. Any of these could make it worthwhile?
For example an unmanned, refrigerated, follow-me cart with a soft hopper could be a multiplier on picking productivity without all the hassles of actual mobile manipulation and perception.