Why grain and produce harvesting are worlds apart
Grain harvesting has been highly automated for decades, and it's worth understanding why that was achievable while so much fruit and vegetable harvesting still isn't. A combine harvester doesn't need to identify individual grains, judge their ripeness one at a time, or handle them delicately, it can cut and thresh an entire field's worth of a uniform crop in bulk, treating the whole plant as raw material to be processed rather than as a collection of individual, fragile objects to be picked out carefully. That's a bulk-processing problem, well suited to mechanical automation.
Delicate produce, like soft fruit, tree fruit, or many vegetables, is the opposite kind of problem. Each individual item has to be identified, evaluated, and handled as a distinct object, since damaging one strawberry or bruising one peach can make it unsellable in a way that damaging a single grain doesn't matter for a bulk grain harvest. This turns harvesting from a bulk-processing task into a repeated, high-precision manipulation task performed thousands of times per field, which is a much harder robotics problem, more similar in spirit to fine manipulation tasks than to industrial-scale mechanical processing.